From 025713b8f625fb388803c8ec9b21691c5848c07f Mon Sep 17 00:00:00 2001 From: xibici Date: Fri, 21 Jul 2023 16:13:56 +0900 Subject: [PATCH] first commit --- .dockerignore | 222 ++++ .gitattributes | 2 + .github/FUNDING.yml | 5 + .github/ISSUE_TEMPLATE/bug-report.yml | 85 ++ .github/ISSUE_TEMPLATE/config.yml | 8 + .github/ISSUE_TEMPLATE/feature-request.yml | 50 + .github/ISSUE_TEMPLATE/question.yml | 33 + .github/PULL_REQUEST_TEMPLATE.md | 7 + .github/dependabot.yml | 23 + .github/workflows/ci-testing.yml | 93 ++ .github/workflows/codeql-analysis.yml | 54 + .github/workflows/greetings.yml | 59 + .github/workflows/rebase.yml | 21 + .github/workflows/stale.yml | 38 + .gitignore | 256 ++++ .pre-commit-config.yaml | 66 + 2.4.1' | 0 CONTRIBUTING.md | 94 ++ Dockerfile | 64 + LICENSE | 674 ++++++++++ README.md | 306 +++++ data/Argoverse.yaml | 67 + data/GlobalWheat2020.yaml | 53 + data/Objects365.yaml | 112 ++ data/SKU-110K.yaml | 52 + data/VOC.yaml | 80 ++ data/VisDrone.yaml | 61 + data/coco.yaml | 44 + data/coco128.yaml | 30 + data/hand.yaml | 19 + data/hyps/hyp.Objects365.yaml | 34 + data/hyps/hyp.VOC.yaml | 40 + data/hyps/hyp.scratch-high.yaml | 34 + data/hyps/hyp.scratch-low.yaml | 34 + data/hyps/hyp.scratch-med.yaml | 34 + data/images/bus.jpg | Bin 0 -> 487438 bytes data/images/zidane.jpg | Bin 0 -> 168949 bytes data/scripts/download_weights.sh | 20 + data/scripts/get_coco.sh | 27 + data/scripts/get_coco128.sh | 17 + data/xView.yaml | 102 ++ detect.py | 260 ++++ export.py | 562 +++++++++ hubconf.py | 143 +++ models/__init__.py | 0 models/common.py | 677 ++++++++++ models/experimental.py | 120 ++ models/hub/anchors.yaml | 59 + models/hub/yolov3-spp.yaml | 51 + models/hub/yolov3-tiny.yaml | 41 + models/hub/yolov3.yaml | 51 + models/hub/yolov5-bifpn.yaml | 48 + models/hub/yolov5-fpn.yaml | 42 + models/hub/yolov5-p2.yaml | 54 + models/hub/yolov5-p34.yaml | 41 + models/hub/yolov5-p6.yaml | 56 + models/hub/yolov5-p7.yaml | 67 + models/hub/yolov5-panet.yaml | 48 + models/hub/yolov5l6.yaml | 60 + models/hub/yolov5m6.yaml | 60 + models/hub/yolov5n6.yaml | 60 + models/hub/yolov5s-ghost.yaml | 48 + models/hub/yolov5s-transformer.yaml | 48 + models/hub/yolov5s6.yaml | 60 + models/hub/yolov5x6.yaml | 60 + models/tf.py | 464 +++++++ models/yolo.py | 329 +++++ models/yolov5l.yaml | 48 + models/yolov5m.yaml | 48 + models/yolov5n.yaml | 48 + models/yolov5s.yaml | 48 + models/yolov5x.yaml | 48 + requirements.txt | 37 + setup.cfg | 45 + train.py | 644 ++++++++++ tutorial.ipynb | 1102 +++++++++++++++++ utils/__init__.py | 37 + utils/activations.py | 101 ++ utils/augmentations.py | 277 +++++ utils/autoanchor.py | 166 +++ utils/autobatch.py | 57 + utils/aws/__init__.py | 0 utils/aws/mime.sh | 26 + utils/aws/resume.py | 40 + utils/aws/userdata.sh | 27 + utils/benchmarks.py | 92 ++ utils/callbacks.py | 78 ++ utils/datasets.py | 1040 ++++++++++++++++ utils/downloads.py | 153 +++ utils/flask_rest_api/README.md | 73 ++ utils/flask_rest_api/example_request.py | 13 + utils/flask_rest_api/restapi.py | 37 + utils/general.py | 880 +++++++++++++ utils/google_app_engine/Dockerfile | 25 + .../additional_requirements.txt | 4 + utils/google_app_engine/app.yaml | 14 + utils/loggers/__init__.py | 168 +++ utils/loggers/wandb/README.md | 152 +++ utils/loggers/wandb/__init__.py | 0 utils/loggers/wandb/log_dataset.py | 27 + utils/loggers/wandb/sweep.py | 41 + utils/loggers/wandb/sweep.yaml | 143 +++ utils/loggers/wandb/wandb_utils.py | 562 +++++++++ utils/loss.py | 224 ++++ utils/metrics.py | 342 +++++ utils/plots.py | 471 +++++++ utils/torch_utils.py | 329 +++++ val.py | 383 ++++++ 108 files changed, 14579 insertions(+) create mode 100644 .dockerignore create mode 100644 .gitattributes create mode 100644 .github/FUNDING.yml create mode 100644 .github/ISSUE_TEMPLATE/bug-report.yml create mode 100644 .github/ISSUE_TEMPLATE/config.yml create mode 100644 .github/ISSUE_TEMPLATE/feature-request.yml create mode 100644 .github/ISSUE_TEMPLATE/question.yml create mode 100644 .github/PULL_REQUEST_TEMPLATE.md create mode 100644 .github/dependabot.yml create mode 100644 .github/workflows/ci-testing.yml create mode 100644 .github/workflows/codeql-analysis.yml create mode 100644 .github/workflows/greetings.yml create mode 100644 .github/workflows/rebase.yml create mode 100644 .github/workflows/stale.yml create mode 100644 .gitignore create mode 100644 .pre-commit-config.yaml create mode 100644 2.4.1' create mode 100644 CONTRIBUTING.md create mode 100644 Dockerfile create mode 100644 LICENSE create mode 100644 README.md create mode 100644 data/Argoverse.yaml create mode 100644 data/GlobalWheat2020.yaml create mode 100644 data/Objects365.yaml create mode 100644 data/SKU-110K.yaml create mode 100644 data/VOC.yaml create mode 100644 data/VisDrone.yaml create mode 100644 data/coco.yaml create mode 100644 data/coco128.yaml create mode 100644 data/hand.yaml create mode 100644 data/hyps/hyp.Objects365.yaml create mode 100644 data/hyps/hyp.VOC.yaml create mode 100644 data/hyps/hyp.scratch-high.yaml create mode 100644 data/hyps/hyp.scratch-low.yaml create mode 100644 data/hyps/hyp.scratch-med.yaml create mode 100644 data/images/bus.jpg create mode 100644 data/images/zidane.jpg create mode 100644 data/scripts/download_weights.sh create mode 100644 data/scripts/get_coco.sh create mode 100644 data/scripts/get_coco128.sh create mode 100644 data/xView.yaml create mode 100644 detect.py create mode 100644 export.py create mode 100644 hubconf.py create mode 100644 models/__init__.py create mode 100644 models/common.py create mode 100644 models/experimental.py create mode 100644 models/hub/anchors.yaml create mode 100644 models/hub/yolov3-spp.yaml create mode 100644 models/hub/yolov3-tiny.yaml create mode 100644 models/hub/yolov3.yaml create mode 100644 models/hub/yolov5-bifpn.yaml create mode 100644 models/hub/yolov5-fpn.yaml create mode 100644 models/hub/yolov5-p2.yaml create mode 100644 models/hub/yolov5-p34.yaml create mode 100644 models/hub/yolov5-p6.yaml create mode 100644 models/hub/yolov5-p7.yaml create mode 100644 models/hub/yolov5-panet.yaml create mode 100644 models/hub/yolov5l6.yaml create mode 100644 models/hub/yolov5m6.yaml create mode 100644 models/hub/yolov5n6.yaml create mode 100644 models/hub/yolov5s-ghost.yaml create mode 100644 models/hub/yolov5s-transformer.yaml create mode 100644 models/hub/yolov5s6.yaml create mode 100644 models/hub/yolov5x6.yaml create mode 100644 models/tf.py create mode 100644 models/yolo.py create mode 100644 models/yolov5l.yaml create mode 100644 models/yolov5m.yaml create mode 100644 models/yolov5n.yaml create mode 100644 models/yolov5s.yaml create mode 100644 models/yolov5x.yaml create mode 100644 requirements.txt create mode 100644 setup.cfg create mode 100644 train.py create mode 100644 tutorial.ipynb create mode 100644 utils/__init__.py create mode 100644 utils/activations.py create mode 100644 utils/augmentations.py create mode 100644 utils/autoanchor.py create mode 100644 utils/autobatch.py create mode 100644 utils/aws/__init__.py create mode 100644 utils/aws/mime.sh create mode 100644 utils/aws/resume.py create mode 100644 utils/aws/userdata.sh create mode 100644 utils/benchmarks.py create mode 100644 utils/callbacks.py create mode 100644 utils/datasets.py create mode 100644 utils/downloads.py create mode 100644 utils/flask_rest_api/README.md create mode 100644 utils/flask_rest_api/example_request.py create mode 100644 utils/flask_rest_api/restapi.py create mode 100644 utils/general.py create mode 100644 utils/google_app_engine/Dockerfile create mode 100644 utils/google_app_engine/additional_requirements.txt create mode 100644 utils/google_app_engine/app.yaml create mode 100644 utils/loggers/__init__.py create mode 100644 utils/loggers/wandb/README.md create mode 100644 utils/loggers/wandb/__init__.py create mode 100644 utils/loggers/wandb/log_dataset.py create mode 100644 utils/loggers/wandb/sweep.py create mode 100644 utils/loggers/wandb/sweep.yaml create mode 100644 utils/loggers/wandb/wandb_utils.py create mode 100644 utils/loss.py create mode 100644 utils/metrics.py create mode 100644 utils/plots.py create mode 100644 utils/torch_utils.py create mode 100644 val.py diff --git a/.dockerignore b/.dockerignore new file mode 100644 index 0000000..af51ccc --- /dev/null +++ b/.dockerignore @@ -0,0 +1,222 @@ +# Repo-specific DockerIgnore ------------------------------------------------------------------------------------------- +#.git +.cache +.idea +runs +output +coco +storage.googleapis.com + +data/samples/* +**/results*.csv +*.jpg + +# Neural Network weights ----------------------------------------------------------------------------------------------- +**/*.pt +**/*.pth +**/*.onnx +**/*.engine +**/*.mlmodel +**/*.torchscript +**/*.torchscript.pt +**/*.tflite +**/*.h5 +**/*.pb +*_saved_model/ +*_web_model/ +*_openvino_model/ + +# Below Copied From .gitignore ----------------------------------------------------------------------------------------- +# Below Copied From .gitignore ----------------------------------------------------------------------------------------- + + +# GitHub Python GitIgnore ---------------------------------------------------------------------------------------------- +# Byte-compiled / optimized / DLL files +__pycache__/ +*.py[cod] +*$py.class + +# C extensions +*.so + +# Distribution / packaging +.Python +env/ +build/ +develop-eggs/ +dist/ +downloads/ +eggs/ +.eggs/ +lib/ +lib64/ +parts/ +sdist/ +var/ +wheels/ +*.egg-info/ +wandb/ +.installed.cfg +*.egg + +# PyInstaller +# Usually these files are written by a python script from a template +# before PyInstaller builds the exe, so as to inject date/other infos into it. +*.manifest +*.spec + +# Installer logs +pip-log.txt +pip-delete-this-directory.txt + +# Unit test / coverage reports +htmlcov/ +.tox/ +.coverage +.coverage.* +.cache +nosetests.xml +coverage.xml +*.cover +.hypothesis/ + +# Translations +*.mo +*.pot + +# Django stuff: +*.log +local_settings.py + +# Flask stuff: +instance/ +.webassets-cache + +# Scrapy stuff: +.scrapy + +# Sphinx documentation +docs/_build/ + +# PyBuilder +target/ + +# Jupyter Notebook +.ipynb_checkpoints + +# pyenv +.python-version + +# celery beat schedule file +celerybeat-schedule + +# SageMath parsed files +*.sage.py + +# dotenv +.env + +# virtualenv +.venv* +venv*/ +ENV*/ + +# Spyder project settings +.spyderproject +.spyproject + +# Rope project settings +.ropeproject + +# mkdocs documentation +/site + +# mypy +.mypy_cache/ + + +# https://github.com/github/gitignore/blob/master/Global/macOS.gitignore ----------------------------------------------- + +# General +.DS_Store +.AppleDouble +.LSOverride + +# Icon must end with two \r +Icon +Icon? + +# Thumbnails +._* + +# Files that might appear in the root of a volume +.DocumentRevisions-V100 +.fseventsd +.Spotlight-V100 +.TemporaryItems +.Trashes +.VolumeIcon.icns +.com.apple.timemachine.donotpresent + +# Directories potentially created on remote AFP share +.AppleDB +.AppleDesktop +Network Trash Folder +Temporary Items +.apdisk + + +# https://github.com/github/gitignore/blob/master/Global/JetBrains.gitignore +# Covers JetBrains IDEs: IntelliJ, RubyMine, PhpStorm, AppCode, PyCharm, CLion, Android Studio and WebStorm +# Reference: https://intellij-support.jetbrains.com/hc/en-us/articles/206544839 + +# User-specific stuff: +.idea/* +.idea/**/workspace.xml +.idea/**/tasks.xml +.idea/dictionaries +.html # Bokeh Plots +.pg # TensorFlow Frozen Graphs +.avi # videos + +# Sensitive or high-churn files: +.idea/**/dataSources/ +.idea/**/dataSources.ids +.idea/**/dataSources.local.xml +.idea/**/sqlDataSources.xml +.idea/**/dynamic.xml +.idea/**/uiDesigner.xml + +# Gradle: +.idea/**/gradle.xml +.idea/**/libraries + +# CMake +cmake-build-debug/ +cmake-build-release/ + +# Mongo Explorer plugin: +.idea/**/mongoSettings.xml + +## File-based project format: +*.iws + +## Plugin-specific files: + +# IntelliJ +out/ + +# mpeltonen/sbt-idea plugin +.idea_modules/ + +# JIRA plugin +atlassian-ide-plugin.xml + +# Cursive Clojure plugin +.idea/replstate.xml + +# Crashlytics plugin (for Android Studio and IntelliJ) +com_crashlytics_export_strings.xml +crashlytics.properties +crashlytics-build.properties +fabric.properties diff --git a/.gitattributes b/.gitattributes new file mode 100644 index 0000000..dad4239 --- /dev/null +++ b/.gitattributes @@ -0,0 +1,2 @@ +# this drop notebooks from GitHub language stats +*.ipynb linguist-vendored diff --git a/.github/FUNDING.yml b/.github/FUNDING.yml new file mode 100644 index 0000000..3da386f --- /dev/null +++ b/.github/FUNDING.yml @@ -0,0 +1,5 @@ +# These are supported funding model platforms + +github: glenn-jocher +patreon: ultralytics +open_collective: ultralytics diff --git a/.github/ISSUE_TEMPLATE/bug-report.yml b/.github/ISSUE_TEMPLATE/bug-report.yml new file mode 100644 index 0000000..fcb6413 --- /dev/null +++ b/.github/ISSUE_TEMPLATE/bug-report.yml @@ -0,0 +1,85 @@ +name: 🐛 Bug Report +# title: " " +description: Problems with YOLOv5 +labels: [bug, triage] +body: + - type: markdown + attributes: + value: | + Thank you for submitting a YOLOv5 🐛 Bug Report! + + - type: checkboxes + attributes: + label: Search before asking + description: > + Please search the [issues](https://github.com/ultralytics/yolov5/issues) to see if a similar bug report already exists. + options: + - label: > + I have searched the YOLOv5 [issues](https://github.com/ultralytics/yolov5/issues) and found no similar bug report. + required: true + + - type: dropdown + attributes: + label: YOLOv5 Component + description: | + Please select the part of YOLOv5 where you found the bug. + multiple: true + options: + - "Training" + - "Validation" + - "Detection" + - "Export" + - "PyTorch Hub" + - "Multi-GPU" + - "Evolution" + - "Integrations" + - "Other" + validations: + required: false + + - type: textarea + attributes: + label: Bug + description: Provide console output with error messages and/or screenshots of the bug. + placeholder: | + 💡 ProTip! Include as much information as possible (screenshots, logs, tracebacks etc.) to receive the most helpful response. + validations: + required: true + + - type: textarea + attributes: + label: Environment + description: Please specify the software and hardware you used to produce the bug. + placeholder: | + - YOLO: YOLOv5 🚀 v6.0-67-g60e42e1 torch 1.9.0+cu111 CUDA:0 (A100-SXM4-40GB, 40536MiB) + - OS: Ubuntu 20.04 + - Python: 3.9.0 + validations: + required: false + + - type: textarea + attributes: + label: Minimal Reproducible Example + description: > + When asking a question, people will be better able to provide help if you provide code that they can easily understand and use to **reproduce** the problem. + This is referred to by community members as creating a [minimal reproducible example](https://stackoverflow.com/help/minimal-reproducible-example). + placeholder: | + ``` + # Code to reproduce your issue here + ``` + validations: + required: false + + - type: textarea + attributes: + label: Additional + description: Anything else you would like to share? + + - type: checkboxes + attributes: + label: Are you willing to submit a PR? + description: > + (Optional) We encourage you to submit a [Pull Request](https://github.com/ultralytics/yolov5/pulls) (PR) to help improve YOLOv5 for everyone, especially if you have a good understanding of how to implement a fix or feature. + See the YOLOv5 [Contributing Guide](https://github.com/ultralytics/yolov5/blob/master/CONTRIBUTING.md) to get started. + options: + - label: Yes I'd like to help by submitting a PR! diff --git a/.github/ISSUE_TEMPLATE/config.yml b/.github/ISSUE_TEMPLATE/config.yml new file mode 100644 index 0000000..f388d7b --- /dev/null +++ b/.github/ISSUE_TEMPLATE/config.yml @@ -0,0 +1,8 @@ +blank_issues_enabled: true +contact_links: + - name: Slack + url: https://join.slack.com/t/ultralytics/shared_invite/zt-w29ei8bp-jczz7QYUmDtgo6r6KcMIAg + about: Ask on Ultralytics Slack Forum + - name: Stack Overflow + url: https://stackoverflow.com/search?q=YOLOv5 + about: Ask on Stack Overflow with 'YOLOv5' tag diff --git a/.github/ISSUE_TEMPLATE/feature-request.yml b/.github/ISSUE_TEMPLATE/feature-request.yml new file mode 100644 index 0000000..68ef985 --- /dev/null +++ b/.github/ISSUE_TEMPLATE/feature-request.yml @@ -0,0 +1,50 @@ +name: 🚀 Feature Request +description: Suggest a YOLOv5 idea +# title: " " +labels: [enhancement] +body: + - type: markdown + attributes: + value: | + Thank you for submitting a YOLOv5 🚀 Feature Request! + + - type: checkboxes + attributes: + label: Search before asking + description: > + Please search the [issues](https://github.com/ultralytics/yolov5/issues) to see if a similar feature request already exists. + options: + - label: > + I have searched the YOLOv5 [issues](https://github.com/ultralytics/yolov5/issues) and found no similar feature requests. + required: true + + - type: textarea + attributes: + label: Description + description: A short description of your feature. + placeholder: | + What new feature would you like to see in YOLOv5? + validations: + required: true + + - type: textarea + attributes: + label: Use case + description: | + Describe the use case of your feature request. It will help us understand and prioritize the feature request. + placeholder: | + How would this feature be used, and who would use it? + + - type: textarea + attributes: + label: Additional + description: Anything else you would like to share? + + - type: checkboxes + attributes: + label: Are you willing to submit a PR? + description: > + (Optional) We encourage you to submit a [Pull Request](https://github.com/ultralytics/yolov5/pulls) (PR) to help improve YOLOv5 for everyone, especially if you have a good understanding of how to implement a fix or feature. + See the YOLOv5 [Contributing Guide](https://github.com/ultralytics/yolov5/blob/master/CONTRIBUTING.md) to get started. + options: + - label: Yes I'd like to help by submitting a PR! diff --git a/.github/ISSUE_TEMPLATE/question.yml b/.github/ISSUE_TEMPLATE/question.yml new file mode 100644 index 0000000..8e0993c --- /dev/null +++ b/.github/ISSUE_TEMPLATE/question.yml @@ -0,0 +1,33 @@ +name: ❓ Question +description: Ask a YOLOv5 question +# title: " " +labels: [question] +body: + - type: markdown + attributes: + value: | + Thank you for asking a YOLOv5 ❓ Question! + + - type: checkboxes + attributes: + label: Search before asking + description: > + Please search the [issues](https://github.com/ultralytics/yolov5/issues) and [discussions](https://github.com/ultralytics/yolov5/discussions) to see if a similar question already exists. + options: + - label: > + I have searched the YOLOv5 [issues](https://github.com/ultralytics/yolov5/issues) and [discussions](https://github.com/ultralytics/yolov5/discussions) and found no similar questions. + required: true + + - type: textarea + attributes: + label: Question + description: What is your question? + placeholder: | + 💡 ProTip! Include as much information as possible (screenshots, logs, tracebacks etc.) to receive the most helpful response. + validations: + required: true + + - type: textarea + attributes: + label: Additional + description: Anything else you would like to share? diff --git a/.github/PULL_REQUEST_TEMPLATE.md b/.github/PULL_REQUEST_TEMPLATE.md new file mode 100644 index 0000000..7a3e1b7 --- /dev/null +++ b/.github/PULL_REQUEST_TEMPLATE.md @@ -0,0 +1,7 @@ +Thank you for submitting a YOLOv5 🚀 Pull Request! We want to make contributing to YOLOv5 as easy and transparent as possible. A few tips to get you started: + +- Search existing YOLOv5 [PRs](https://github.com/ultralytics/yolov5/pull) to see if a similar PR already exists. +- Link this PR to a YOLOv5 [issue](https://github.com/ultralytics/yolov5/issues) to help us understand what bug fix or feature is being implemented. +- Provide before and after profiling/inference/training results to help us quantify the improvement your PR provides (if applicable). + +Please see our ✅ [Contributing Guide](https://github.com/ultralytics/yolov5/blob/master/CONTRIBUTING.md) for more details. diff --git a/.github/dependabot.yml b/.github/dependabot.yml new file mode 100644 index 0000000..c1b3d5d --- /dev/null +++ b/.github/dependabot.yml @@ -0,0 +1,23 @@ +version: 2 +updates: + - package-ecosystem: pip + directory: "/" + schedule: + interval: weekly + time: "04:00" + open-pull-requests-limit: 10 + reviewers: + - glenn-jocher + labels: + - dependencies + + - package-ecosystem: github-actions + directory: "/" + schedule: + interval: weekly + time: "04:00" + open-pull-requests-limit: 5 + reviewers: + - glenn-jocher + labels: + - dependencies diff --git a/.github/workflows/ci-testing.yml b/.github/workflows/ci-testing.yml new file mode 100644 index 0000000..5cf1613 --- /dev/null +++ b/.github/workflows/ci-testing.yml @@ -0,0 +1,93 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +name: CI CPU testing + +on: # https://help.github.com/en/actions/reference/events-that-trigger-workflows + push: + branches: [ master ] + pull_request: + # The branches below must be a subset of the branches above + branches: [ master ] + schedule: + - cron: '0 0 * * *' # Runs at 00:00 UTC every day + +jobs: + cpu-tests: + + runs-on: ${{ matrix.os }} + strategy: + fail-fast: false + matrix: + os: [ ubuntu-latest, macos-latest, windows-latest ] + python-version: [ 3.9 ] + model: [ 'yolov5n' ] # models to test + + # Timeout: https://stackoverflow.com/a/59076067/4521646 + timeout-minutes: 60 + steps: + - uses: actions/checkout@v2 + - name: Set up Python ${{ matrix.python-version }} + uses: actions/setup-python@v2 + with: + python-version: ${{ matrix.python-version }} + + # Note: This uses an internal pip API and may not always work + # https://github.com/actions/cache/blob/master/examples.md#multiple-oss-in-a-workflow + - name: Get pip cache + id: pip-cache + run: | + python -c "from pip._internal.locations import USER_CACHE_DIR; print('::set-output name=dir::' + USER_CACHE_DIR)" + + - name: Cache pip + uses: actions/cache@v2.1.7 + with: + path: ${{ steps.pip-cache.outputs.dir }} + key: ${{ runner.os }}-${{ matrix.python-version }}-pip-${{ hashFiles('requirements.txt') }} + restore-keys: | + ${{ runner.os }}-${{ matrix.python-version }}-pip- + + # Known Keras 2.7.0 issue: https://github.com/ultralytics/yolov5/pull/5486 + - name: Install dependencies + run: | + python -m pip install --upgrade pip + pip install -qr requirements.txt -f https://download.pytorch.org/whl/cpu/torch_stable.html + pip install -q onnx tensorflow-cpu keras==2.6.0 # wandb # extras + python --version + pip --version + pip list + shell: bash + + # - name: W&B login + # run: wandb login 345011b3fb26dc8337fd9b20e53857c1d403f2aa + + # - name: Download data + # run: | + # curl -L -o tmp.zip https://github.com/ultralytics/yolov5/releases/download/v1.0/coco128.zip + # unzip -q tmp.zip -d ../datasets + + - name: Tests workflow + run: | + # export PYTHONPATH="$PWD" # to run '$ python *.py' files in subdirectories + d=cpu # device + weights=runs/train/exp/weights/best.pt + + # Train + python train.py --img 64 --batch 32 --weights ${{ matrix.model }}.pt --cfg ${{ matrix.model }}.yaml --epochs 1 --device $d + # Val + python val.py --img 64 --batch 32 --weights ${{ matrix.model }}.pt --device $d + python val.py --img 64 --batch 32 --weights $weights --device $d + # Detect + python detect.py --weights ${{ matrix.model }}.pt --device $d + python detect.py --weights $weights --device $d + python hubconf.py # hub + # Export + python models/yolo.py --cfg ${{ matrix.model }}.yaml # build PyTorch model + python models/tf.py --weights ${{ matrix.model }}.pt # build TensorFlow model + python export.py --weights ${{ matrix.model }}.pt --img 64 --include torchscript onnx # export + # Python + python - <=3.7.0**](https://www.python.org/) with all [requirements.txt](https://github.com/ultralytics/yolov5/blob/master/requirements.txt) installed including [**PyTorch>=1.7**](https://pytorch.org/get-started/locally/). To get started: + ```bash + git clone https://github.com/ultralytics/yolov5 # clone + cd yolov5 + pip install -r requirements.txt # install + ``` + + ## Environments + + YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including [CUDA](https://developer.nvidia.com/cuda)/[CUDNN](https://developer.nvidia.com/cudnn), [Python](https://www.python.org/) and [PyTorch](https://pytorch.org/) preinstalled): + + - **Google Colab and Kaggle** notebooks with free GPU: Open In Colab Open In Kaggle + - **Google Cloud** Deep Learning VM. See [GCP Quickstart Guide](https://github.com/ultralytics/yolov5/wiki/GCP-Quickstart) + - **Amazon** Deep Learning AMI. See [AWS Quickstart Guide](https://github.com/ultralytics/yolov5/wiki/AWS-Quickstart) + - **Docker Image**. See [Docker Quickstart Guide](https://github.com/ultralytics/yolov5/wiki/Docker-Quickstart) Docker Pulls + + + ## Status + + CI CPU testing + + If this badge is green, all [YOLOv5 GitHub Actions](https://github.com/ultralytics/yolov5/actions) Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training ([train.py](https://github.com/ultralytics/yolov5/blob/master/train.py)), validation ([val.py](https://github.com/ultralytics/yolov5/blob/master/val.py)), inference ([detect.py](https://github.com/ultralytics/yolov5/blob/master/detect.py)) and export ([export.py](https://github.com/ultralytics/yolov5/blob/master/export.py)) on MacOS, Windows, and Ubuntu every 24 hours and on every commit. diff --git a/.github/workflows/rebase.yml b/.github/workflows/rebase.yml new file mode 100644 index 0000000..a4db1ef --- /dev/null +++ b/.github/workflows/rebase.yml @@ -0,0 +1,21 @@ +# https://github.com/marketplace/actions/automatic-rebase + +name: Automatic Rebase +on: + issue_comment: + types: [created] +jobs: + rebase: + name: Rebase + if: github.event.issue.pull_request != '' && contains(github.event.comment.body, '/rebase') + runs-on: ubuntu-latest + steps: + - name: Checkout the latest code + uses: actions/checkout@v2 + with: + token: ${{ secrets.ACTIONS_TOKEN }} + fetch-depth: 0 # otherwise, you will fail to push refs to dest repo + - name: Automatic Rebase + uses: cirrus-actions/rebase@1.5 + env: + GITHUB_TOKEN: ${{ secrets.ACTIONS_TOKEN }} diff --git a/.github/workflows/stale.yml b/.github/workflows/stale.yml new file mode 100644 index 0000000..7a83950 --- /dev/null +++ b/.github/workflows/stale.yml @@ -0,0 +1,38 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +name: Close stale issues +on: + schedule: + - cron: '0 0 * * *' # Runs at 00:00 UTC every day + +jobs: + stale: + runs-on: ubuntu-latest + steps: + - uses: actions/stale@v4 + with: + repo-token: ${{ secrets.GITHUB_TOKEN }} + stale-issue-message: | + 👋 Hello, this issue has been automatically marked as stale because it has not had recent activity. Please note it will be closed if no further activity occurs. + + Access additional [YOLOv5](https://ultralytics.com/yolov5) 🚀 resources: + - **Wiki** – https://github.com/ultralytics/yolov5/wiki + - **Tutorials** – https://github.com/ultralytics/yolov5#tutorials + - **Docs** – https://docs.ultralytics.com + + Access additional [Ultralytics](https://ultralytics.com) ⚡ resources: + - **Ultralytics HUB** – https://ultralytics.com/hub + - **Vision API** – https://ultralytics.com/yolov5 + - **About Us** – https://ultralytics.com/about + - **Join Our Team** – https://ultralytics.com/work + - **Contact Us** – https://ultralytics.com/contact + + Feel free to inform us of any other **issues** you discover or **feature requests** that come to mind in the future. Pull Requests (PRs) are also always welcomed! + + Thank you for your contributions to YOLOv5 🚀 and Vision AI ⭐! + + stale-pr-message: 'This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions YOLOv5 🚀 and Vision AI ⭐.' + days-before-stale: 30 + days-before-close: 5 + exempt-issue-labels: 'documentation,tutorial,TODO' + operations-per-run: 100 # The maximum number of operations per run, used to control rate limiting. diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..69a0084 --- /dev/null +++ b/.gitignore @@ -0,0 +1,256 @@ +# Repo-specific GitIgnore ---------------------------------------------------------------------------------------------- +*.jpg +*.jpeg +*.png +*.bmp +*.tif +*.tiff +*.heic +*.JPG +*.JPEG +*.PNG +*.BMP +*.TIF +*.TIFF +*.HEIC +*.mp4 +*.mov +*.MOV +*.avi +*.data +*.json +*.cfg +!setup.cfg +!cfg/yolov3*.cfg + +storage.googleapis.com +runs/* +data/* +data/images/* +!data/*.yaml +!data/hyps +!data/scripts +!data/images +!data/images/zidane.jpg +!data/images/bus.jpg +!data/*.sh + +results*.csv + +# Datasets ------------------------------------------------------------------------------------------------------------- +coco/ +coco128/ +VOC/ + +# MATLAB GitIgnore ----------------------------------------------------------------------------------------------------- +*.m~ +*.mat +!targets*.mat + +# Neural Network weights ----------------------------------------------------------------------------------------------- +*.weights +*.pt +*.pb +*.onnx +*.engine +*.mlmodel +*.torchscript +*.tflite +*.h5 +*_saved_model/ +*_web_model/ +*_openvino_model/ +darknet53.conv.74 +yolov3-tiny.conv.15 + +# GitHub Python GitIgnore ---------------------------------------------------------------------------------------------- +# Byte-compiled / optimized / DLL files +__pycache__/ +*.py[cod] +*$py.class + +# C extensions +*.so + +# Distribution / packaging +.Python +env/ +build/ +develop-eggs/ +dist/ +downloads/ +eggs/ +.eggs/ +lib/ +lib64/ +parts/ +sdist/ +var/ +wheels/ +*.egg-info/ +/wandb/ +.installed.cfg +*.egg + + +# PyInstaller +# Usually these files are written by a python script from a template +# before PyInstaller builds the exe, so as to inject date/other infos into it. +*.manifest +*.spec + +# Installer logs +pip-log.txt +pip-delete-this-directory.txt + +# Unit test / coverage reports +htmlcov/ +.tox/ +.coverage +.coverage.* +.cache +nosetests.xml +coverage.xml +*.cover +.hypothesis/ + +# Translations +*.mo +*.pot + +# Django stuff: +*.log +local_settings.py + +# Flask stuff: +instance/ +.webassets-cache + +# Scrapy stuff: +.scrapy + +# Sphinx documentation +docs/_build/ + +# PyBuilder +target/ + +# Jupyter Notebook +.ipynb_checkpoints + +# pyenv +.python-version + +# celery beat schedule file +celerybeat-schedule + +# SageMath parsed files +*.sage.py + +# dotenv +.env + +# virtualenv +.venv* +venv*/ +ENV*/ + +# Spyder project settings +.spyderproject +.spyproject + +# Rope project settings +.ropeproject + +# mkdocs documentation +/site + +# mypy +.mypy_cache/ + + +# https://github.com/github/gitignore/blob/master/Global/macOS.gitignore ----------------------------------------------- + +# General +.DS_Store +.AppleDouble +.LSOverride + +# Icon must end with two \r +Icon +Icon? + +# Thumbnails +._* + +# Files that might appear in the root of a volume +.DocumentRevisions-V100 +.fseventsd +.Spotlight-V100 +.TemporaryItems +.Trashes +.VolumeIcon.icns +.com.apple.timemachine.donotpresent + +# Directories potentially created on remote AFP share +.AppleDB +.AppleDesktop +Network Trash Folder +Temporary Items +.apdisk + + +# https://github.com/github/gitignore/blob/master/Global/JetBrains.gitignore +# Covers JetBrains IDEs: IntelliJ, RubyMine, PhpStorm, AppCode, PyCharm, CLion, Android Studio and WebStorm +# Reference: https://intellij-support.jetbrains.com/hc/en-us/articles/206544839 + +# User-specific stuff: +.idea/* +.idea/**/workspace.xml +.idea/**/tasks.xml +.idea/dictionaries +.html # Bokeh Plots +.pg # TensorFlow Frozen Graphs +.avi # videos + +# Sensitive or high-churn files: +.idea/**/dataSources/ +.idea/**/dataSources.ids +.idea/**/dataSources.local.xml +.idea/**/sqlDataSources.xml +.idea/**/dynamic.xml +.idea/**/uiDesigner.xml + +# Gradle: +.idea/**/gradle.xml +.idea/**/libraries + +# CMake +cmake-build-debug/ +cmake-build-release/ + +# Mongo Explorer plugin: +.idea/**/mongoSettings.xml + +## File-based project format: +*.iws + +## Plugin-specific files: + +# IntelliJ +out/ + +# mpeltonen/sbt-idea plugin +.idea_modules/ + +# JIRA plugin +atlassian-ide-plugin.xml + +# Cursive Clojure plugin +.idea/replstate.xml + +# Crashlytics plugin (for Android Studio and IntelliJ) +com_crashlytics_export_strings.xml +crashlytics.properties +crashlytics-build.properties +fabric.properties diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml new file mode 100644 index 0000000..526a560 --- /dev/null +++ b/.pre-commit-config.yaml @@ -0,0 +1,66 @@ +# Define hooks for code formations +# Will be applied on any updated commit files if a user has installed and linked commit hook + +default_language_version: + python: python3.8 + +# Define bot property if installed via https://github.com/marketplace/pre-commit-ci +ci: + autofix_prs: true + autoupdate_commit_msg: '[pre-commit.ci] pre-commit suggestions' + autoupdate_schedule: quarterly + # submodules: true + +repos: + - repo: https://github.com/pre-commit/pre-commit-hooks + rev: v4.1.0 + hooks: + - id: end-of-file-fixer + - id: trailing-whitespace + - id: check-case-conflict + - id: check-yaml + - id: check-toml + - id: pretty-format-json + - id: check-docstring-first + + - repo: https://github.com/asottile/pyupgrade + rev: v2.31.0 + hooks: + - id: pyupgrade + args: [--py36-plus] + name: Upgrade code + + - repo: https://github.com/PyCQA/isort + rev: 5.10.1 + hooks: + - id: isort + name: Sort imports + + # TODO + #- repo: https://github.com/pre-commit/mirrors-yapf + # rev: v0.31.0 + # hooks: + # - id: yapf + # name: formatting + + # TODO + #- repo: https://github.com/executablebooks/mdformat + # rev: 0.7.7 + # hooks: + # - id: mdformat + # additional_dependencies: + # - mdformat-gfm + # - mdformat-black + # - mdformat_frontmatter + + # TODO + #- repo: https://github.com/asottile/yesqa + # rev: v1.2.3 + # hooks: + # - id: yesqa + + - repo: https://github.com/PyCQA/flake8 + rev: 4.0.1 + hooks: + - id: flake8 + name: PEP8 diff --git a/2.4.1' b/2.4.1' new file mode 100644 index 0000000..e69de29 diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md new file mode 100644 index 0000000..ebde03a --- /dev/null +++ b/CONTRIBUTING.md @@ -0,0 +1,94 @@ +## Contributing to YOLOv5 🚀 + +We love your input! We want to make contributing to YOLOv5 as easy and transparent as possible, whether it's: + +- Reporting a bug +- Discussing the current state of the code +- Submitting a fix +- Proposing a new feature +- Becoming a maintainer + +YOLOv5 works so well due to our combined community effort, and for every small improvement you contribute you will be +helping push the frontiers of what's possible in AI 😃! + +## Submitting a Pull Request (PR) 🛠ī¸ + +Submitting a PR is easy! This example shows how to submit a PR for updating `requirements.txt` in 4 steps: + +### 1. Select File to Update + +Select `requirements.txt` to update by clicking on it in GitHub. +

PR_step1

+ +### 2. Click 'Edit this file' + +Button is in top-right corner. +

PR_step2

+ +### 3. Make Changes + +Change `matplotlib` version from `3.2.2` to `3.3`. +

PR_step3

+ +### 4. Preview Changes and Submit PR + +Click on the **Preview changes** tab to verify your updates. At the bottom of the screen select 'Create a **new branch** +for this commit', assign your branch a descriptive name such as `fix/matplotlib_version` and click the green **Propose +changes** button. All done, your PR is now submitted to YOLOv5 for review and approval 😃! +

PR_step4

+ +### PR recommendations + +To allow your work to be integrated as seamlessly as possible, we advise you to: + +- ✅ Verify your PR is **up-to-date with upstream/master.** If your PR is behind upstream/master an + automatic [GitHub Actions](https://github.com/ultralytics/yolov5/blob/master/.github/workflows/rebase.yml) merge may + be attempted by writing /rebase in a new comment, or by running the following code, replacing 'feature' with the name + of your local branch: + +```bash +git remote add upstream https://github.com/ultralytics/yolov5.git +git fetch upstream +# git checkout feature # <--- replace 'feature' with local branch name +git merge upstream/master +git push -u origin -f +``` + +- ✅ Verify all Continuous Integration (CI) **checks are passing**. +- ✅ Reduce changes to the absolute **minimum** required for your bug fix or feature addition. _"It is not daily increase + but daily decrease, hack away the unessential. The closer to the source, the less wastage there is."_ — Bruce Lee + +## Submitting a Bug Report 🐛 + +If you spot a problem with YOLOv5 please submit a Bug Report! + +For us to start investigating a possible problem we need to be able to reproduce it ourselves first. We've created a few +short guidelines below to help users provide what we need in order to get started. + +When asking a question, people will be better able to provide help if you provide **code** that they can easily +understand and use to **reproduce** the problem. This is referred to by community members as creating +a [minimum reproducible example](https://stackoverflow.com/help/minimal-reproducible-example). Your code that reproduces +the problem should be: + +* ✅ **Minimal** – Use as little code as possible that still produces the same problem +* ✅ **Complete** – Provide **all** parts someone else needs to reproduce your problem in the question itself +* ✅ **Reproducible** – Test the code you're about to provide to make sure it reproduces the problem + +In addition to the above requirements, for [Ultralytics](https://ultralytics.com/) to provide assistance your code +should be: + +* ✅ **Current** – Verify that your code is up-to-date with current + GitHub [master](https://github.com/ultralytics/yolov5/tree/master), and if necessary `git pull` or `git clone` a new + copy to ensure your problem has not already been resolved by previous commits. +* ✅ **Unmodified** – Your problem must be reproducible without any modifications to the codebase in this + repository. [Ultralytics](https://ultralytics.com/) does not provide support for custom code ⚠ī¸. + +If you believe your problem meets all of the above criteria, please close this issue and raise a new one using the 🐛 ** +Bug Report** [template](https://github.com/ultralytics/yolov5/issues/new/choose) and providing +a [minimum reproducible example](https://stackoverflow.com/help/minimal-reproducible-example) to help us better +understand and diagnose your problem. + +## License + +By contributing, you agree that your contributions will be licensed under +the [GPL-3.0 license](https://choosealicense.com/licenses/gpl-3.0/) diff --git a/Dockerfile b/Dockerfile new file mode 100644 index 0000000..489dd04 --- /dev/null +++ b/Dockerfile @@ -0,0 +1,64 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Start FROM Nvidia PyTorch image https://ngc.nvidia.com/catalog/containers/nvidia:pytorch +FROM nvcr.io/nvidia/pytorch:21.10-py3 + +# Install linux packages +RUN apt update && apt install -y zip htop screen libgl1-mesa-glx + +# Install python dependencies +COPY requirements.txt . +RUN python -m pip install --upgrade pip +RUN pip uninstall -y torch torchvision torchtext +RUN pip install --no-cache -r requirements.txt albumentations wandb gsutil notebook \ + torch==1.10.2+cu113 torchvision==0.11.3+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html +# RUN pip install --no-cache -U torch torchvision + +# Create working directory +RUN mkdir -p /usr/src/app +WORKDIR /usr/src/app + +# Copy contents +COPY . /usr/src/app + +# Downloads to user config dir +ADD https://ultralytics.com/assets/Arial.ttf /root/.config/Ultralytics/ + +# Set environment variables +# ENV HOME=/usr/src/app + + +# Usage Examples ------------------------------------------------------------------------------------------------------- + +# Build and Push +# t=ultralytics/yolov5:latest && sudo docker build -t $t . && sudo docker push $t + +# Pull and Run +# t=ultralytics/yolov5:latest && sudo docker pull $t && sudo docker run -it --ipc=host --gpus all $t + +# Pull and Run with local directory access +# t=ultralytics/yolov5:latest && sudo docker pull $t && sudo docker run -it --ipc=host --gpus all -v "$(pwd)"/datasets:/usr/src/datasets $t + +# Kill all +# sudo docker kill $(sudo docker ps -q) + +# Kill all image-based +# sudo docker kill $(sudo docker ps -qa --filter ancestor=ultralytics/yolov5:latest) + +# Bash into running container +# sudo docker exec -it 5a9b5863d93d bash + +# Bash into stopped container +# id=$(sudo docker ps -qa) && sudo docker start $id && sudo docker exec -it $id bash + +# Clean up +# docker system prune -a --volumes + +# Update Ubuntu drivers +# https://www.maketecheasier.com/install-nvidia-drivers-ubuntu/ + +# DDP test +# python -m torch.distributed.run --nproc_per_node 2 --master_port 1 train.py --epochs 3 + +# GCP VM from Image +# docker.io/ultralytics/yolov5:latest diff --git a/LICENSE b/LICENSE new file mode 100644 index 0000000..92b370f --- /dev/null +++ b/LICENSE @@ -0,0 +1,674 @@ +GNU GENERAL PUBLIC LICENSE + Version 3, 29 June 2007 + + Copyright (C) 2007 Free Software Foundation, Inc. + Everyone is permitted to copy and distribute verbatim copies + of this license document, but changing it is not allowed. + + Preamble + + The GNU General Public License is a free, copyleft license for +software and other kinds of works. + + The licenses for most software and other practical works are designed +to take away your freedom to share and change the works. By contrast, +the GNU General Public License is intended to guarantee your freedom to +share and change all versions of a program--to make sure it remains free +software for all its users. We, the Free Software Foundation, use the +GNU General Public License for most of our software; it applies also to +any other work released this way by its authors. You can apply it to +your programs, too. + + When we speak of free software, we are referring to freedom, not +price. Our General Public Licenses are designed to make sure that you +have the freedom to distribute copies of free software (and charge for +them if you wish), that you receive source code or can get it if you +want it, that you can change the software or use pieces of it in new +free programs, and that you know you can do these things. + + To protect your rights, we need to prevent others from denying you +these rights or asking you to surrender the rights. Therefore, you have +certain responsibilities if you distribute copies of the software, or if +you modify it: responsibilities to respect the freedom of others. + + For example, if you distribute copies of such a program, whether +gratis or for a fee, you must pass on to the recipients the same +freedoms that you received. You must make sure that they, too, receive +or can get the source code. And you must show them these terms so they +know their rights. + + Developers that use the GNU GPL protect your rights with two steps: +(1) assert copyright on the software, and (2) offer you this License +giving you legal permission to copy, distribute and/or modify it. + + For the developers' and authors' protection, the GPL clearly explains +that there is no warranty for this free software. For both users' and +authors' sake, the GPL requires that modified versions be marked as +changed, so that their problems will not be attributed erroneously to +authors of previous versions. + + Some devices are designed to deny users access to install or run +modified versions of the software inside them, although the manufacturer +can do so. This is fundamentally incompatible with the aim of +protecting users' freedom to change the software. The systematic +pattern of such abuse occurs in the area of products for individuals to +use, which is precisely where it is most unacceptable. Therefore, we +have designed this version of the GPL to prohibit the practice for those +products. If such problems arise substantially in other domains, we +stand ready to extend this provision to those domains in future versions +of the GPL, as needed to protect the freedom of users. + + Finally, every program is threatened constantly by software patents. +States should not allow patents to restrict development and use of +software on general-purpose computers, but in those that do, we wish to +avoid the special danger that patents applied to a free program could +make it effectively proprietary. To prevent this, the GPL assures that +patents cannot be used to render the program non-free. + + The precise terms and conditions for copying, distribution and +modification follow. + + TERMS AND CONDITIONS + + 0. Definitions. + + "This License" refers to version 3 of the GNU General Public License. + + "Copyright" also means copyright-like laws that apply to other kinds of +works, such as semiconductor masks. + + "The Program" refers to any copyrightable work licensed under this +License. Each licensee is addressed as "you". "Licensees" and +"recipients" may be individuals or organizations. + + To "modify" a work means to copy from or adapt all or part of the work +in a fashion requiring copyright permission, other than the making of an +exact copy. The resulting work is called a "modified version" of the +earlier work or a work "based on" the earlier work. + + A "covered work" means either the unmodified Program or a work based +on the Program. + + To "propagate" a work means to do anything with it that, without +permission, would make you directly or secondarily liable for +infringement under applicable copyright law, except executing it on a +computer or modifying a private copy. Propagation includes copying, +distribution (with or without modification), making available to the +public, and in some countries other activities as well. + + To "convey" a work means any kind of propagation that enables other +parties to make or receive copies. Mere interaction with a user through +a computer network, with no transfer of a copy, is not conveying. + + An interactive user interface displays "Appropriate Legal Notices" +to the extent that it includes a convenient and prominently visible +feature that (1) displays an appropriate copyright notice, and (2) +tells the user that there is no warranty for the work (except to the +extent that warranties are provided), that licensees may convey the +work under this License, and how to view a copy of this License. If +the interface presents a list of user commands or options, such as a +menu, a prominent item in the list meets this criterion. + + 1. Source Code. + + The "source code" for a work means the preferred form of the work +for making modifications to it. "Object code" means any non-source +form of a work. + + A "Standard Interface" means an interface that either is an official +standard defined by a recognized standards body, or, in the case of +interfaces specified for a particular programming language, one that +is widely used among developers working in that language. + + The "System Libraries" of an executable work include anything, other +than the work as a whole, that (a) is included in the normal form of +packaging a Major Component, but which is not part of that Major +Component, and (b) serves only to enable use of the work with that +Major Component, or to implement a Standard Interface for which an +implementation is available to the public in source code form. A +"Major Component", in this context, means a major essential component +(kernel, window system, and so on) of the specific operating system +(if any) on which the executable work runs, or a compiler used to +produce the work, or an object code interpreter used to run it. + + The "Corresponding Source" for a work in object code form means all +the source code needed to generate, install, and (for an executable +work) run the object code and to modify the work, including scripts to +control those activities. However, it does not include the work's +System Libraries, or general-purpose tools or generally available free +programs which are used unmodified in performing those activities but +which are not part of the work. For example, Corresponding Source +includes interface definition files associated with source files for +the work, and the source code for shared libraries and dynamically +linked subprograms that the work is specifically designed to require, +such as by intimate data communication or control flow between those +subprograms and other parts of the work. + + The Corresponding Source need not include anything that users +can regenerate automatically from other parts of the Corresponding +Source. + + The Corresponding Source for a work in source code form is that +same work. + + 2. Basic Permissions. + + All rights granted under this License are granted for the term of +copyright on the Program, and are irrevocable provided the stated +conditions are met. This License explicitly affirms your unlimited +permission to run the unmodified Program. The output from running a +covered work is covered by this License only if the output, given its +content, constitutes a covered work. This License acknowledges your +rights of fair use or other equivalent, as provided by copyright law. + + You may make, run and propagate covered works that you do not +convey, without conditions so long as your license otherwise remains +in force. You may convey covered works to others for the sole purpose +of having them make modifications exclusively for you, or provide you +with facilities for running those works, provided that you comply with +the terms of this License in conveying all material for which you do +not control copyright. Those thus making or running the covered works +for you must do so exclusively on your behalf, under your direction +and control, on terms that prohibit them from making any copies of +your copyrighted material outside their relationship with you. + + Conveying under any other circumstances is permitted solely under +the conditions stated below. Sublicensing is not allowed; section 10 +makes it unnecessary. + + 3. Protecting Users' Legal Rights From Anti-Circumvention Law. + + No covered work shall be deemed part of an effective technological +measure under any applicable law fulfilling obligations under article +11 of the WIPO copyright treaty adopted on 20 December 1996, or +similar laws prohibiting or restricting circumvention of such +measures. + + When you convey a covered work, you waive any legal power to forbid +circumvention of technological measures to the extent such circumvention +is effected by exercising rights under this License with respect to +the covered work, and you disclaim any intention to limit operation or +modification of the work as a means of enforcing, against the work's +users, your or third parties' legal rights to forbid circumvention of +technological measures. + + 4. Conveying Verbatim Copies. + + You may convey verbatim copies of the Program's source code as you +receive it, in any medium, provided that you conspicuously and +appropriately publish on each copy an appropriate copyright notice; +keep intact all notices stating that this License and any +non-permissive terms added in accord with section 7 apply to the code; +keep intact all notices of the absence of any warranty; and give all +recipients a copy of this License along with the Program. + + You may charge any price or no price for each copy that you convey, +and you may offer support or warranty protection for a fee. + + 5. Conveying Modified Source Versions. + + You may convey a work based on the Program, or the modifications to +produce it from the Program, in the form of source code under the +terms of section 4, provided that you also meet all of these conditions: + + a) The work must carry prominent notices stating that you modified + it, and giving a relevant date. + + b) The work must carry prominent notices stating that it is + released under this License and any conditions added under section + 7. This requirement modifies the requirement in section 4 to + "keep intact all notices". + + c) You must license the entire work, as a whole, under this + License to anyone who comes into possession of a copy. This + License will therefore apply, along with any applicable section 7 + additional terms, to the whole of the work, and all its parts, + regardless of how they are packaged. This License gives no + permission to license the work in any other way, but it does not + invalidate such permission if you have separately received it. + + d) If the work has interactive user interfaces, each must display + Appropriate Legal Notices; however, if the Program has interactive + interfaces that do not display Appropriate Legal Notices, your + work need not make them do so. + + A compilation of a covered work with other separate and independent +works, which are not by their nature extensions of the covered work, +and which are not combined with it such as to form a larger program, +in or on a volume of a storage or distribution medium, is called an +"aggregate" if the compilation and its resulting copyright are not +used to limit the access or legal rights of the compilation's users +beyond what the individual works permit. Inclusion of a covered work +in an aggregate does not cause this License to apply to the other +parts of the aggregate. + + 6. Conveying Non-Source Forms. + + You may convey a covered work in object code form under the terms +of sections 4 and 5, provided that you also convey the +machine-readable Corresponding Source under the terms of this License, +in one of these ways: + + a) Convey the object code in, or embodied in, a physical product + (including a physical distribution medium), accompanied by the + Corresponding Source fixed on a durable physical medium + customarily used for software interchange. + + b) Convey the object code in, or embodied in, a physical product + (including a physical distribution medium), accompanied by a + written offer, valid for at least three years and valid for as + long as you offer spare parts or customer support for that product + model, to give anyone who possesses the object code either (1) a + copy of the Corresponding Source for all the software in the + product that is covered by this License, on a durable physical + medium customarily used for software interchange, for a price no + more than your reasonable cost of physically performing this + conveying of source, or (2) access to copy the + Corresponding Source from a network server at no charge. + + c) Convey individual copies of the object code with a copy of the + written offer to provide the Corresponding Source. This + alternative is allowed only occasionally and noncommercially, and + only if you received the object code with such an offer, in accord + with subsection 6b. + + d) Convey the object code by offering access from a designated + place (gratis or for a charge), and offer equivalent access to the + Corresponding Source in the same way through the same place at no + further charge. You need not require recipients to copy the + Corresponding Source along with the object code. If the place to + copy the object code is a network server, the Corresponding Source + may be on a different server (operated by you or a third party) + that supports equivalent copying facilities, provided you maintain + clear directions next to the object code saying where to find the + Corresponding Source. Regardless of what server hosts the + Corresponding Source, you remain obligated to ensure that it is + available for as long as needed to satisfy these requirements. + + e) Convey the object code using peer-to-peer transmission, provided + you inform other peers where the object code and Corresponding + Source of the work are being offered to the general public at no + charge under subsection 6d. + + A separable portion of the object code, whose source code is excluded +from the Corresponding Source as a System Library, need not be +included in conveying the object code work. + + A "User Product" is either (1) a "consumer product", which means any +tangible personal property which is normally used for personal, family, +or household purposes, or (2) anything designed or sold for incorporation +into a dwelling. In determining whether a product is a consumer product, +doubtful cases shall be resolved in favor of coverage. For a particular +product received by a particular user, "normally used" refers to a +typical or common use of that class of product, regardless of the status +of the particular user or of the way in which the particular user +actually uses, or expects or is expected to use, the product. A product +is a consumer product regardless of whether the product has substantial +commercial, industrial or non-consumer uses, unless such uses represent +the only significant mode of use of the product. + + "Installation Information" for a User Product means any methods, +procedures, authorization keys, or other information required to install +and execute modified versions of a covered work in that User Product from +a modified version of its Corresponding Source. The information must +suffice to ensure that the continued functioning of the modified object +code is in no case prevented or interfered with solely because +modification has been made. + + If you convey an object code work under this section in, or with, or +specifically for use in, a User Product, and the conveying occurs as +part of a transaction in which the right of possession and use of the +User Product is transferred to the recipient in perpetuity or for a +fixed term (regardless of how the transaction is characterized), the +Corresponding Source conveyed under this section must be accompanied +by the Installation Information. But this requirement does not apply +if neither you nor any third party retains the ability to install +modified object code on the User Product (for example, the work has +been installed in ROM). + + The requirement to provide Installation Information does not include a +requirement to continue to provide support service, warranty, or updates +for a work that has been modified or installed by the recipient, or for +the User Product in which it has been modified or installed. Access to a +network may be denied when the modification itself materially and +adversely affects the operation of the network or violates the rules and +protocols for communication across the network. + + Corresponding Source conveyed, and Installation Information provided, +in accord with this section must be in a format that is publicly +documented (and with an implementation available to the public in +source code form), and must require no special password or key for +unpacking, reading or copying. + + 7. Additional Terms. + + "Additional permissions" are terms that supplement the terms of this +License by making exceptions from one or more of its conditions. +Additional permissions that are applicable to the entire Program shall +be treated as though they were included in this License, to the extent +that they are valid under applicable law. If additional permissions +apply only to part of the Program, that part may be used separately +under those permissions, but the entire Program remains governed by +this License without regard to the additional permissions. + + When you convey a copy of a covered work, you may at your option +remove any additional permissions from that copy, or from any part of +it. (Additional permissions may be written to require their own +removal in certain cases when you modify the work.) You may place +additional permissions on material, added by you to a covered work, +for which you have or can give appropriate copyright permission. + + Notwithstanding any other provision of this License, for material you +add to a covered work, you may (if authorized by the copyright holders of +that material) supplement the terms of this License with terms: + + a) Disclaiming warranty or limiting liability differently from the + terms of sections 15 and 16 of this License; or + + b) Requiring preservation of specified reasonable legal notices or + author attributions in that material or in the Appropriate Legal + Notices displayed by works containing it; or + + c) Prohibiting misrepresentation of the origin of that material, or + requiring that modified versions of such material be marked in + reasonable ways as different from the original version; or + + d) Limiting the use for publicity purposes of names of licensors or + authors of the material; or + + e) Declining to grant rights under trademark law for use of some + trade names, trademarks, or service marks; or + + f) Requiring indemnification of licensors and authors of that + material by anyone who conveys the material (or modified versions of + it) with contractual assumptions of liability to the recipient, for + any liability that these contractual assumptions directly impose on + those licensors and authors. + + All other non-permissive additional terms are considered "further +restrictions" within the meaning of section 10. If the Program as you +received it, or any part of it, contains a notice stating that it is +governed by this License along with a term that is a further +restriction, you may remove that term. If a license document contains +a further restriction but permits relicensing or conveying under this +License, you may add to a covered work material governed by the terms +of that license document, provided that the further restriction does +not survive such relicensing or conveying. + + If you add terms to a covered work in accord with this section, you +must place, in the relevant source files, a statement of the +additional terms that apply to those files, or a notice indicating +where to find the applicable terms. + + Additional terms, permissive or non-permissive, may be stated in the +form of a separately written license, or stated as exceptions; +the above requirements apply either way. + + 8. Termination. + + You may not propagate or modify a covered work except as expressly +provided under this License. Any attempt otherwise to propagate or +modify it is void, and will automatically terminate your rights under +this License (including any patent licenses granted under the third +paragraph of section 11). + + However, if you cease all violation of this License, then your +license from a particular copyright holder is reinstated (a) +provisionally, unless and until the copyright holder explicitly and +finally terminates your license, and (b) permanently, if the copyright +holder fails to notify you of the violation by some reasonable means +prior to 60 days after the cessation. + + Moreover, your license from a particular copyright holder is +reinstated permanently if the copyright holder notifies you of the +violation by some reasonable means, this is the first time you have +received notice of violation of this License (for any work) from that +copyright holder, and you cure the violation prior to 30 days after +your receipt of the notice. + + Termination of your rights under this section does not terminate the +licenses of parties who have received copies or rights from you under +this License. If your rights have been terminated and not permanently +reinstated, you do not qualify to receive new licenses for the same +material under section 10. + + 9. Acceptance Not Required for Having Copies. + + You are not required to accept this License in order to receive or +run a copy of the Program. Ancillary propagation of a covered work +occurring solely as a consequence of using peer-to-peer transmission +to receive a copy likewise does not require acceptance. However, +nothing other than this License grants you permission to propagate or +modify any covered work. These actions infringe copyright if you do +not accept this License. Therefore, by modifying or propagating a +covered work, you indicate your acceptance of this License to do so. + + 10. Automatic Licensing of Downstream Recipients. + + Each time you convey a covered work, the recipient automatically +receives a license from the original licensors, to run, modify and +propagate that work, subject to this License. You are not responsible +for enforcing compliance by third parties with this License. + + An "entity transaction" is a transaction transferring control of an +organization, or substantially all assets of one, or subdividing an +organization, or merging organizations. If propagation of a covered +work results from an entity transaction, each party to that +transaction who receives a copy of the work also receives whatever +licenses to the work the party's predecessor in interest had or could +give under the previous paragraph, plus a right to possession of the +Corresponding Source of the work from the predecessor in interest, if +the predecessor has it or can get it with reasonable efforts. + + You may not impose any further restrictions on the exercise of the +rights granted or affirmed under this License. For example, you may +not impose a license fee, royalty, or other charge for exercise of +rights granted under this License, and you may not initiate litigation +(including a cross-claim or counterclaim in a lawsuit) alleging that +any patent claim is infringed by making, using, selling, offering for +sale, or importing the Program or any portion of it. + + 11. Patents. + + A "contributor" is a copyright holder who authorizes use under this +License of the Program or a work on which the Program is based. The +work thus licensed is called the contributor's "contributor version". + + A contributor's "essential patent claims" are all patent claims +owned or controlled by the contributor, whether already acquired or +hereafter acquired, that would be infringed by some manner, permitted +by this License, of making, using, or selling its contributor version, +but do not include claims that would be infringed only as a +consequence of further modification of the contributor version. For +purposes of this definition, "control" includes the right to grant +patent sublicenses in a manner consistent with the requirements of +this License. + + Each contributor grants you a non-exclusive, worldwide, royalty-free +patent license under the contributor's essential patent claims, to +make, use, sell, offer for sale, import and otherwise run, modify and +propagate the contents of its contributor version. + + In the following three paragraphs, a "patent license" is any express +agreement or commitment, however denominated, not to enforce a patent +(such as an express permission to practice a patent or covenant not to +sue for patent infringement). To "grant" such a patent license to a +party means to make such an agreement or commitment not to enforce a +patent against the party. + + If you convey a covered work, knowingly relying on a patent license, +and the Corresponding Source of the work is not available for anyone +to copy, free of charge and under the terms of this License, through a +publicly available network server or other readily accessible means, +then you must either (1) cause the Corresponding Source to be so +available, or (2) arrange to deprive yourself of the benefit of the +patent license for this particular work, or (3) arrange, in a manner +consistent with the requirements of this License, to extend the patent +license to downstream recipients. "Knowingly relying" means you have +actual knowledge that, but for the patent license, your conveying the +covered work in a country, or your recipient's use of the covered work +in a country, would infringe one or more identifiable patents in that +country that you have reason to believe are valid. + + If, pursuant to or in connection with a single transaction or +arrangement, you convey, or propagate by procuring conveyance of, a +covered work, and grant a patent license to some of the parties +receiving the covered work authorizing them to use, propagate, modify +or convey a specific copy of the covered work, then the patent license +you grant is automatically extended to all recipients of the covered +work and works based on it. + + A patent license is "discriminatory" if it does not include within +the scope of its coverage, prohibits the exercise of, or is +conditioned on the non-exercise of one or more of the rights that are +specifically granted under this License. You may not convey a covered +work if you are a party to an arrangement with a third party that is +in the business of distributing software, under which you make payment +to the third party based on the extent of your activity of conveying +the work, and under which the third party grants, to any of the +parties who would receive the covered work from you, a discriminatory +patent license (a) in connection with copies of the covered work +conveyed by you (or copies made from those copies), or (b) primarily +for and in connection with specific products or compilations that +contain the covered work, unless you entered into that arrangement, +or that patent license was granted, prior to 28 March 2007. + + Nothing in this License shall be construed as excluding or limiting +any implied license or other defenses to infringement that may +otherwise be available to you under applicable patent law. + + 12. No Surrender of Others' Freedom. + + If conditions are imposed on you (whether by court order, agreement or +otherwise) that contradict the conditions of this License, they do not +excuse you from the conditions of this License. If you cannot convey a +covered work so as to satisfy simultaneously your obligations under this +License and any other pertinent obligations, then as a consequence you may +not convey it at all. For example, if you agree to terms that obligate you +to collect a royalty for further conveying from those to whom you convey +the Program, the only way you could satisfy both those terms and this +License would be to refrain entirely from conveying the Program. + + 13. Use with the GNU Affero General Public License. + + Notwithstanding any other provision of this License, you have +permission to link or combine any covered work with a work licensed +under version 3 of the GNU Affero General Public License into a single +combined work, and to convey the resulting work. The terms of this +License will continue to apply to the part which is the covered work, +but the special requirements of the GNU Affero General Public License, +section 13, concerning interaction through a network will apply to the +combination as such. + + 14. Revised Versions of this License. + + The Free Software Foundation may publish revised and/or new versions of +the GNU General Public License from time to time. Such new versions will +be similar in spirit to the present version, but may differ in detail to +address new problems or concerns. + + Each version is given a distinguishing version number. If the +Program specifies that a certain numbered version of the GNU General +Public License "or any later version" applies to it, you have the +option of following the terms and conditions either of that numbered +version or of any later version published by the Free Software +Foundation. If the Program does not specify a version number of the +GNU General Public License, you may choose any version ever published +by the Free Software Foundation. + + If the Program specifies that a proxy can decide which future +versions of the GNU General Public License can be used, that proxy's +public statement of acceptance of a version permanently authorizes you +to choose that version for the Program. + + Later license versions may give you additional or different +permissions. However, no additional obligations are imposed on any +author or copyright holder as a result of your choosing to follow a +later version. + + 15. Disclaimer of Warranty. + + THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY +APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT +HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY +OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, +THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR +PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM +IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF +ALL NECESSARY SERVICING, REPAIR OR CORRECTION. + + 16. Limitation of Liability. + + IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING +WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS +THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY +GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE +USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF +DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD +PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), +EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF +SUCH DAMAGES. + + 17. Interpretation of Sections 15 and 16. + + If the disclaimer of warranty and limitation of liability provided +above cannot be given local legal effect according to their terms, +reviewing courts shall apply local law that most closely approximates +an absolute waiver of all civil liability in connection with the +Program, unless a warranty or assumption of liability accompanies a +copy of the Program in return for a fee. + + END OF TERMS AND CONDITIONS + + How to Apply These Terms to Your New Programs + + If you develop a new program, and you want it to be of the greatest +possible use to the public, the best way to achieve this is to make it +free software which everyone can redistribute and change under these terms. + + To do so, attach the following notices to the program. It is safest +to attach them to the start of each source file to most effectively +state the exclusion of warranty; and each file should have at least +the "copyright" line and a pointer to where the full notice is found. + + + Copyright (C) + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . + +Also add information on how to contact you by electronic and paper mail. + + If the program does terminal interaction, make it output a short +notice like this when it starts in an interactive mode: + + Copyright (C) + This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. + This is free software, and you are welcome to redistribute it + under certain conditions; type `show c' for details. + +The hypothetical commands `show w' and `show c' should show the appropriate +parts of the General Public License. Of course, your program's commands +might be different; for a GUI interface, you would use an "about box". + + You should also get your employer (if you work as a programmer) or school, +if any, to sign a "copyright disclaimer" for the program, if necessary. +For more information on this, and how to apply and follow the GNU GPL, see +. + + The GNU General Public License does not permit incorporating your program +into proprietary programs. If your program is a subroutine library, you +may consider it more useful to permit linking proprietary applications with +the library. If this is what you want to do, use the GNU Lesser General +Public License instead of this License. But first, please read +. diff --git a/README.md b/README.md new file mode 100644 index 0000000..b51d3b4 --- /dev/null +++ b/README.md @@ -0,0 +1,306 @@ +#for study purpose + +
+

+ + +

+
+
+ CI CPU testing + YOLOv5 Citation + Docker Pulls +
+ Open In Colab + Open In Kaggle + Join Forum +
+ +
+

+YOLOv5 🚀 is a family of object detection architectures and models pretrained on the COCO dataset, and represents Ultralytics + open-source research into future vision AI methods, incorporating lessons learned and best practices evolved over thousands of hours of research and development. +

+ + + + + +
+ +##
Documentation
+ +See the [YOLOv5 Docs](https://docs.ultralytics.com) for full documentation on training, testing and deployment. + +##
Quick Start Examples
+ +
+Install + +Clone repo and install [requirements.txt](https://github.com/ultralytics/yolov5/blob/master/requirements.txt) in a +[**Python>=3.7.0**](https://www.python.org/) environment, including +[**PyTorch>=1.7**](https://pytorch.org/get-started/locally/). + +```bash +git clone https://github.com/ultralytics/yolov5 # clone +cd yolov5 +pip install -r requirements.txt # install +``` + +
+ +
+Inference + +Inference with YOLOv5 and [PyTorch Hub](https://github.com/ultralytics/yolov5/issues/36) +. [Models](https://github.com/ultralytics/yolov5/tree/master/models) download automatically from the latest +YOLOv5 [release](https://github.com/ultralytics/yolov5/releases). + +```python +import torch + +# Model +model = torch.hub.load('ultralytics/yolov5', 'yolov5s') # or yolov5m, yolov5l, yolov5x, custom + +# Images +img = 'https://ultralytics.com/images/zidane.jpg' # or file, Path, PIL, OpenCV, numpy, list + +# Inference +results = model(img) + +# Results +results.print() # or .show(), .save(), .crop(), .pandas(), etc. +``` + +
+ + + +
+Inference with detect.py + +`detect.py` runs inference on a variety of sources, downloading [models](https://github.com/ultralytics/yolov5/tree/master/models) automatically from +the latest YOLOv5 [release](https://github.com/ultralytics/yolov5/releases) and saving results to `runs/detect`. + +```bash +python detect.py --source 0 # webcam + img.jpg # image + vid.mp4 # video + path/ # directory + path/*.jpg # glob + 'https://youtu.be/Zgi9g1ksQHc' # YouTube + 'rtsp://example.com/media.mp4' # RTSP, RTMP, HTTP stream +``` + +
+ +
+Training + +The commands below reproduce YOLOv5 [COCO](https://github.com/ultralytics/yolov5/blob/master/data/scripts/get_coco.sh) +results. [Models](https://github.com/ultralytics/yolov5/tree/master/models) +and [datasets](https://github.com/ultralytics/yolov5/tree/master/data) download automatically from the latest +YOLOv5 [release](https://github.com/ultralytics/yolov5/releases). Training times for YOLOv5n/s/m/l/x are +1/2/4/6/8 days on a V100 GPU ([Multi-GPU](https://github.com/ultralytics/yolov5/issues/475) times faster). Use the +largest `--batch-size` possible, or pass `--batch-size -1` for +YOLOv5 [AutoBatch](https://github.com/ultralytics/yolov5/pull/5092). Batch sizes shown for V100-16GB. + +```bash +python train.py --data coco.yaml --cfg yolov5n.yaml --weights '' --batch-size 128 + yolov5s 64 + yolov5m 40 + yolov5l 24 + yolov5x 16 +``` + + + +
+ +
+Tutorials + +* [Train Custom Data](https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data)  🚀 RECOMMENDED +* [Tips for Best Training Results](https://github.com/ultralytics/yolov5/wiki/Tips-for-Best-Training-Results)  ☘ī¸ + RECOMMENDED +* [Weights & Biases Logging](https://github.com/ultralytics/yolov5/issues/1289)  🌟 NEW +* [Roboflow for Datasets, Labeling, and Active Learning](https://github.com/ultralytics/yolov5/issues/4975)  🌟 NEW +* [Multi-GPU Training](https://github.com/ultralytics/yolov5/issues/475) +* [PyTorch Hub](https://github.com/ultralytics/yolov5/issues/36)  ⭐ NEW +* [TFLite, ONNX, CoreML, TensorRT Export](https://github.com/ultralytics/yolov5/issues/251) 🚀 +* [Test-Time Augmentation (TTA)](https://github.com/ultralytics/yolov5/issues/303) +* [Model Ensembling](https://github.com/ultralytics/yolov5/issues/318) +* [Model Pruning/Sparsity](https://github.com/ultralytics/yolov5/issues/304) +* [Hyperparameter Evolution](https://github.com/ultralytics/yolov5/issues/607) +* [Transfer Learning with Frozen Layers](https://github.com/ultralytics/yolov5/issues/1314)  ⭐ NEW +* [TensorRT Deployment](https://github.com/wang-xinyu/tensorrtx) + +
+ +##
Environments
+ +Get started in seconds with our verified environments. Click each icon below for details. + + + +##
Integrations
+ + + +|Weights and Biases|Roboflow ⭐ NEW| +|:-:|:-:| +|Automatically track and visualize all your YOLOv5 training runs in the cloud with [Weights & Biases](https://wandb.ai/site?utm_campaign=repo_yolo_readme)|Label and export your custom datasets directly to YOLOv5 for training with [Roboflow](https://roboflow.com/?ref=ultralytics) | + + + + +##
Why YOLOv5
+ +

+
+ YOLOv5-P5 640 Figure (click to expand) + +

+
+
+ Figure Notes (click to expand) + +* **COCO AP val** denotes mAP@0.5:0.95 metric measured on the 5000-image [COCO val2017](http://cocodataset.org) dataset over various inference sizes from 256 to 1536. +* **GPU Speed** measures average inference time per image on [COCO val2017](http://cocodataset.org) dataset using a [AWS p3.2xlarge](https://aws.amazon.com/ec2/instance-types/p3/) V100 instance at batch-size 32. +* **EfficientDet** data from [google/automl](https://github.com/google/automl) at batch size 8. +* **Reproduce** by `python val.py --task study --data coco.yaml --iou 0.7 --weights yolov5n6.pt yolov5s6.pt yolov5m6.pt yolov5l6.pt yolov5x6.pt` +
+ +### Pretrained Checkpoints + +[assets]: https://github.com/ultralytics/yolov5/releases + +[TTA]: https://github.com/ultralytics/yolov5/issues/303 + +|Model |size
(pixels) |mAPval
0.5:0.95 |mAPval
0.5 |Speed
CPU b1
(ms) |Speed
V100 b1
(ms) |Speed
V100 b32
(ms) |params
(M) |FLOPs
@640 (B) +|--- |--- |--- |--- |--- |--- |--- |--- |--- +|[YOLOv5n][assets] |640 |28.0 |45.7 |**45** |**6.3**|**0.6**|**1.9**|**4.5** +|[YOLOv5s][assets] |640 |37.4 |56.8 |98 |6.4 |0.9 |7.2 |16.5 +|[YOLOv5m][assets] |640 |45.4 |64.1 |224 |8.2 |1.7 |21.2 |49.0 +|[YOLOv5l][assets] |640 |49.0 |67.3 |430 |10.1 |2.7 |46.5 |109.1 +|[YOLOv5x][assets] |640 |50.7 |68.9 |766 |12.1 |4.8 |86.7 |205.7 +| | | | | | | | | +|[YOLOv5n6][assets] |1280 |36.0 |54.4 |153 |8.1 |2.1 |3.2 |4.6 +|[YOLOv5s6][assets] |1280 |44.8 |63.7 |385 |8.2 |3.6 |16.8 |12.6 +|[YOLOv5m6][assets] |1280 |51.3 |69.3 |887 |11.1 |6.8 |35.7 |50.0 +|[YOLOv5l6][assets] |1280 |53.7 |71.3 |1784 |15.8 |10.5 |76.8 |111.4 +|[YOLOv5x6][assets]
+ [TTA][TTA]|1280
1536 |55.0
**55.8** |72.7
**72.7** |3136
- |26.2
- |19.4
- |140.7
- |209.8
- + +
+ Table Notes (click to expand) + +* All checkpoints are trained to 300 epochs with default settings. Nano and Small models use [hyp.scratch-low.yaml](https://github.com/ultralytics/yolov5/blob/master/data/hyps/hyp.scratch-low.yaml) hyps, all others use [hyp.scratch-high.yaml](https://github.com/ultralytics/yolov5/blob/master/data/hyps/hyp.scratch-high.yaml). +* **mAPval** values are for single-model single-scale on [COCO val2017](http://cocodataset.org) dataset.
Reproduce by `python val.py --data coco.yaml --img 640 --conf 0.001 --iou 0.65` +* **Speed** averaged over COCO val images using a [AWS p3.2xlarge](https://aws.amazon.com/ec2/instance-types/p3/) instance. NMS times (~1 ms/img) not included.
Reproduce by `python val.py --data coco.yaml --img 640 --task speed --batch 1` +* **TTA** [Test Time Augmentation](https://github.com/ultralytics/yolov5/issues/303) includes reflection and scale augmentations.
Reproduce by `python val.py --data coco.yaml --img 1536 --iou 0.7 --augment` + +
+ +##
Contribute
+ +We love your input! We want to make contributing to YOLOv5 as easy and transparent as possible. Please see our [Contributing Guide](CONTRIBUTING.md) to get started, and fill out the [YOLOv5 Survey](https://ultralytics.com/survey?utm_source=github&utm_medium=social&utm_campaign=Survey) to send us feedback on your experiences. Thank you to all our contributors! + + + +##
Contact
+ +For YOLOv5 bugs and feature requests please visit [GitHub Issues](https://github.com/ultralytics/yolov5/issues). For business inquiries or +professional support requests please visit [https://ultralytics.com/contact](https://ultralytics.com/contact). + +
+ + diff --git a/data/Argoverse.yaml b/data/Argoverse.yaml new file mode 100644 index 0000000..312791b --- /dev/null +++ b/data/Argoverse.yaml @@ -0,0 +1,67 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# Argoverse-HD dataset (ring-front-center camera) http://www.cs.cmu.edu/~mengtial/proj/streaming/ by Argo AI +# Example usage: python train.py --data Argoverse.yaml +# parent +# ├── yolov5 +# └── datasets +# └── Argoverse ← downloads here + + +# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..] +path: ../datasets/Argoverse # dataset root dir +train: Argoverse-1.1/images/train/ # train images (relative to 'path') 39384 images +val: Argoverse-1.1/images/val/ # val images (relative to 'path') 15062 images +test: Argoverse-1.1/images/test/ # test images (optional) https://eval.ai/web/challenges/challenge-page/800/overview + +# Classes +nc: 8 # number of classes +names: ['person', 'bicycle', 'car', 'motorcycle', 'bus', 'truck', 'traffic_light', 'stop_sign'] # class names + + +# Download script/URL (optional) --------------------------------------------------------------------------------------- +download: | + import json + + from tqdm import tqdm + from utils.general import download, Path + + + def argoverse2yolo(set): + labels = {} + a = json.load(open(set, "rb")) + for annot in tqdm(a['annotations'], desc=f"Converting {set} to YOLOv5 format..."): + img_id = annot['image_id'] + img_name = a['images'][img_id]['name'] + img_label_name = img_name[:-3] + "txt" + + cls = annot['category_id'] # instance class id + x_center, y_center, width, height = annot['bbox'] + x_center = (x_center + width / 2) / 1920.0 # offset and scale + y_center = (y_center + height / 2) / 1200.0 # offset and scale + width /= 1920.0 # scale + height /= 1200.0 # scale + + img_dir = set.parents[2] / 'Argoverse-1.1' / 'labels' / a['seq_dirs'][a['images'][annot['image_id']]['sid']] + if not img_dir.exists(): + img_dir.mkdir(parents=True, exist_ok=True) + + k = str(img_dir / img_label_name) + if k not in labels: + labels[k] = [] + labels[k].append(f"{cls} {x_center} {y_center} {width} {height}\n") + + for k in labels: + with open(k, "w") as f: + f.writelines(labels[k]) + + + # Download + dir = Path('../datasets/Argoverse') # dataset root dir + urls = ['https://argoverse-hd.s3.us-east-2.amazonaws.com/Argoverse-HD-Full.zip'] + download(urls, dir=dir, delete=False) + + # Convert + annotations_dir = 'Argoverse-HD/annotations/' + (dir / 'Argoverse-1.1' / 'tracking').rename(dir / 'Argoverse-1.1' / 'images') # rename 'tracking' to 'images' + for d in "train.json", "val.json": + argoverse2yolo(dir / annotations_dir / d) # convert VisDrone annotations to YOLO labels diff --git a/data/GlobalWheat2020.yaml b/data/GlobalWheat2020.yaml new file mode 100644 index 0000000..869dace --- /dev/null +++ b/data/GlobalWheat2020.yaml @@ -0,0 +1,53 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# Global Wheat 2020 dataset http://www.global-wheat.com/ by University of Saskatchewan +# Example usage: python train.py --data GlobalWheat2020.yaml +# parent +# ├── yolov5 +# └── datasets +# └── GlobalWheat2020 ← downloads here + + +# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..] +path: ../datasets/GlobalWheat2020 # dataset root dir +train: # train images (relative to 'path') 3422 images + - images/arvalis_1 + - images/arvalis_2 + - images/arvalis_3 + - images/ethz_1 + - images/rres_1 + - images/inrae_1 + - images/usask_1 +val: # val images (relative to 'path') 748 images (WARNING: train set contains ethz_1) + - images/ethz_1 +test: # test images (optional) 1276 images + - images/utokyo_1 + - images/utokyo_2 + - images/nau_1 + - images/uq_1 + +# Classes +nc: 1 # number of classes +names: ['wheat_head'] # class names + + +# Download script/URL (optional) --------------------------------------------------------------------------------------- +download: | + from utils.general import download, Path + + # Download + dir = Path(yaml['path']) # dataset root dir + urls = ['https://zenodo.org/record/4298502/files/global-wheat-codalab-official.zip', + 'https://github.com/ultralytics/yolov5/releases/download/v1.0/GlobalWheat2020_labels.zip'] + download(urls, dir=dir) + + # Make Directories + for p in 'annotations', 'images', 'labels': + (dir / p).mkdir(parents=True, exist_ok=True) + + # Move + for p in 'arvalis_1', 'arvalis_2', 'arvalis_3', 'ethz_1', 'rres_1', 'inrae_1', 'usask_1', \ + 'utokyo_1', 'utokyo_2', 'nau_1', 'uq_1': + (dir / p).rename(dir / 'images' / p) # move to /images + f = (dir / p).with_suffix('.json') # json file + if f.exists(): + f.rename((dir / 'annotations' / p).with_suffix('.json')) # move to /annotations diff --git a/data/Objects365.yaml b/data/Objects365.yaml new file mode 100644 index 0000000..4c7cf3f --- /dev/null +++ b/data/Objects365.yaml @@ -0,0 +1,112 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# Objects365 dataset https://www.objects365.org/ by Megvii +# Example usage: python train.py --data Objects365.yaml +# parent +# ├── yolov5 +# └── datasets +# └── Objects365 ← downloads here + + +# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..] +path: ../datasets/Objects365 # dataset root dir +train: images/train # train images (relative to 'path') 1742289 images +val: images/val # val images (relative to 'path') 80000 images +test: # test images (optional) + +# Classes +nc: 365 # number of classes +names: ['Person', 'Sneakers', 'Chair', 'Other Shoes', 'Hat', 'Car', 'Lamp', 'Glasses', 'Bottle', 'Desk', 'Cup', + 'Street Lights', 'Cabinet/shelf', 'Handbag/Satchel', 'Bracelet', 'Plate', 'Picture/Frame', 'Helmet', 'Book', + 'Gloves', 'Storage box', 'Boat', 'Leather Shoes', 'Flower', 'Bench', 'Potted Plant', 'Bowl/Basin', 'Flag', + 'Pillow', 'Boots', 'Vase', 'Microphone', 'Necklace', 'Ring', 'SUV', 'Wine Glass', 'Belt', 'Monitor/TV', + 'Backpack', 'Umbrella', 'Traffic Light', 'Speaker', 'Watch', 'Tie', 'Trash bin Can', 'Slippers', 'Bicycle', + 'Stool', 'Barrel/bucket', 'Van', 'Couch', 'Sandals', 'Basket', 'Drum', 'Pen/Pencil', 'Bus', 'Wild Bird', + 'High Heels', 'Motorcycle', 'Guitar', 'Carpet', 'Cell Phone', 'Bread', 'Camera', 'Canned', 'Truck', + 'Traffic cone', 'Cymbal', 'Lifesaver', 'Towel', 'Stuffed Toy', 'Candle', 'Sailboat', 'Laptop', 'Awning', + 'Bed', 'Faucet', 'Tent', 'Horse', 'Mirror', 'Power outlet', 'Sink', 'Apple', 'Air Conditioner', 'Knife', + 'Hockey Stick', 'Paddle', 'Pickup Truck', 'Fork', 'Traffic Sign', 'Balloon', 'Tripod', 'Dog', 'Spoon', 'Clock', + 'Pot', 'Cow', 'Cake', 'Dinning Table', 'Sheep', 'Hanger', 'Blackboard/Whiteboard', 'Napkin', 'Other Fish', + 'Orange/Tangerine', 'Toiletry', 'Keyboard', 'Tomato', 'Lantern', 'Machinery Vehicle', 'Fan', + 'Green Vegetables', 'Banana', 'Baseball Glove', 'Airplane', 'Mouse', 'Train', 'Pumpkin', 'Soccer', 'Skiboard', + 'Luggage', 'Nightstand', 'Tea pot', 'Telephone', 'Trolley', 'Head Phone', 'Sports Car', 'Stop Sign', + 'Dessert', 'Scooter', 'Stroller', 'Crane', 'Remote', 'Refrigerator', 'Oven', 'Lemon', 'Duck', 'Baseball Bat', + 'Surveillance Camera', 'Cat', 'Jug', 'Broccoli', 'Piano', 'Pizza', 'Elephant', 'Skateboard', 'Surfboard', + 'Gun', 'Skating and Skiing shoes', 'Gas stove', 'Donut', 'Bow Tie', 'Carrot', 'Toilet', 'Kite', 'Strawberry', + 'Other Balls', 'Shovel', 'Pepper', 'Computer Box', 'Toilet Paper', 'Cleaning Products', 'Chopsticks', + 'Microwave', 'Pigeon', 'Baseball', 'Cutting/chopping Board', 'Coffee Table', 'Side Table', 'Scissors', + 'Marker', 'Pie', 'Ladder', 'Snowboard', 'Cookies', 'Radiator', 'Fire Hydrant', 'Basketball', 'Zebra', 'Grape', + 'Giraffe', 'Potato', 'Sausage', 'Tricycle', 'Violin', 'Egg', 'Fire Extinguisher', 'Candy', 'Fire Truck', + 'Billiards', 'Converter', 'Bathtub', 'Wheelchair', 'Golf Club', 'Briefcase', 'Cucumber', 'Cigar/Cigarette', + 'Paint Brush', 'Pear', 'Heavy Truck', 'Hamburger', 'Extractor', 'Extension Cord', 'Tong', 'Tennis Racket', + 'Folder', 'American Football', 'earphone', 'Mask', 'Kettle', 'Tennis', 'Ship', 'Swing', 'Coffee Machine', + 'Slide', 'Carriage', 'Onion', 'Green beans', 'Projector', 'Frisbee', 'Washing Machine/Drying Machine', + 'Chicken', 'Printer', 'Watermelon', 'Saxophone', 'Tissue', 'Toothbrush', 'Ice cream', 'Hot-air balloon', + 'Cello', 'French Fries', 'Scale', 'Trophy', 'Cabbage', 'Hot dog', 'Blender', 'Peach', 'Rice', 'Wallet/Purse', + 'Volleyball', 'Deer', 'Goose', 'Tape', 'Tablet', 'Cosmetics', 'Trumpet', 'Pineapple', 'Golf Ball', + 'Ambulance', 'Parking meter', 'Mango', 'Key', 'Hurdle', 'Fishing Rod', 'Medal', 'Flute', 'Brush', 'Penguin', + 'Megaphone', 'Corn', 'Lettuce', 'Garlic', 'Swan', 'Helicopter', 'Green Onion', 'Sandwich', 'Nuts', + 'Speed Limit Sign', 'Induction Cooker', 'Broom', 'Trombone', 'Plum', 'Rickshaw', 'Goldfish', 'Kiwi fruit', + 'Router/modem', 'Poker Card', 'Toaster', 'Shrimp', 'Sushi', 'Cheese', 'Notepaper', 'Cherry', 'Pliers', 'CD', + 'Pasta', 'Hammer', 'Cue', 'Avocado', 'Hamimelon', 'Flask', 'Mushroom', 'Screwdriver', 'Soap', 'Recorder', + 'Bear', 'Eggplant', 'Board Eraser', 'Coconut', 'Tape Measure/Ruler', 'Pig', 'Showerhead', 'Globe', 'Chips', + 'Steak', 'Crosswalk Sign', 'Stapler', 'Camel', 'Formula 1', 'Pomegranate', 'Dishwasher', 'Crab', + 'Hoverboard', 'Meat ball', 'Rice Cooker', 'Tuba', 'Calculator', 'Papaya', 'Antelope', 'Parrot', 'Seal', + 'Butterfly', 'Dumbbell', 'Donkey', 'Lion', 'Urinal', 'Dolphin', 'Electric Drill', 'Hair Dryer', 'Egg tart', + 'Jellyfish', 'Treadmill', 'Lighter', 'Grapefruit', 'Game board', 'Mop', 'Radish', 'Baozi', 'Target', 'French', + 'Spring Rolls', 'Monkey', 'Rabbit', 'Pencil Case', 'Yak', 'Red Cabbage', 'Binoculars', 'Asparagus', 'Barbell', + 'Scallop', 'Noddles', 'Comb', 'Dumpling', 'Oyster', 'Table Tennis paddle', 'Cosmetics Brush/Eyeliner Pencil', + 'Chainsaw', 'Eraser', 'Lobster', 'Durian', 'Okra', 'Lipstick', 'Cosmetics Mirror', 'Curling', 'Table Tennis'] + + +# Download script/URL (optional) --------------------------------------------------------------------------------------- +download: | + from pycocotools.coco import COCO + from tqdm import tqdm + + from utils.general import Path, download, np, xyxy2xywhn + + # Make Directories + dir = Path(yaml['path']) # dataset root dir + for p in 'images', 'labels': + (dir / p).mkdir(parents=True, exist_ok=True) + for q in 'train', 'val': + (dir / p / q).mkdir(parents=True, exist_ok=True) + + # Train, Val Splits + for split, patches in [('train', 50 + 1), ('val', 43 + 1)]: + print(f"Processing {split} in {patches} patches ...") + images, labels = dir / 'images' / split, dir / 'labels' / split + + # Download + url = f"https://dorc.ks3-cn-beijing.ksyun.com/data-set/2020Objects365%E6%95%B0%E6%8D%AE%E9%9B%86/{split}/" + if split == 'train': + download([f'{url}zhiyuan_objv2_{split}.tar.gz'], dir=dir, delete=False) # annotations json + download([f'{url}patch{i}.tar.gz' for i in range(patches)], dir=images, curl=True, delete=False, threads=8) + elif split == 'val': + download([f'{url}zhiyuan_objv2_{split}.json'], dir=dir, delete=False) # annotations json + download([f'{url}images/v1/patch{i}.tar.gz' for i in range(15 + 1)], dir=images, curl=True, delete=False, threads=8) + download([f'{url}images/v2/patch{i}.tar.gz' for i in range(16, patches)], dir=images, curl=True, delete=False, threads=8) + + # Move + for f in tqdm(images.rglob('*.jpg'), desc=f'Moving {split} images'): + f.rename(images / f.name) # move to /images/{split} + + # Labels + coco = COCO(dir / f'zhiyuan_objv2_{split}.json') + names = [x["name"] for x in coco.loadCats(coco.getCatIds())] + for cid, cat in enumerate(names): + catIds = coco.getCatIds(catNms=[cat]) + imgIds = coco.getImgIds(catIds=catIds) + for im in tqdm(coco.loadImgs(imgIds), desc=f'Class {cid + 1}/{len(names)} {cat}'): + width, height = im["width"], im["height"] + path = Path(im["file_name"]) # image filename + try: + with open(labels / path.with_suffix('.txt').name, 'a') as file: + annIds = coco.getAnnIds(imgIds=im["id"], catIds=catIds, iscrowd=None) + for a in coco.loadAnns(annIds): + x, y, w, h = a['bbox'] # bounding box in xywh (xy top-left corner) + xyxy = np.array([x, y, x + w, y + h])[None] # pixels(1,4) + x, y, w, h = xyxy2xywhn(xyxy, w=width, h=height, clip=True)[0] # normalized and clipped + file.write(f"{cid} {x:.5f} {y:.5f} {w:.5f} {h:.5f}\n") + except Exception as e: + print(e) diff --git a/data/SKU-110K.yaml b/data/SKU-110K.yaml new file mode 100644 index 0000000..9481b7a --- /dev/null +++ b/data/SKU-110K.yaml @@ -0,0 +1,52 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# SKU-110K retail items dataset https://github.com/eg4000/SKU110K_CVPR19 by Trax Retail +# Example usage: python train.py --data SKU-110K.yaml +# parent +# ├── yolov5 +# └── datasets +# └── SKU-110K ← downloads here + + +# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..] +path: ../datasets/SKU-110K # dataset root dir +train: train.txt # train images (relative to 'path') 8219 images +val: val.txt # val images (relative to 'path') 588 images +test: test.txt # test images (optional) 2936 images + +# Classes +nc: 1 # number of classes +names: ['object'] # class names + + +# Download script/URL (optional) --------------------------------------------------------------------------------------- +download: | + import shutil + from tqdm import tqdm + from utils.general import np, pd, Path, download, xyxy2xywh + + # Download + dir = Path(yaml['path']) # dataset root dir + parent = Path(dir.parent) # download dir + urls = ['http://trax-geometry.s3.amazonaws.com/cvpr_challenge/SKU110K_fixed.tar.gz'] + download(urls, dir=parent, delete=False) + + # Rename directories + if dir.exists(): + shutil.rmtree(dir) + (parent / 'SKU110K_fixed').rename(dir) # rename dir + (dir / 'labels').mkdir(parents=True, exist_ok=True) # create labels dir + + # Convert labels + names = 'image', 'x1', 'y1', 'x2', 'y2', 'class', 'image_width', 'image_height' # column names + for d in 'annotations_train.csv', 'annotations_val.csv', 'annotations_test.csv': + x = pd.read_csv(dir / 'annotations' / d, names=names).values # annotations + images, unique_images = x[:, 0], np.unique(x[:, 0]) + with open((dir / d).with_suffix('.txt').__str__().replace('annotations_', ''), 'w') as f: + f.writelines(f'./images/{s}\n' for s in unique_images) + for im in tqdm(unique_images, desc=f'Converting {dir / d}'): + cls = 0 # single-class dataset + with open((dir / 'labels' / im).with_suffix('.txt'), 'a') as f: + for r in x[images == im]: + w, h = r[6], r[7] # image width, height + xywh = xyxy2xywh(np.array([[r[1] / w, r[2] / h, r[3] / w, r[4] / h]]))[0] # instance + f.write(f"{cls} {xywh[0]:.5f} {xywh[1]:.5f} {xywh[2]:.5f} {xywh[3]:.5f}\n") # write label diff --git a/data/VOC.yaml b/data/VOC.yaml new file mode 100644 index 0000000..975d564 --- /dev/null +++ b/data/VOC.yaml @@ -0,0 +1,80 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# PASCAL VOC dataset http://host.robots.ox.ac.uk/pascal/VOC by University of Oxford +# Example usage: python train.py --data VOC.yaml +# parent +# ├── yolov5 +# └── datasets +# └── VOC ← downloads here + + +# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..] +path: ../datasets/VOC +train: # train images (relative to 'path') 16551 images + - images/train2012 + - images/train2007 + - images/val2012 + - images/val2007 +val: # val images (relative to 'path') 4952 images + - images/test2007 +test: # test images (optional) + - images/test2007 + +# Classes +nc: 20 # number of classes +names: ['aeroplane', 'bicycle', 'bird', 'boat', 'bottle', 'bus', 'car', 'cat', 'chair', 'cow', 'diningtable', 'dog', + 'horse', 'motorbike', 'person', 'pottedplant', 'sheep', 'sofa', 'train', 'tvmonitor'] # class names + + +# Download script/URL (optional) --------------------------------------------------------------------------------------- +download: | + import xml.etree.ElementTree as ET + + from tqdm import tqdm + from utils.general import download, Path + + + def convert_label(path, lb_path, year, image_id): + def convert_box(size, box): + dw, dh = 1. / size[0], 1. / size[1] + x, y, w, h = (box[0] + box[1]) / 2.0 - 1, (box[2] + box[3]) / 2.0 - 1, box[1] - box[0], box[3] - box[2] + return x * dw, y * dh, w * dw, h * dh + + in_file = open(path / f'VOC{year}/Annotations/{image_id}.xml') + out_file = open(lb_path, 'w') + tree = ET.parse(in_file) + root = tree.getroot() + size = root.find('size') + w = int(size.find('width').text) + h = int(size.find('height').text) + + for obj in root.iter('object'): + cls = obj.find('name').text + if cls in yaml['names'] and not int(obj.find('difficult').text) == 1: + xmlbox = obj.find('bndbox') + bb = convert_box((w, h), [float(xmlbox.find(x).text) for x in ('xmin', 'xmax', 'ymin', 'ymax')]) + cls_id = yaml['names'].index(cls) # class id + out_file.write(" ".join([str(a) for a in (cls_id, *bb)]) + '\n') + + + # Download + dir = Path(yaml['path']) # dataset root dir + url = 'https://github.com/ultralytics/yolov5/releases/download/v1.0/' + urls = [url + 'VOCtrainval_06-Nov-2007.zip', # 446MB, 5012 images + url + 'VOCtest_06-Nov-2007.zip', # 438MB, 4953 images + url + 'VOCtrainval_11-May-2012.zip'] # 1.95GB, 17126 images + download(urls, dir=dir / 'images', delete=False) + + # Convert + path = dir / f'images/VOCdevkit' + for year, image_set in ('2012', 'train'), ('2012', 'val'), ('2007', 'train'), ('2007', 'val'), ('2007', 'test'): + imgs_path = dir / 'images' / f'{image_set}{year}' + lbs_path = dir / 'labels' / f'{image_set}{year}' + imgs_path.mkdir(exist_ok=True, parents=True) + lbs_path.mkdir(exist_ok=True, parents=True) + + image_ids = open(path / f'VOC{year}/ImageSets/Main/{image_set}.txt').read().strip().split() + for id in tqdm(image_ids, desc=f'{image_set}{year}'): + f = path / f'VOC{year}/JPEGImages/{id}.jpg' # old img path + lb_path = (lbs_path / f.name).with_suffix('.txt') # new label path + f.rename(imgs_path / f.name) # move image + convert_label(path, lb_path, year, id) # convert labels to YOLO format diff --git a/data/VisDrone.yaml b/data/VisDrone.yaml new file mode 100644 index 0000000..83a5c7d --- /dev/null +++ b/data/VisDrone.yaml @@ -0,0 +1,61 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# VisDrone2019-DET dataset https://github.com/VisDrone/VisDrone-Dataset by Tianjin University +# Example usage: python train.py --data VisDrone.yaml +# parent +# ├── yolov5 +# └── datasets +# └── VisDrone ← downloads here + + +# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..] +path: ../datasets/VisDrone # dataset root dir +train: VisDrone2019-DET-train/images # train images (relative to 'path') 6471 images +val: VisDrone2019-DET-val/images # val images (relative to 'path') 548 images +test: VisDrone2019-DET-test-dev/images # test images (optional) 1610 images + +# Classes +nc: 10 # number of classes +names: ['pedestrian', 'people', 'bicycle', 'car', 'van', 'truck', 'tricycle', 'awning-tricycle', 'bus', 'motor'] + + +# Download script/URL (optional) --------------------------------------------------------------------------------------- +download: | + from utils.general import download, os, Path + + def visdrone2yolo(dir): + from PIL import Image + from tqdm import tqdm + + def convert_box(size, box): + # Convert VisDrone box to YOLO xywh box + dw = 1. / size[0] + dh = 1. / size[1] + return (box[0] + box[2] / 2) * dw, (box[1] + box[3] / 2) * dh, box[2] * dw, box[3] * dh + + (dir / 'labels').mkdir(parents=True, exist_ok=True) # make labels directory + pbar = tqdm((dir / 'annotations').glob('*.txt'), desc=f'Converting {dir}') + for f in pbar: + img_size = Image.open((dir / 'images' / f.name).with_suffix('.jpg')).size + lines = [] + with open(f, 'r') as file: # read annotation.txt + for row in [x.split(',') for x in file.read().strip().splitlines()]: + if row[4] == '0': # VisDrone 'ignored regions' class 0 + continue + cls = int(row[5]) - 1 + box = convert_box(img_size, tuple(map(int, row[:4]))) + lines.append(f"{cls} {' '.join(f'{x:.6f}' for x in box)}\n") + with open(str(f).replace(os.sep + 'annotations' + os.sep, os.sep + 'labels' + os.sep), 'w') as fl: + fl.writelines(lines) # write label.txt + + + # Download + dir = Path(yaml['path']) # dataset root dir + urls = ['https://github.com/ultralytics/yolov5/releases/download/v1.0/VisDrone2019-DET-train.zip', + 'https://github.com/ultralytics/yolov5/releases/download/v1.0/VisDrone2019-DET-val.zip', + 'https://github.com/ultralytics/yolov5/releases/download/v1.0/VisDrone2019-DET-test-dev.zip', + 'https://github.com/ultralytics/yolov5/releases/download/v1.0/VisDrone2019-DET-test-challenge.zip'] + download(urls, dir=dir) + + # Convert + for d in 'VisDrone2019-DET-train', 'VisDrone2019-DET-val', 'VisDrone2019-DET-test-dev': + visdrone2yolo(dir / d) # convert VisDrone annotations to YOLO labels diff --git a/data/coco.yaml b/data/coco.yaml new file mode 100644 index 0000000..3ed7e48 --- /dev/null +++ b/data/coco.yaml @@ -0,0 +1,44 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# COCO 2017 dataset http://cocodataset.org by Microsoft +# Example usage: python train.py --data coco.yaml +# parent +# ├── yolov5 +# └── datasets +# └── coco ← downloads here + + +# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..] +path: ../datasets/coco # dataset root dir +train: train2017.txt # train images (relative to 'path') 118287 images +val: val2017.txt # val images (relative to 'path') 5000 images +test: test-dev2017.txt # 20288 of 40670 images, submit to https://competitions.codalab.org/competitions/20794 + +# Classes +nc: 80 # number of classes +names: ['person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light', + 'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow', + 'elephant', 'bear', 'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee', + 'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat', 'baseball glove', 'skateboard', 'surfboard', + 'tennis racket', 'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple', + 'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch', + 'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote', 'keyboard', 'cell phone', + 'microwave', 'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors', 'teddy bear', + 'hair drier', 'toothbrush'] # class names + + +# Download script/URL (optional) +download: | + from utils.general import download, Path + + # Download labels + segments = False # segment or box labels + dir = Path(yaml['path']) # dataset root dir + url = 'https://github.com/ultralytics/yolov5/releases/download/v1.0/' + urls = [url + ('coco2017labels-segments.zip' if segments else 'coco2017labels.zip')] # labels + download(urls, dir=dir.parent) + + # Download data + urls = ['http://images.cocodataset.org/zips/train2017.zip', # 19G, 118k images + 'http://images.cocodataset.org/zips/val2017.zip', # 1G, 5k images + 'http://images.cocodataset.org/zips/test2017.zip'] # 7G, 41k images (optional) + download(urls, dir=dir / 'images', threads=3) diff --git a/data/coco128.yaml b/data/coco128.yaml new file mode 100644 index 0000000..d07c704 --- /dev/null +++ b/data/coco128.yaml @@ -0,0 +1,30 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# COCO128 dataset https://www.kaggle.com/ultralytics/coco128 (first 128 images from COCO train2017) by Ultralytics +# Example usage: python train.py --data coco128.yaml +# parent +# ├── yolov5 +# └── datasets +# └── coco128 ← downloads here + + +# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..] +path: ../datasets/coco128 # dataset root dir +train: images/train2017 # train images (relative to 'path') 128 images +val: images/train2017 # val images (relative to 'path') 128 images +test: # test images (optional) + +# Classes +nc: 80 # number of classes +names: ['person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light', + 'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow', + 'elephant', 'bear', 'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee', + 'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat', 'baseball glove', 'skateboard', 'surfboard', + 'tennis racket', 'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple', + 'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch', + 'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote', 'keyboard', 'cell phone', + 'microwave', 'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors', 'teddy bear', + 'hair drier', 'toothbrush'] # class names + + +# Download script/URL (optional) +download: https://ultralytics.com/assets/coco128.zip diff --git a/data/hand.yaml b/data/hand.yaml new file mode 100644 index 0000000..50e67cb --- /dev/null +++ b/data/hand.yaml @@ -0,0 +1,19 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# COCO 2017 dataset http://cocodataset.org by Microsoft +# Example usage: python train.py --data coco.yaml +# parent +# ├── yolov5 +# └── datasets +# └── coco ← downloads here + + +# Train/val/test sets as 1) dir: path/to/imgs, 2) file: path/to/imgs.txt, or 3) list: [path/to/imgs1, path/to/imgs2, ..] +path: ../hand # dataset root dir +train: train.txt # train images (relative to 'path') 118287 images +val: val.txt # val images (relative to 'path') 5000 images +test: test.txt # 20288 of 40670 images, submit to https://competitions.codalab.org/competitions/20794 + +# Classes +nc: 1 # number of classes +names: ['hand'] # class names + diff --git a/data/hyps/hyp.Objects365.yaml b/data/hyps/hyp.Objects365.yaml new file mode 100644 index 0000000..7497174 --- /dev/null +++ b/data/hyps/hyp.Objects365.yaml @@ -0,0 +1,34 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# Hyperparameters for Objects365 training +# python train.py --weights yolov5m.pt --data Objects365.yaml --evolve +# See Hyperparameter Evolution tutorial for details https://github.com/ultralytics/yolov5#tutorials + +lr0: 0.00258 +lrf: 0.17 +momentum: 0.779 +weight_decay: 0.00058 +warmup_epochs: 1.33 +warmup_momentum: 0.86 +warmup_bias_lr: 0.0711 +box: 0.0539 +cls: 0.299 +cls_pw: 0.825 +obj: 0.632 +obj_pw: 1.0 +iou_t: 0.2 +anchor_t: 3.44 +anchors: 3.2 +fl_gamma: 0.0 +hsv_h: 0.0188 +hsv_s: 0.704 +hsv_v: 0.36 +degrees: 0.0 +translate: 0.0902 +scale: 0.491 +shear: 0.0 +perspective: 0.0 +flipud: 0.0 +fliplr: 0.5 +mosaic: 1.0 +mixup: 0.0 +copy_paste: 0.0 diff --git a/data/hyps/hyp.VOC.yaml b/data/hyps/hyp.VOC.yaml new file mode 100644 index 0000000..aa952c5 --- /dev/null +++ b/data/hyps/hyp.VOC.yaml @@ -0,0 +1,40 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# Hyperparameters for VOC training +# python train.py --batch 128 --weights yolov5m6.pt --data VOC.yaml --epochs 50 --img 512 --hyp hyp.scratch-med.yaml --evolve +# See Hyperparameter Evolution tutorial for details https://github.com/ultralytics/yolov5#tutorials + +# YOLOv5 Hyperparameter Evolution Results +# Best generation: 319 +# Last generation: 434 +# metrics/precision, metrics/recall, metrics/mAP_0.5, metrics/mAP_0.5:0.95, val/box_loss, val/obj_loss, val/cls_loss +# 0.86236, 0.86184, 0.91274, 0.72647, 0.0077056, 0.0042449, 0.0013846 + +lr0: 0.0033 +lrf: 0.15184 +momentum: 0.74747 +weight_decay: 0.00025 +warmup_epochs: 3.4278 +warmup_momentum: 0.59032 +warmup_bias_lr: 0.18742 +box: 0.02 +cls: 0.21563 +cls_pw: 0.5 +obj: 0.50843 +obj_pw: 0.6729 +iou_t: 0.2 +anchor_t: 3.4172 +fl_gamma: 0.0 +hsv_h: 0.01032 +hsv_s: 0.5562 +hsv_v: 0.28255 +degrees: 0.0 +translate: 0.04575 +scale: 0.73711 +shear: 0.0 +perspective: 0.0 +flipud: 0.0 +fliplr: 0.5 +mosaic: 0.87158 +mixup: 0.04294 +copy_paste: 0.0 +anchors: 3.3556 diff --git a/data/hyps/hyp.scratch-high.yaml b/data/hyps/hyp.scratch-high.yaml new file mode 100644 index 0000000..123cc84 --- /dev/null +++ b/data/hyps/hyp.scratch-high.yaml @@ -0,0 +1,34 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# Hyperparameters for high-augmentation COCO training from scratch +# python train.py --batch 32 --cfg yolov5m6.yaml --weights '' --data coco.yaml --img 1280 --epochs 300 +# See tutorials for hyperparameter evolution https://github.com/ultralytics/yolov5#tutorials + +lr0: 0.01 # initial learning rate (SGD=1E-2, Adam=1E-3) +lrf: 0.1 # final OneCycleLR learning rate (lr0 * lrf) +momentum: 0.937 # SGD momentum/Adam beta1 +weight_decay: 0.0005 # optimizer weight decay 5e-4 +warmup_epochs: 3.0 # warmup epochs (fractions ok) +warmup_momentum: 0.8 # warmup initial momentum +warmup_bias_lr: 0.1 # warmup initial bias lr +box: 0.05 # box loss gain +cls: 0.3 # cls loss gain +cls_pw: 1.0 # cls BCELoss positive_weight +obj: 0.7 # obj loss gain (scale with pixels) +obj_pw: 1.0 # obj BCELoss positive_weight +iou_t: 0.20 # IoU training threshold +anchor_t: 4.0 # anchor-multiple threshold +# anchors: 3 # anchors per output layer (0 to ignore) +fl_gamma: 0.0 # focal loss gamma (efficientDet default gamma=1.5) +hsv_h: 0.015 # image HSV-Hue augmentation (fraction) +hsv_s: 0.7 # image HSV-Saturation augmentation (fraction) +hsv_v: 0.4 # image HSV-Value augmentation (fraction) +degrees: 0.0 # image rotation (+/- deg) +translate: 0.1 # image translation (+/- fraction) +scale: 0.9 # image scale (+/- gain) +shear: 0.0 # image shear (+/- deg) +perspective: 0.0 # image perspective (+/- fraction), range 0-0.001 +flipud: 0.0 # image flip up-down (probability) +fliplr: 0.5 # image flip left-right (probability) +mosaic: 1.0 # image mosaic (probability) +mixup: 0.1 # image mixup (probability) +copy_paste: 0.1 # segment copy-paste (probability) diff --git a/data/hyps/hyp.scratch-low.yaml b/data/hyps/hyp.scratch-low.yaml new file mode 100644 index 0000000..b9ef1d5 --- /dev/null +++ b/data/hyps/hyp.scratch-low.yaml @@ -0,0 +1,34 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# Hyperparameters for low-augmentation COCO training from scratch +# python train.py --batch 64 --cfg yolov5n6.yaml --weights '' --data coco.yaml --img 640 --epochs 300 --linear +# See tutorials for hyperparameter evolution https://github.com/ultralytics/yolov5#tutorials + +lr0: 0.01 # initial learning rate (SGD=1E-2, Adam=1E-3) +lrf: 0.01 # final OneCycleLR learning rate (lr0 * lrf) +momentum: 0.937 # SGD momentum/Adam beta1 +weight_decay: 0.0005 # optimizer weight decay 5e-4 +warmup_epochs: 3.0 # warmup epochs (fractions ok) +warmup_momentum: 0.8 # warmup initial momentum +warmup_bias_lr: 0.1 # warmup initial bias lr +box: 0.05 # box loss gain +cls: 0.5 # cls loss gain +cls_pw: 1.0 # cls BCELoss positive_weight +obj: 1.0 # obj loss gain (scale with pixels) +obj_pw: 1.0 # obj BCELoss positive_weight +iou_t: 0.20 # IoU training threshold +anchor_t: 4.0 # anchor-multiple threshold +# anchors: 3 # anchors per output layer (0 to ignore) +fl_gamma: 0.0 # focal loss gamma (efficientDet default gamma=1.5) +hsv_h: 0.015 # image HSV-Hue augmentation (fraction) +hsv_s: 0.7 # image HSV-Saturation augmentation (fraction) +hsv_v: 0.4 # image HSV-Value augmentation (fraction) +degrees: 0.0 # image rotation (+/- deg) +translate: 0.1 # image translation (+/- fraction) +scale: 0.5 # image scale (+/- gain) +shear: 0.0 # image shear (+/- deg) +perspective: 0.0 # image perspective (+/- fraction), range 0-0.001 +flipud: 0.0 # image flip up-down (probability) +fliplr: 0.5 # image flip left-right (probability) +mosaic: 1.0 # image mosaic (probability) +mixup: 0.0 # image mixup (probability) +copy_paste: 0.0 # segment copy-paste (probability) diff --git a/data/hyps/hyp.scratch-med.yaml b/data/hyps/hyp.scratch-med.yaml new file mode 100644 index 0000000..d6867d7 --- /dev/null +++ b/data/hyps/hyp.scratch-med.yaml @@ -0,0 +1,34 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# Hyperparameters for medium-augmentation COCO training from scratch +# python train.py --batch 32 --cfg yolov5m6.yaml --weights '' --data coco.yaml --img 1280 --epochs 300 +# See tutorials for hyperparameter evolution https://github.com/ultralytics/yolov5#tutorials + +lr0: 0.01 # initial learning rate (SGD=1E-2, Adam=1E-3) +lrf: 0.1 # final OneCycleLR learning rate (lr0 * lrf) +momentum: 0.937 # SGD momentum/Adam beta1 +weight_decay: 0.0005 # optimizer weight decay 5e-4 +warmup_epochs: 3.0 # warmup epochs (fractions ok) +warmup_momentum: 0.8 # warmup initial momentum +warmup_bias_lr: 0.1 # warmup initial bias lr +box: 0.05 # box loss gain +cls: 0.3 # cls loss gain +cls_pw: 1.0 # cls BCELoss positive_weight +obj: 0.7 # obj loss gain (scale with pixels) +obj_pw: 1.0 # obj BCELoss positive_weight +iou_t: 0.20 # IoU training threshold +anchor_t: 4.0 # anchor-multiple threshold +# anchors: 3 # anchors per output layer (0 to ignore) +fl_gamma: 0.0 # focal loss gamma (efficientDet default gamma=1.5) +hsv_h: 0.015 # image HSV-Hue augmentation (fraction) +hsv_s: 0.7 # image HSV-Saturation augmentation (fraction) +hsv_v: 0.4 # image HSV-Value augmentation (fraction) +degrees: 0.0 # image rotation (+/- deg) +translate: 0.1 # image translation (+/- fraction) +scale: 0.9 # image scale (+/- gain) +shear: 0.0 # image shear (+/- deg) +perspective: 0.0 # image perspective (+/- fraction), range 0-0.001 +flipud: 0.0 # image flip up-down (probability) +fliplr: 0.5 # image flip left-right (probability) +mosaic: 1.0 # image mosaic (probability) +mixup: 0.1 # image mixup (probability) +copy_paste: 0.0 # segment copy-paste (probability) diff --git a/data/images/bus.jpg b/data/images/bus.jpg new file mode 100644 index 0000000000000000000000000000000000000000..b43e311165c785f000eb7493ff8fb662d06a3f83 GIT binary patch literal 487438 zcmeFYbyODL*Ec*DA)V4KCDNT2AR-_jqKI^Ncc+4McY{cYlr%_p2}pN$ch@`USHHjY ztovT;S?_xPdw=KR%zS3gKKtywPs~1NhP&~*c>q&NTv8l>Kmgzc_yg`15Z%O_j12%l zMh2h<000p{g)1o8qRR#s;EfbnZ1OACDp zrMLGaQ2TpLexKRcnc0{*0f3#AjgOy|gP)a+f(<-!^K)I_N2wvPif4y5P7$TpJnjO9wb!b`Bsi z@?ZSvkIrnYte;^1cn_BT6YL)h$NL%opSr--$@s1ReNO-~Vg8}tW7z+|@c&@YvzZV6 zgUj%++XQl<@??Vd0>j10{8);zj_CSu`JjxAYeZx0Bc3MzYk_)P$STH zFwEd3CuqTayBYxitPbdvKRlpje{mTo`47Dm=uaPlBLk>@ z(Lo*HXaSl(JPj*>x&%g#S0XBdw$Q(o);tsq5Im`h= zpa8N0V+1gOifV#296_y87w1gHgu0Am0N!VJRz zNd)hE0#NX`4S5YC4ak7{c>oE37-$bOj23v`9Iysn1JVFL%o`YSKp5Z!`Po2zKL|BU z6l4kT1wMmbNCw`+JcXeHEJ4jEKpi=O9DoXD3vvnA0dO*AULEnA_ zGC;3>0y2RKz~X+^|CL~V^$ZaHIsYaDFsOfe%P=_I!NEU3KzwIlrfq2Z&OzJEUSHYH z*j(RMSKCbgoxYW&t`R2WAvmw8%rzYoVhbd8gChYE3!n5c6AQ24Z-KuioqKpcy8bf> z-`$-|0&x08cXv-H0ALjM?rxb8fVc^ROypn&^@If6pV@fWL^wFOxS04jIK`MaIoQRS zp7B0sW#Z!$eawYKDCYUMkaYhLMW+xz)0?AllzMq|b%_iXa zULSBy{ykg$AI}^L9c^7x3O##mGYVa8bA20a3T|_A3Ii7Q`{xdT0s0IaivY|k@BpHF zDd1dza1I`S(G~CMaDV8Cm>|0M@^4Q+_(caXA{{vA{8JW@71ZNbS?T+-Xn%QLfFAut zm%pcD{iVMK=aYZw1b^vD0O>C}3MgL-^y9t!`?}2Ti-U;{%=dcS&)MJ^^WGzRVE+BI z2Y$r`ARSrt7acil95jRt#P@UheYrp0d-VI7)fE7$Spl>|us%|}Kj(kvd5{Y(+ZO;H z8s9A-;~Cl6S@E;5SlBXa>sjjPGwWKKvp8v6v9K|-vH(zFCo9nF`gRmL`i909f|PqT z4U`nddV-XSTr#XOR-*bw#xGoK^xwL?lGk-H)#cNp6c(a@I`KQ1Tbb+IX;V0vn_1ZM zI|)+$63!3edol|p#V-~+Q$b24nb#DemNxnnT+E!ztdwAB8$APlx#!~llm)K@DgRN@ z(b19Fk%QUN#*l@LkB^Uqm7Rs1oeAV%vURqw({^IAu%-GV;W;=WY>ch!j4drF?j>sL zSlZhOQi3)AR>|Dzuh{?5_CJhPMwWJ#wnmo!3;W;OzZ`a7$ckUoMqk@b|2dciDev9R z#LB_M$|nCOAYukXhu@4cmU_kp&i^ape@L^mk+-xo6QukdXHio8t(O13^zRD6R{7s* zV*%aD@{eEd;ok-lqWEw6uLu6?f&Y5ozaIFn2mb4U|NnX5-w;pV0*oLW!2k}p+X0-w zt)Aa;4+RG^E5Q3qQU>lmE(HK#5{L=_OH3m{0>-*901OF?Fz|qlgO!_ug9JthK)jDZ zSUFhPNMLyW^8)&Z>9;JJU$QXK-j+vFJh&RsMwth^gcarQ7GXFgMMX7V%gKsMz7+el zQ3gY;4embOS8i@$XCo)^j6z9Sg#uv#jMNdps2mTV(AKrJ5_$dl<-N>5=U?*w9!^Jp zMasYs)4i;2q8Kk4PfQu!goh&oo48(o+5TsMtOv%D;H_&g>+4$C*nwg9eSA*kXlHeg z{XvXl11bgLx002=-D`y*HLnAwi-#a>z7P`!g6kvn$0KmP^f7g5T z=9e!act8H7dCviWV8h+rUCqBVsRjTj21_DW|4XCe1s6=w(O4~;Jm%RH`4vc{T z037Cki!5buUGnpOtz&Dz!T#$$1oauXD4e~!JE8^Gb5Q_r6MlDh_4)4ZCIeg|O#?uq z#a$zSg$TzBOAia71Yocruvn10*82r31RnA${Z<^j{%7@!gp2|*l!MDS2rMiN94tKi zedmU_-)|K+^#9vK}QpO{=&Tv}dPU0dJS zJUBc$J~=%*zqq`&3j)CYHVZudw(S3~3*7F4z`()5!Xez-1%YwAHyjHN{t?>)Y>_tz z+SU&#**_rSJc~##ZbG8sklV-Au^B+dqvl+oIk-3Nmu3HVhI#$Jvh1&6|JpSLh=Qa2 zKEZ&0urM%SQ^A4>4m1Hw@DJdBr3e2cgkK5qJ|X>0cc7485@2B9z(1r1@DKjp`@aTv zGvGD>$K51=3JY#2V8LR6TTVAP`&kQR)8Q3ubM3sHf~F(dHL6v}Y8|*gG-ncUJ7ld# zYk6cI(K^1J^FFy$%r#!?32|^}>QK3r{}HoddzQ!`T0Y2lTt-~5z+JV(JqQ)Oy?9%B zoTq5@K#x1FM2{?r{)I3rZi7mezT;|Boubi9Wnd&_FSk^b@7_-vf8EC5b`9ye7Ld;+RKVL$i%u0L zGFFFGA!Iip;*_HaoNIxb!cyQPdDPCzN=|rVs-XHfk^@tHi7ffyT&nRQE%EjkuhYRm zH2d1RQ-0C8#4%l3Z)G`N{}wFgDfRJeWG|=SNU{ahe z@m-vHUumAK3K`)jM|}z!M%-^~Po(ule}dX@tGtp$V@ux7{aHIhg8AcUsz+K#o&7}` zC*|;0t=*~NhCE+evtcd10#@>Rj<5uUl6F@(6<9f!?F7_hkh3DWQM(Mml6)2~g76n7 zxQ*FKoZ5-=dUlG*Y@xP?$-Zr5DfaIl${=lW!TpReOe^#?)j!_ zXmyZf)ime{x(ab-=zhF-}W0Atv$} z`pNm>MZL(+gho!*gpb4NiS^mY5(w4f^ECHfE1djXNY#`Rpqk~>M44JY9T18Th))#r zQF_TNUt{xdjiuWE+O&oY_h&h;PyJUE=X_l6uXO$O_@Sm9WnungPn>+hf}q5FKC+Vr zvKSgFvmy_DC^{n^n%l9Z<`G(&jk#vHNe`6s<=go)+aqSh$IG7jL`Picd9NfSx<{I8 zAu^+AE%9z?>Uv$)iP;TFL?}EA8d2I{@xWLY!mn2qM%{8GQXlzY$rE8rMG#0k$67{x z@m&ACKwR=zlqor~#_X!y4 zzs3IXmifI_P0XACt_HN}4VTy`zAL7keE-r|=9hsoMH7$h*YVV6wU_PstXMqtFpy13 zBuwuBB&Kn{43)CRdUUI=Ty)CA**4sLEtOI?p$B@O1&K~KyILLb)(Lo#+nhToSg9h~ zv^P^Bx5NH?u@i%0!~q}udbF0cPqB6a54ZFORhVAI>ynl^2*>YD3g04a9Wd2C%r*Kx zTY8g|DM~-n=59eGFlGeHP9?gcrpiX|;1--7@7qYCxH1#At$mPJ25QxbgPbdB)c7EQZ6o+sI%mYQf}FLtb2wKG&_ zBWr|1Sg#f5)saR@I>_9Uu?(?d@vQXcj+?#FhT zYeBJIWPw*Qov)yX8LsdHSWpdl2iOE7cY`4y->aNMyOD*lZ&?|1Xh9-C#D91yv zraSem6YdyyyJ^2ZXYaq5UF@QO+V@U5HG^RpP)Q$<07!Kq-#NrW?}vuF^GrJT-$p0ZO8{@wm0& zoCdwpG`$zY_gZFA2Ay=kU!Ex~@OT5=x_ok2dGm)uNKc8T#y=m^4tTZH&>As12tBsSf%BKEY*S;gy zo@9PLwDGuz<=hXWZEs?A+o+y%2&%fjJ2(@|+wZTN!g=^3Jz9c7d0Pn=%4ky|@h03_ z9V)q`>F;9ZXsDoxwbA?>4L4@_iNylq%t{ALeao%{$+BZnc23H$OAV(pE-R+RWJYG% zf`gyktVJYsdBlsu`tEc2tz`P^>SJnbw$Nm9k->Vh^R$+VCp0+|x)x z%~9T*W%jxQME3Rjf`lg~@vD6=a{L&3VwF~Oy66ZYUXc%+;l>>Idut_}m#UbyY0a64 z9w4$r&px$yYdl5S{qB@>+T31CT!fE7W5aR2KPb^R)u3UJqJH+nk61;vGH=o!GG+TN z^Os|B%Ck;_tLFT>w(!@e(1j|zU~liLU18R<#mZ28G^jw!;FpSu!qxp=YphnrSCsgp zw;MZ`Ypt?7GGu!#9&CPBRByZxLN~efsSGgo3Z|_NSPA6-dJlznT2IgW#-uAF%*#?> z^s7e}g1ZeMo*t1AYvuxBdofRW-RxnZ%dWHy3k5u?RbH3G7lo;hzXy`mUA`wR}gb`{hG{=`cw-e2GlRxP9Z;vxaFLnOXo0A!?I<2mu zE`B7VD<9_d~^{tef~he11e)Fo~|05sKZh^fxKPW zATsisw7pwk>RRfYIa0}#RhFslaHf~Pu52KK)z?*7%|FFp(qoU}AY7!VU_paD=R^Pg z@Mg!fV@<`Rw7cC@Wmf;^`99K{;RB1$XZ)I;qWR|tS8s0n>PgR|_B|IqIwtyE)9@R% zF626W&;$De6Vse%?Vwq!?NU!ds`NVHAiFvbX(RmFWx6SA?x}Hnp53~2#GYo`+_psS z#%0b<@4jiXw|Mc;?s1C4!3j>qhXrNY>dM4z#(~~?S?)=t>G!*>352&I<@i_31D48K zIKht8}?gfaIHmELs@0pS{X zcLTIs(AaaKx)y=tWd!-T1XQU;Pb>2x(%8x41pxuy(go3#XPe88|6rlH+*2Bc_n&sm zPL`+!8!#G-?*Lpe>#L_umaU1xmqIzZ3p)mM7N6;9Lj?*%WkQ;(E~3k{8=DZ|mU4?~ z<`d|7#}y;b<S!9@-eApp`w=bRoa>?TwfJ1K*AG&cDc@EtuHeXgO-MT1qE>>rbpT?8#>I~GVBBNTh| zx3i;UXbUD!FU)?R)IHmhN^5D*llkamGaSmZZ)7kg5IV#ol;n+z16Ye>PQClPR z#;Kv;d!s%`2i%gNaC2=-N8E=o(Eu8(??$H|9O>XAYTF5SWufSE{jQ7cRR<`-NYuny zm4n+ADHl+Ym`(s)6gG9XZfh$~{%xS7iHBoohwF=1EHa^C@@h^#mDDG#+g_X-`|B_~d2K@nE%zV)~O-#Yledh-lF;>3-nK(fn|?QCvGvCt^ivzR6|&)Si+ zJZ57*)B4BGq0Z8qV!-08*QsOY^JH3p5P>nPoAA&s%SzROV8-~f_yGO4(3fyBRImdm zxhUUANEU16&gl$4GG1$6;bS28TA5?dz1g44YTrN#yWvK3)F; zYi-5AqbX)U%GxrkKS_w;Ty+(T*Plm5XO<+w--Ao@ zP2O6EC~#ZG)Acc&WGtQDyggmJNuVjI2zAacsui`R9^!buVZxcl6s?8bVVmJ>2t-7A z!MjmiIXitnJX&nn@S&L#1GkH_q=+s z09GE1IpeRf_CxYFxOOUh7I$PK2wl{a;O=OwuI!aDe0hchNRzIS8%;Rm<4_I6ZhABlq0J1X#sN`{HZW-6vm zuv9Z~4c|8KP&20g)29=bkL)=l^MOTHM1<_q?-TBTEx9(E9_uoUn9ljj;V~Z@+o`~d zGh$Msj?~3Py5V*tG~dRvB9im+1K6fpjKOU;(WL^N{+3Y%6`v$Y$FFLlrl>}SPnEhj z=$A-k9A@d9R|m7N81o~UpVC`d2V0*jHs0dh#NGjhvOcp^mN?VrCLMCayB13UV18r^iR^Bs={c9IKM<;f*Uybp~`RhqcRPx4BMCKk-w zUQMh*59EpSphcOg{pPx{*Srvdr$42MrEf~vMRQ(I^{u{m94qrFMl|r2&W+f* zNXf*5pu(%~Ayd77F)U-l7M1g}YGMeiEj&Hax^2JI?XYFh!an&mmdpOS6-`!b#J-0u ze3>#<>x0A=6ZA%dhG2Cv+@LlgaatB!H3Z2~e@N(z(ufAf8c23Va zyTKUW`#49+rRkMcD%@Ph(~$}{lJ@d|Ho?(8+r))2!lZe=uC1-Tz4SfJvW>IJ&bQ~w zKD>{!T;o0p8(DgL;@T5(Ho%*X9&X|l(@z^$s^*^{68b*tF|;Y}C#3qM2L~>+>*1aJ z?f>({p*z78aDtI?JgSPTkhw_YV{kQzD<+0R)pY!eQvr|=BWARdA~&s$(( zD$Z5SY9C}_sVFI3&NGL5K%hN7=1!zyQ{REcAD+TRTaY8GYIX8;+T)9Xu(a> z!V^$YW!QS36iSL!&fmXlJ6XLx&|EPKec_mJ29*^8sF zaNR?$Bg1BAy{L}FsizXx?y6n_hGvn|%HRCm&QR!y+%DH`A7Y3vHUl=_yl03;yyb8qfrEDc zf_m;%4!?q4>|{uCsN8cp36kUMHm9|nhuV(r`=T1&FGOit5T%%Hs{$Y$vO@%d3}eXQ zz2V5A58O@LgpGr)LLQ%b9^hpdbH)y~noD(|;Mm_>t{H3ChVpLhscNg71}++1Sf0My zpDp+_t0wlPv6DI^z&uG2R;Q3V(i?h^c&WGu^VK*4fso83t(E8^?o|6cV(|{ZHyJd# zyctq|9=ko??W6bflbM`U@MUsx;H7IvG275*^>H*xFPEbTZ%xG>BaZ&ma=h%Filyvl zH{Hu&1<&;e?*QQi6_|-n-vUYCFJz_q53{!)2*to@T&CUqyg~VC*FZbp9WSqg)X&Ge zv!M}Xwpr|+@>J*!km?x3!_+Z+twYD|UZYbZt+cbm4n@VS+UsMupu$(L%)Ppyclhb< zt9vtLPOr*W&^e>Z_(`|mjW0>z@T1aRr4JfUcrcNZ*SOXyZE|FYtO>49{2U0syN<*+ zDnpj;fH9??0?_W=8i%M`7X^PqY@jySc1Kl0N9AYaDF>&ac86%hj#MLV_t71&nVFJ2 z^}XGy?G?vlis7oH1x++sBkIx1E@>IDkMzk)sWBc*d!flA??RMFr_pO23RsvZCJ(Sh znV)hJNVV7#Tk;PX>EfX~qQ`U?#~M?;ESyUELABMm`rTc5*$%lHxn)0phJOlQoWx$R zu4eSy*mZp(XFEZ3I~Rt{$uE{T*?KB8rbQ-Sh|4g z{bLy!Guoq1_L!ZKnf~3I&si-4_jHy!r#L;nmVXOMENNoc&DBC z<-pX37K#M9^Kr)9iaIklLHD(Nh-cr%O0+8l9{wZEf6Ppe!>_t%5Z6FK7r+ zrl?q)P>g*QB(nWjPpCHk68mWl1{~zAl(Y&j)C*BK3^&78-yTWWC?xy@2pg>4*d=?B zIEnf&JV(%3kb$5}a@}ZgbWLD^lo{66-m2v+)htJ7>J%IxJItK%7TSuPa6D=*V^8L% z0`QDuF6075c47@^>+tYHow$AG$re+Tw!Uo4wkr=2iTES(lDA#om@0{I)xMCrZnX_f z$c^BsZ*2~*yF?!`Pbm0|W+RF_pJoyxLv+OL9CDjx@4SS8&k+=q|51RtY%!z0q$UODLAAdM#V3e!SrAC=$#|%w+T<#J>Za1*-!OEz895 zA+)cTh6uE580cTrx4TX8PR^uJBpK`bh28=5#jjj!$_o!ZbapJXy|XR$s&2BZNwX+# z{@Rzc>RoI?ulo7pqa7@ zWU5*P1pV;yk-pk;krrG^2BsH>B z2Z`J*>4pBuQnhczzT^S*#KUGqTK!4(`lf;-!4VygULSN}OCGWVj&|1!5*=NJj)m$^ zBb>##k?N4-8X;q(&iyJ!S2uL|#;Acm-6Wwnx z75}_$UNb-iK0oo@VQ1x&$U=wC%!h}awd~bfUhvL$fIlZOLrr>a51Eu(W+*_)5!$#` zVDD3^e(PhaBgwN&e};>+_?pXbpdp++lQ6uag8#6=G{d4$x0J^GO7k#DY36Z$Pp zHuA4&m}-z1gXJQHjK&Wk(ESO`mBgbE%jf4|y6MDd0&S}TdURNY$2W%^!SwOFOe+ZO za%tG8{+LgDt7)WkuG%t(Ro+v;A~*P``Vfa8)>z|-e#j8*i&it-A*lUQkp_dH5bk>7 zgAB(k6=SxI0gv_PKB7`J^4dWw@ypBJH97UB!!4`0l+~ZM`}h{9T&WIN+Q|3CW&Owq zi=&jz6_-%Xbt`Q#!=8`kb450rLB97Zy?>F8ykwbHQ!i^7;PJ$>iQm&G40V0!N8NsQ zMHGct_2_ZW9e{2wQsYV(uA$7K2MOcU%_V5}%&T!nCw7RR+Zc_zwR(Lh4QC)e)!Fl& zH@26XiiH(t?Xwf9wqKX=3=Z` zU-hmW6C$(Nq35}w%8x_wUh2UhS!5JR-OhnK;%g#fC*!w|rq&bfN7g1LXAdWqVtaPj z1Q0G#hgmtOP&55j)xwG4#`@A8-oo7W3*(sU4V7cSZ>hxe^)WHWqKS8`@;;%O$seRQ zjDtk)TrYOD{NqLh9 zEyu5nR1(61;m4)AhX%Fm=%$$A%3n@fq^qgpTU(U%_uPMx7BtD#VxofJUv9j{ShCid zD7Cpk{}>{(8vWKJ)Jg?4OzPwtiG)vibh2krLW?G2-f2v~7Z(P;lLks#kZcvvU<_QB zF3FT)`m$6|L_fT;A41~ZuJ-W4PyAMn4WeMYD7whmQ&*(h-fxiv7j39cRKl$syYy#) z;OntT3^BHwjHXlmsqx7a#Y0!WYwYJb2D)bIs*ibjQsYQ(Tu#Z>yZeok> zu1()!uK3%Ek_z~rh^`US4pY9Md4_X*#Di|bissHt%x@D`&3Plsovrk}y`yxzO=~;x zG&l&ib6l(y?!hFI_7RKjUdM`(2g@*d8`%?e(YDR8(aPc=cPX6kDH&H(lV=aA{4j{y zYV0gBpyp7!WjBFEw1+L&3$f*4DIRIrL*Vqp#H8Ov7Z5gA}_<5>kJhR9yehS zsE8WwyNVz9w0|MUtvJ-F3kemHlV+6b*cGuS?SHv`Kp?5`&UCOWZc=j}eUf0hIfPJq zebg;1eIspOriO{7Ij!>Xha#w>D^8*tn+fKGx|EX6;VKnriR3B=ha_WUC(pz8ZEwWU zJT+YGn^mcQk|C;jP!QcJdFqC}_arat4@rG^p0X&5F}3c(9QCEm*8V2lDS)Fjy>zAd z>#mxf7I#xSN1w!Tjrn2^%D8?_iFxPyl-Vl6q8{kW)`iW*bSSjCZc4z;(Qv@30xzGg zYB(gN@9@bYBgqnPyC?EOtz=%-SB_Tk?RjSKad7euvRHlyop@{1F(&L|bxZv{aQEFX zBj4O%$~f$U2F{Pn{dp@D`3JJW8Vof9*GxiR2r@rd*P=t*Fp{=3KimOjEFH3N8)*cb zuC7opwj?b=f))YfoKD=$=YTci0DwwD%%`cH4 zo0WS*q0mwktL9_XV%16fp6eoM-)-}IPNVSobEtr>c}~+!Ax*ni=C115w7ER)mdxDu>%bs6*A%Ht+TPb-8i-wqF??gQvZ6~7h+SnvY}GDm zcs0VZBe5V1zb=e^&9^{~(Nf71^`a-~qYbQeRuw`Yn(Uan_Wz6mEvNq7Ap<&1$m06+m2;h-II1v|>a8!VxIH)glyB1S}An zKZ7uIkyzEfeTKIXP_4XzFWdqC}f1vCu$eHXgfwrN$<;lxoXt5Kl4c2il+|nvdKou(w z)Qk_boE^I}!*>10KhV0jxauaV_TxF3wvUeDC5}eWZ^n92nLDDei?=;!Zl|?wrs(QJ z0wN+AX4AzpJrL|yJ)A?j*5B1cmYSPqfve5Uj`(Z&{ked(1B2p46jLPH^0Lrgy?8D9 zVlQgBxmFRV9_$5j&%w?QMPXcf>X0DuIHRCQB{;FC8IewEc11bmVYufE@`x^wIA|%zXWr@9(lTTM`!`d4Vp%SKx8*b_-BDUIs)32JJh(zEv-z;m@ z8h$U;Jt|D*QYnefkmIiC@Jm}}&4Fj+uvvBVkiaTblNN4AN#G8dn~I3U!G(`paBi9j zX|)haSZqk&!`f81)ZF%4(2JdQ)=rWc)OrGE&atZa(Hyy&`Y4oN@#t(}2ICeRBS0lI z&s8{j^!udbx0b3Ze8Shx>YwIeUIiw z5X*;0z!b+;9rE1#L#(wAG@PNkk3(rd>WSJ}6~dF~icj<|i z_z@euZ}xud>~v8->eX4cn9`3&AD*UCtR5Z zv?ReCeVH+9bgb49nz*+Em5H@!1ofn*vWBR|CaLVPo2*$rJ;V$%0(vhnb~__EC@6 z53P+ZpYQOgeixiukoBG)^0b%>Ov4&9G(N#Va1#HSA^8MZ!TrtL3bSq#g!{d>@w5c< z#8v93E}f887s;~qg<@A4N;Vj%Oe!wgeYZR45ll^42vS}`M-f**>8dF4juEayh?rI! z-k=QWp5|saO750TcPcW$sTc}`ym)zI7PQ< zB_Q}?N^?-`RDwir=*ZG5<~23tXBRr;Tj(Qa8y91KZqlaaguY)YLfk07SEiWjKKX7C zhF1JC7?ZCZ-x`*;i_~<&Lf?EPI^6SeWTgQUW?D@dhKJQ&cP}`#WwH($F6XSRO1=*C9Ja}1h6;Crz3u*c=ZeZC8N4W6@u@D>TeA)2G&l|9KW|kRSL^ceb1$#<5vM!$ zJ?E(%bRGdE!?fYN?wi@8>NSY-H{cgEJAF268d+ptiw^F9cDxlm#ftsxuvk59@?PsN z-yjEN$!XvD2zs5<8k2&QZV9hphje1xz|R7PXJv`|FG)lnxul(>NVpvLNE-_;SUU|~ zg^P!LPH$c#FkR#rs=>%|d986;7si9MJMiZ4Mz*Pca!)?hGDzr=Ym&5Sl0D*}ZB zc6E`O^t%^JgpF7bw#XRI6KLv^V}FX-eld9{%UOuWt6hP{d4`<~P=>tr)p3c@$Y)sh z3_#j{{Iq1hFysKEg@2liak!Rap;u|kA?~2=OW;8Q8AAy7(`vSg&NS`2kV562 z1T%2~mSmC?qbY-;>=2{drK1HVv8>T@m>$%{NtBOA1uI8d(bVbE?Mp^ZGf}Y2)!aecu(a$AFV<*>k?mqs!W>BrT z9m~YhVnJ%rgKA?ZY!)1qiVxO$Iw*`Xjtpf3CEcD5`4QQp?KU!s-*4hQpx6AB-^;%b zvRE+NnI=hZSk&@A$UbPXA(&Ecu{Vo)9)UzDEWlo3KHX%UxZPcH;@>JB6i09A(a}5D zIF_Nn@n%(0uugVk3EWM+1Fmr}#A**3f}vDP=2RX027ZN`3JwmpEJH8fZWbHo(YNL( z4hBn_(+lZDq_)5Q!3)c^-E{}RKhlK1W*DgFx*bVF;|lY;MDQ=G<6kyOG+*)a-_}Zc zJ8|l2{HQqT>gCI}Mut<$Py3o&=PTfX;Nm8@mhIqd-XhELAg%S5zG^r<>0RK_OK zU<{8}>4=ZSbwx!r{X_3v*W2oEBF(Wyiu-%l*rlgW_x-ku7c7J3Fpj@yjk(F*0f?D# zZ#Bipa~_ms(#o`p zBTb-r>cT|{mtp(;r_DFvLi9?Fw@BwP3o}yAhLt5(?3xnri!*9fz62VBHc?bvk2?b! z^{citHE6@E&3yFbLLv@3k$DzUG3HG#63@>-qmFM?Lx&?$~qZo-qbW=M_^~P68{uxh2}9e zRU?>|WJ?}JrR#!Ctg9O{UCTYx{9t}nSsQ=1X#ysF0I!8_ zz17WE1K3ly1NIl`zE=eMtzY=6pYmydU%bwAO5)%+DLwsu#2&>o zn#%u$;7ZPAUnnO`Wv!)XR?Im=yv(q*p*S@I=MKQy<*Z&z6iCmJVN;G7dr~crsi&i~ zmJ7f(71coP1@99=4ZboN+qFJ)(C5H{1jI!x}%S)SB480F3*bFJ;r6;tO5LW4{ZC*V(ko@ce+& zP8?mJUYgLy;yEhR6-%AuqYX3|Z*XE;$Fc4wC}0w@tej+=RTg>t)FmPka*FWfv}hSB zHInx8&Gg3cja`hXZ@i?6B z5%D#XW)Q-mPY|w_@7TK3>rcPw%zwZ`8EL&MxaFltFP+D|SmyXj-pl#x+76cFdq0M8 zmv=$Lyy5u%J@~rCeZo``mkQy&Zfr-6J92Zpq`oeN3}SMfYjW!b2-Gp*=DfRXO?jBo zM809dbZW3$uY661EOCxhBEm-M2(Aj#<8CkIn!j>I zA<6BUv{Jn${v>!?&Nb!pJiaW4j!4JmHIFErdIZMTSV$Od-i;7`I8NPNv?&DgJ)|N(Y?2Z30>*rFC zf4gjjcuQJeL*I;@I*;@)_Rdn^;QDnBUWxU<&Ga=a`l}p`)h8@Ez6*scOhYS6! zf-VFZO{&g&;yDmf2ZC~%lxGKF;pwC}p$w_2o1xn*WQSWS+{a_b5@$`?2d9i5@{L~= z%oMD^dA!2E7*p_Ib`u!eK~E~E(6vByv?|T^_{2*7MzaIICGDK)@s999jO@9J2GRW% zX9hN~hvTTiSc5^%kX|1!PJW8>aGhbt6W+7QqQP*-#FC}`7V&`gBd#>`v_L<%ITNc< z$w1o6b*Za!Pi`^QZ(zY`2!Y4LRk-RlH0=E?ic=1Twz<)As!U4DbKBjoV1;PTedSlI z=PvdR`@4Be$+gi}@+j;_<0Z~!64h{HD`y{s=QGp3`L!C=4;UX1%l^!w&lFqaT^XR& z=`EvYy0*qQJC(C;S{Es*azi?x?{DMA+SEj?RW#32;e{t^zGR5z+c?ZshIA4m;Ff!6s%u!N-!@d(9ZOtT&d2$hA-mx`Bhx8P zJnHFzgBMHw3V`#_ef0L&tQ(niTcp=1eikFga9eNaXzcP@))b?a<&_>7BqzW8#EkAj zULgFsL7`l3&6=7;3vo|VD+u~tuPp+kQ#(wR6Z zWGGx#xRyF>2siUb(HY`WfeoIESZ}8Esq}>k;=X?-}_jnI#Mm7B6 zRoy86Hp=%et@CqFzSol9J; z5UynG-D-lLHm0z^7C(#+JJUl^8`&s{9+>10Nuy}vm7w|@NBCyz^`yEw{pu>-xD^j1 zBCUwBa-sUY$1>*9(S`<>-KcM|JBo{V%BkxI{$&nKVdTWVJ`QIc`CxB!*(F7Dg7QwP z-PE+c!u2;C>QnC<8k3y5HWH7h@&`s-sr<({FABMOHdK+Jr?5gT2tT9tP^-Wq$A_7LZhP;#r)$AaDr8;8Qvw?>Ma2d=_ zf(M-)eE8~ zIa>R?yQxSbfmy^OSg)-2q<*|#i6BOdk4kj?Dw&X$uJWJWu&I?Bnc5kBP*3qbo zimoF%n)U}6B%?7?FLa5m9M%Yd(PQhF^`e=s&nB;PgafFO2Cv!XQP8+og&(S$;B#M0 zeS3Z1x0sb;*B~sHjI*Fdw+6_<(i$sSYU9X=c1Gz22z zF!{%D#a31I;!JyHy{cSGN$EJWa8$~ryj{;su{P4e9b1M21RA5Mcxuw-Lvc34As7c3 zamV$n8-EdLI(k~f_AhT7t|3*1UhUio!=3;qi7TwOo5R*9(s*`duz9l6E$y?&Bys#r zGz{1$)T#Vy(1vS=3cZ5g}ALi|ecq@2oa19mGxbe`WCZ0E|GB! z+beRhFf5q$UKoBNy*ESfu9}juju8C(#re;!Nv}?yGj>NKs|#Ak;XjQ(;FaIAj-ld& z@lK93n;9p#id&?YNfvI=LdB#L8%9Vz;N1^Vv2l3<6(kQXLNu| z!^sd(E@jg$WYdW&mc{HS@bo<7jzJj6uh>@j0pn{|O-dgWEY{-UCMx#v?~*hp3hc_m zc{urkf-%Kuc)!LzA^3mcX{FcYRX_++C?m{#;Z!s%1{tx=+*Wlm=(YRAcSA6<(O=U~ z>+U}-yc_#U{6F!PuKxfG>lRvrt~|twRAxeYDO`sK7&t0A)Y_-Ss3BHpF*s6LlpGQD z8O?iti@pnd67ipftu>F0{sFbR*PN)lyR)^93y(Hn1njRG+g!9`k_j&3Xl41U$Mp}` z58+j&b$^5YG4Te4h}6Srs9d(MWJv%6JU6PBED6aWE>VI0A=12wWfbo0&c%&`TR-di z?0!am-o7Bxd`a*>Q@FR0UhZ3qWR7mTBc5=xDC{=^7d?46AoE|A-X8d!acAHs@LYOi z!gh5pGfXYpayA1gRF@!woH*;+zkmM#XkXZK#@3z^)#KMBw7k;hh|6(zrG2g`h(ov- z?Jfn%#?O-;(ripwykNZT}wT-;OQK#xF+7&+MxY#E8hGs za(}X|HJKoa_H$(%FubXe5F>G9C{U~R$s}W_HR3vMtETAQE^7-}v`ubEN;%UmW^}q3 z!6I9H83O#k5b>sRE7-h6;=NZ;v=*}XGf&WGgyh_5;^N$e+BSKLszjj=K;&Ro+*7TG zqZLjoMtt@o>osPo%37m=_=)h(Lh+5Q#m(-iuIRV-QKGr9YgsJjbO&jgJ(1jdf39qC zfn05$hV|_yLuS)qx`Dn`w^-be7=7{Op(OPAiwykTPj99C3!}@g+ej~baiN#N+Zm2S zDED*)N`OWQBOUT9o$)@Sr)f6>NtWuzRdP@7beWCJYDXAiQFb}(yq|jX@L7#7;lf&Z zo^C%4TBGKgSMGextzI?$H2YZ=Nw~momP3!c4{g~Q4ZvZ?TIcj_HYxPww9}87e|GE@ z?ieIdCL7LBc9vtXKHRo5#@^@aSA)Pex?;h{L9ew74$cW zeh>K5K(s@l$7y{ugXB9zXb*fOU?FBb&m8ox5%6w}@x#NnT9wC$A-mHrWYh^sB6!OA zKmeNp5}_Cmo8$~Sahlz<;?>M*(rIjTZ+EUotfsGNVs7PW7U5-kR7oT%LyhGMGXUMm zIme}X=YYI@{i&H(*zbio&VI$FFmn2bfNsOGjh&BTVsh-Y3_*J>vfWiWr7@!$wrAwlj=<;qdfn;w#3b3Yu@P z$omX7XUSGPy7fO&e{4N)d<##2UM|+&Tf2K(i@jFp`4e5tV|$B2Tgr`=M`*)t9EN?o ze8&~<{{RL)H253hr|rX~{7dlWr#;obhdvu>5wcFnWi0Z~s31gW=8;AvMBNg~%0VN3 z8BRZ+JAaBAv*!xrhT%koFnVXMY*QiHl8jDk;dih-sj3oz;` z>~?d6V;d`wFm|3t)K+E3%3E;TkJ7$wa7TKaUgtY|HqgwyHwy3{+FCCYOYo~(xA0V} zZ>z~;a+Y$kG=U{Bz8h*w0~~uwhy9^m?R?1jiBEBw@z0I7aOvI-y}Y`8k-=qkDPnjV zVSio+9<|+;E?V^E*{{r1W$w$+`0wLShxMy}_#qaVp!lx(;yp`6@CS#lCzpJY$X>~9 zrSjzoEWTz}3h2@@NPhRqa6gwMy+_tX+;R=DS;SK~BYLESEG~5 z9aBM-)(K_seu-d| z*P}0a+UF$L#GeWH zFX8r^;BOX31^4_c$uB3??m#%Rv9f97(xm&_R$1@Km=F#l3J*b^v*KOcn$N*Mfc_XGiaLA z%&RAvV;Nz_GV@3E&Q3VbIIp7r0A@XQ<3rYbHR0_eQq?ret#`zKVuduzhqaB-;ZrA` zBVpzusBN*3L1W0k$ocEQ`i$2;6!AUF4>6j;6*w6L&pM3bj9`#K9S^mAMf);%Bg6hA z@a4CS{8Onxq}q6ESTo0QHP6{gT--*f97a%~(LP=7Rf}!la@em&F2qWWMN$gwcUL|K zEyK#CXu-F$>u#s?Z~Fl0ve|yl-Zay+k^P?>x`&G=mgzPbW7~6lLq@VAD?8&GijcTD z1mIWJ{{Rs!_01FZR`3>&;ajMM{4;nG-YcC(Hkf>qS}>8Lgv1$vT;Oe73j8+xj(=*6 zbK&QKzAWoDR@2}3Ueiat)Z=#CkUTe%T*V=T&WO=8LS&iMyB=~$uh*SF(6n((tXD8h4oZ|AxDjuJPiA5?tB z@Kg3`y73q7bKwR3rmyArnRTdmV_mzR=H!bVI^5gc>G4Bv1%y&=@*QV=@w>V+9KP%> zJ#+pEiJ^EK;b(-uYJU%WKGUqUeLqY1i{dEl?5vhJWVXV_@JmagJ;G?=sFS<7m{KX*d($40D?t)Dbw8G{{X@p;Iuj8{dD?%N3Zc?+P_Zj!zY~(&oenA)N@~$x;}$p zrhmaNJ{tH-PnPFSw}(jZ6}mOMFPaD2w89Xyir|MQ4sntR!3P9>nO?|N?&2>=ANkkA z=QKU}HP&0J!*s;`ztas2Bf+LCe@ek1(n|eNAg}Xia

L*Ns zBO<5GD-3K?)li)0C$0eKImSkO_C-mybrv@h3v(o^%W|v!&ushFipUx{G5yBR;-;y6 zsu|$J51ixhsqKQsW@2-_Ul{BP)^ga*ZZ$^8;#SVzB|i)RDe?mviII*6T3F6vjeR3u z{RCB&3S%nV_9C$~XtNw*QUl=g{&}mCK@0?OjGiz^=BTdSsxi=tx&-se9$aiu$35!f zElipy%5Pw}SmL{O50Vl@Mj0KmfI$3BMPqR{+oYM1OEBjgXY>7P?fh{{%Gtpk=(beP z%Dm&PZ-$@ml4Vp&lH|83_QhpnVY?iTM;$tI`c^I8otOy{g<;vw-hDD_sB+O-8PI2r zohzNxW|hfOS9W*=anil4ZMkl6VcXRlJ;sf#J;Tc+1>xgwmO~H62EJtdzN4Q{{{Vtg ze$p_qmHz;WcYY2=L6Jci^fVTq%lG zvr#bIOo{l6@CRY(gV@)oU0mt=k{o0x3KRe_v;o)JyllS<8jbTG6T>-uEL5hdE7;^@ z@Ep=3Ew0BSI5xL?z#T`*NFDlOxwG)zT_06%5m?z9JMwbZky{w9x#aD+gAxL9)G#?6 zM-|*$*;(D@NMcoNv>dVJhtf&HyX?IXU+}_0v}XqOTTxMEF*U zccj{8hk!q8-`hXoY*1Nv2jYgm4W+&Sw7R>GQ$jleHCV1+Z?4%C_Rj+!W_S<&3p4v% z_%cO<;%A5aKjPR-hzkg=G-<+c4jvY}hDPg*#~3FAfnGu3&xalu)+C-U5n07|l^a>t z<{ildbLrnXuay2Ncy7zXUL~7I)4z3Yw@48S7TS0u0^PC5=DdpVlwI37@leNL5~~`w zU)OW{4*in9Xc+$hYM+AIU&gB~F5|=TYI9w{Y-g4^T^iDMfsdIV%2mqlPK~sUz#M_~ zpNur?OK*eTCxQ`!Hkoo6k0QcMx5U5CVp;sPg9xgW3_|lP*~)*Knbr>jk3G5 z!RTw}{{R4n`|Lc2j#HF8a!yWur_#QHbPAhd;Cpwkh{VMEicRdz>sZ8_Hhu2(T2y)N zCbyds#>`NwjIIZ$ITYPCZH^Ut9M#1G&vw8+IqB|e&YtH~QfPjeWd)tblW75jza-l` zL!5uIr*BVc?d+Q2=L;;-2Rwsu40A92^&ghw@y&A)tH&>!3HgEGpL*QU?`~tdBIg8j z2iT0)jf-7P;Ccm-)CcdD-V30ohJs>^>GvdDf8+If#_}CHXwKPTbVXb=WQ>8>uHjj5 zt9gE6eCvZVeekzPHux;Iidh8cKYDcjy$N+U8|fh4d33mV30*1TX@VN@CL?hy??q%1MmX6)HT)WXU*8X zt;cM(+H0X0jzAY3fm2>tDr}XRafc@v=DlibOQ@_RwYdU9qmA*Y;P(utf(Jv2<7Kha zZcv0dWe14eNk2f?{&ncrR+5pODf_JRnC+d|j_OMNeZ^BtON6<&Ert%uAOp5Q1b$+= z3)?$wIX-D#HeT|h0gk!cxfRb_YjCykirkBXBxn6o1{p>jeXvezt`zx|(VLZxxKV_>DIR7h%sTesOl-kBxDQsdsa@GW9K+MXHPQEr^pB0x)1I0 zMIo(=b@PCc#~JJ1wdA!gacZh~N13vHe7&F#`|6u<8$li;QJnrAd8VLk7MYsbY!_|0 zz+#7u!#~fT*1DTp{{R(OS_@4#QJy<;Fk>?#-k075NSfKs5a7$-c}yMxZL zbl1gyT~9X=fuk?8_a21*01|v>r^n^V9+7nuKgv?>O*VD_@&q?=yg9+o`s{YD^TU2R zGF#kgI_wQ?6JkWli40=_N=q%fO9t$WjjjPw4S5COi8N?jDE;HUea9RT(z)c*w96}p zxYOW{-boJTcn6Z?;~6apLVp_ij29c#rvCtW1pPGs05i?0lT&NmXW5bJ+HRrY=hQ52 zA-NI8WCDm_---pb8U<5o*~WDS>%!>h6$G6=`8=Du9lrq!-( z-bJ{OjlVQ^2^8qTMmB>Yd5RBxk}3D^SUx1a@m{TMXD*ql>RueW@`NIN7WznI3$(c3 zZ+e%IwmHG$u;2k-Nkfh>bu^l5)qj!aJc_JeG+ocBJYiusi9W++bLHOzypB;Ekq7xu zNxU|ABM1B`d|TjO3N4JQr&z6{Z6R9Onk*LJaKA5JdRNDu8~BUyhvDtSL&d%@gT#7S zkr)0Zx?90>&Ucr)(b!URNotxk>^6&Gb*@|9sybX~f-AOwyJKTp znG)gg$IOv;$;8ZrFpMiOXk>4ZG<-}5fwHm4Z0>x zv)GuVB4}SpV*T@doq;R6l|JK> zj@(y;h@HL@vYamOBz!@wdTNx&su@W9Z{}Pt`rsoZ8HuJ zAaP!MpTkXR(rdqmlF6seW7@yidcD@3v8g!$G~Ga7>7~cf21YmskI{Ww;r^W0 z-)yyW6!JDhyAVSVGnF8(PMEH~I~gX@&9%mIR|Ld)`twT-o7SfN*)&aD6*((d{$YMF zf59z3Xg>;G$*AahkM^uhhVr~mrMnpzXZMY$=(1f~M!osAKrzz3F!5*X=lex?ZrW`_ z{t)YpSuKz&>K+u-S6ht+c#K1BviVz#{Pw^v_le@avURl7G~#Y8U4ry|#iAsCyqVy4 z^sYO{>EN5F(^B#Fi)nUXQ!kk%ymyZ3N`~^JjboHQ-Q;o&eI5^tYeVBno=t3)65B)g znb$mjf8om`;lCMcm)a%l=V-9j*5gZ+#(rfnT}2GCbqlx?&wSTw;SUo0F|@Tdx~GSA z3yB!QU14&B!Jg54PK zdMtZE>0h2+F#iC8O8(B8^2j^~r|A~AClW^51;(S{iHJDM$rY@Z7EZv91KR=HBbxdQ zCmP}8tyMJpqtBj6Rx90iKM5o8zOQwzN2hpWP&U?4NV$t%p5J;iH~>UZ6^bSaIV|J! zuVe8Ch?l{7V&8ac0L6<7nXbMb z_@7;#DAch{XHK-%VHp|uUrLtNeoqCS2Rvl=uYkNo`yl*%@Ns)x8$r>$LnK8;-wf+8 z3x+(gFK4JFoPm#0s$@CHIqP3Vh0o|zx3a7E{{X=}df8;-cQ_qNUi={O*Md>4E)_K` zGDZa^1r}>F^KJnAp{<)CfIOoH{u5j^$HLD9+^}65X(LFBd148{+Kzs7Q_0#+GClag zt}6ck$KM$K6B~;Uh+1XuhGBN*U2jFYg57`s5$BkrwvljqmIs`1T#ldPZ7Wo~{{V!7 zd&ksmWpEX=HiRb7bHXYjgX&i-J?q|~ijq&6T^>bTzMP%bkI?scZG1zbTHQ6=hfuke zD~Kh3y2U1tFe}qx9WjEwg1pbeHhwCy*Y12p;tM-VEw>?I20le(sxXP%` z2N^Z9seDY*H0wxh{3!Z}xRHcqH8B#vJLAo7-e=I@4!qZ;X{W`ywt)A#lr}RXBRMH| z2bxzsHWCTr(379kl}b`q*f986QF3v!ZCLq};tz}UOB=}M@eHs z2F7kMIQypovJHFYgW#)6uYh{j?ICBWTCni`x1?zBHN(5yWdR;Jrn@$$fEcIf?m>XI3$|7|7kw^MFC(shzEfcOcjEw}@)lyNbyxvaa{;|S!Rek8sYFV)-;u1QEz0sSfzKGPPw~#F ziK4XLVQEJV{pSAY_pi!%C--xwH#mo3_SBl=;%^XrqHO;HR6fuby z7I_`80_2{6SHF>9rCSjQ#a~tVoS3SrFlzk{)ArrB(JjB=mzu7eAreDvpy&}oCMHrO zS#1Kg=?gP3C+`!x-oKfz0BAzrjhW^ z;r6+GrCV4}Jh!c4jcmD3vaRgJ3WYI7tX3f^_FcH#3Fj|$s|^cCzPIq*!|A$@pKmfr zEM!I&B6ouYB#2@k7OSHdoq~w{NIk9aW^f)W)4C@*B(+ z(3t+yEU_lkX>w4mB8-AT)|b2QpEHBNZ%$8J9~gXM)@*g{RzI_g3wsG+idYAj7FLE? zB=U0(S)N5gHw^3!Nf_<=A^St>+GFTG8Srhq(WaxN__qH3O90Acg3nUWQgbV+oCIs7 zDZB2OK2gpq9Z==7|=3Nfj7-3&BTRlG2C%AQ1kVH&jf~)Z^86c6oM+VeJN9NE1s|#CmsXFeCKeNviw+Nsgnm~m@ zsTeHa6T?^6`WJw_A9?Zr0K>lyyhWkOp!jo4)kdFbqg*ndv`q!ReECY25<9~S%EXey zHr70lFu!Wgg_i#SvG;{MG4R8}dQ0hkD$)Ekrt2D~h|7U@ajI&UahWYtC@865iy9Rm z?r8?$^5wka_HOYPiF`@%yWvgminR>@@$Uig`SExX8)HFjVIb00yB1>tZwJv64; zZEg6TeiI1{Je4-RtD{ z@JnaGejWI#UkQH6KM{Oes=@Yyr)hVW;tOM!XyjYin6H>r7ue+B;2d#Zpg*)Oo1&c` z!nzK>9Cp%b`p%VTsW4ecbh4gAPBN^>$`7VSdBu2``Q1wUf51OMz(K1?9p{3gzVT=6 zz3}_u?v;Gcp?|`@)KWPA0C%_jJ5sY`AbDlU{n*J5@{xcrPJT%Jlss~pFYGz{U28i0 zZEL4#-YocsG*?4uBsiSO@~&Uyewo2J7#_bse#ScI_UDZ}KjB&BQGMa>3tZSc23Bq9 zs(>6RIoqCj1A*yZ6#Nm<@8A9i&HG1bJ|~^$x?7)!Qq4Z*1?IMvR%iJfXLHBM+*I-o z;lca@SanO&~Q#0jsf=KypQAjT0O_YE6WXLXM#Dd zt=2b;9m_CgXwLz-;PnJolR`fFI-{ZSckE^?SH=2)u}~N65)~|0dv1|3#48X1!5nds z#%t~$7y$Y@u7BmcKl&K2oIhtD1ls7@?e~am^+dlC+j%BEs2v-2T(;e<*+2z8hQ5;V za)G5p%06Rr2>$@b>**Y>bICq*Y2FR753UVlg&*FXJT~wy2;GdHM-0dKn4Bo->s$xRQF7 zX6i}mN&f)r4l9xH%<<|zBGz?Fm}QR8Ngb`Q2qZt~!Gx?zvH4m@V7MQ`bIo*`oced! zm>a0jsn5*Y4;-Et59e4^XH;Ukn#puym&*W>K_dY49ep5Rkpb}AV+1&{{8^{hfsX3Xy_IAT6tGDT`6YTIPV2h+AQ z*0EnwTQ8mB^A{i=Z+hvVDJyJt7(GAF6;VblX>u#IQxcWxdQ{p}ThxZ*1Fz#$uT|rN zZb2Eyt$i#HE1KU658ftCNA5FG;z4N94qF5R?_4GBl#-wA1#BE~&l#@n@UlQAhz<$d zGupiF36k#CSAkhpGmI0Q_u{>*cfxT{TC*Lt85%CRP#AqiKh7)WANVLmsIdP4f=2$) zYaT)m4zdsX>2f*x*VH$65Kkg3VEllRd)Ld~@KKAEPxvH9?GF)c#k2zCo)66=YB=`A zeSd||{uU|Uw7>M0c-&cksLR!UX#OkgwM)BV>nKLr#FE7489o01&MT>dT#%xNjc~jI z4lsRwm5FaPw6Yl11ahH|%so$joxjd%SG$23t!^z!FdxH{*Xldh^pk_!{AQ*3wK~f! zQY)k~IuI8m?IFk1;O8HOb~-)Gb4hlHL1edJcJw=jI{t>dm+hK(BHlG@xX$+IpmjZe zrFFgo)NL+xTl+gan4z<9j@&ejxNbi|j{SRAE;b9;$@v~mb0-`u+U#juL}G>#GQ`Uk z21ZET&;J0bwY5zxMa$fyZY)nAVD#ucsympzTUB7c>mnU-fKN_3=DC}#UTa-maAIwhLECVObHUG+ z4^K~8@Vk51WdPjqxNv%}zE5iUm*8X;lW3Zz6nwWy_>U+MYt+S3wKmVusa3l^miT`H z-b*YXZ3!9Aq38P7#-FrQOK;iX$S z9PLr4NTGiWSL|2AYj|%j?jAwA$&4HxLEd(dLE!OUjsF1fNKF3#!wu+atPb68!~=tz z5A7*`r{!O=9}3@W_vG~vr~ZU%Jf?(nk3;OTs;*cnQvT}vkI(-Aj~){6_kzAZU1&ZT zwiecMM-|7EpluB*Ze?HgLFwt!wNqFa+l=6Y&%Jh^wB%xMj~CaZ$i79jtb2rI*O-Ah zE6B!sisE$3+c}{S!*d*x5^%Wyf&S^iug){nqll=YtdnQx_y*l5x7iX|M}g-9$sH>~ z-A|lKp1o>y(dJBC1Ph*{k&u5%V2#p7UYR-lYv(FdUC%}e#=J00_!#+HJdU;3Xqt3! z{L?n#d!K5s}rvuS0luIvAccO@MS9esyvwS^ofc zj04d3sy1J_Eb0pvQ@8wT?VcID#_o9b?OZIhyP9_xm#-LJO;eK)pkbW#01i!A^2C9G z$r;5~ynXJGAlfiF?lIc5r7JRJ9S=cp1O*si6wVJL?&RjYR%kT#SCocW`T1Aho-Mv_f_TP1g`cxGi2QLKyq97--4O&xS2>$fh5!L_2Ui5=wR_nG z2-m3!Pph%bkF4pz$sa&?Z}yA$5#o`HFB4nYv+!foBa+vX+jNVt{{Y`r@6h~1@K=YU zuxqR2x|2{>@&%Km>SK=^n<20O$Q^6>n)rSDOL$vfQxA!BnRJVlByH3h-ZFk|ge%1( zBM^G+WA9&F{6F}MY2bP8o5J^3dMjTAE@4>aid9Zm{cLif#(3I6>&<<>W56yq_EKs2 zpF4-*NWYDJ$M5Ea;{N~$>T<@`_vo+2R$RuzKFkJx0&Aht^wDvpTFVm&6vVuzAQAM< zeoJ^i_MY)gm`}DgkFVQ>7>4TJOF2m#b2Ym5kEMIYllwvVTT+EDB=F9wq+IRD zlS{j{x`3SaQr7c-Ck?FUQ{HvI{w`+OJqY@aA z&u%#tw6{*W9ifw!%N(9hPg>{huMCm|9S%)q(Mn3?Th%fmGd4lU#YY5LhBGk8jWT|s zCgz^t1O-kQ=jbbL&INrwBaO~?1}E2H7_NDHr%WfI;{GqvZLf53a(+gQmA(D{07{hU zYkA>kiUPi9ggS@haNk4Px}9z<))QGyfTs)f1pfd!r>8^Y{{W+o6(TA)-~usEBx-r;trICdL9ey$;8*a=!3Ynw)Z-oSn6H^*vsaM1xlDiv9mRVptmE*=;c-pf zjXo>*hi4~~;SF+I?V*2htH*C}-!^cfSk}@)M?APcje9?ab-h}W^Y)0!^~h$8vN7qEV_ob}$1)wvS3N?W2>dF( zxmrH+)fu`FmWx9F0EqRE6vKQq8REau?W5klRnng=*E|s-hJ@#@GH^5RUQyv|KN9Jd z5$IM}zOtI`H&l6%k@-9hPQb(xNB|D}SFc!I>RLo!@g=x;7>p4cvTpwXduagvAk|y# zZ|w68S5ngLFRvz&IG*Ihzi23U{{UK^T)!`GpFZ^IX3$Rc6GY1b-po8pZSb>Rg3j{q zQn=IgX%5KWOf$8@N)PV@W(6!$AcMPfWOG?I9u$X0d%N9VUbxmY##qIoUEEIQ7aO;S zgtUCKa6UjW!OwWCt|QTPDD@2*_f*o(7%+->?-jvdMpFJcT|mc0oL5z=YPzS2?Q{L7 zrAsV1bdOTAdt-5qku>OMjx6v)qb?6Sal!WKW|SwSW%uZC)WOa>qEX_VM#|gI(#t^& z?5dH4UBM6lIBap;el_PG7?Z_*Gw_G}A>InJv6|Qv^90tATr`9*03>e8r0wAEE9+2c zo+9|K;rClzN5Xo%x@;kkKAbIet77}O#J18Kl~BFf9zAQgweghtq|l3v8faycD7Mo) zDKix~&y^!2ybKe7!~!_Ybi#Ea?u?~ME$VzCx>tiVPZH1JUlP8fqP)RM*x56|@(e7D zu)E1_6druFM&SK3;##llk^3)vVevaf;AQomOs=Kf7FEag^Vg?aC^`7POx_g~3l zz~jAsn)u_v1L7K3n@k|r188?_%V9!k0h$S;ZaLUPESV(Z zI0K$YuUznV!%LqJm5WW))us@g=5poT+Xg~hmOnlLC$Do}{sW3JaE+V4uk~|+wo^sx zvp~HjTLRYQ{KDnMOZdX_6_5D@W9a z0&=8u1Xrr~ixDWd7s~v<;M~GFeFXJaK5G4vz5@Iq{h0LqM#sR~mZ^JrtfjcrX4Pel zo<_Ibu%vNC6iN5nI~Sh$BEJ6c{GSe5QlM;X9X37>uid?sm@&k;9PQoNsY=x~^dP=_MXc0TO5)}b*N_TiXg zAoe}$ox0TRA%UJ*6@0#1C$>1RoO~za?~OXfg>kHS!{JAYd^jxE&6Hl-MRZJ9$dOXU zS*F@S=wl8$pgdQb+y2yl2t0DJ#o^6=Pm<~1{yoLaWzN%r3{Fx&eTvtB@_NvI@;yaf zoR_-Jui|#Od#>qM4Hq(7@_`9bRYAzl7&y-+znV|kPWikS`$=ftF7WNWhN-NLerrn| zF}_KNxtNGj=I&)^tnK58Lp+Xm959U}nUzoUC#d-2Mr}^+-%wq<;dWNWcPT^2_xAq) z8vb*B5qxR!2f-f}T6o&fMXKphY7<^b6zWRc$Y;*<4xF$6DJ`9u#?i%orQzK-A8MbM z>U>`kJ!5|}`YZ82fG>2#)jltLVbsrw{9Lv-enSLh({zXpynw-^TDfm1Fh9Mza@RK* z`JG}B!yocj@RAKJJSPPD+#XbEYi|dZBsq-vQ57KL=D{0!VCOaZWAQKk3RUBu7TV9H z{5bH9zl9>YjpCN(9YRS9EAZ>~FCbd{%1AB_@Oe@VetLKk&Kuk5Z0CU<+0r=FZv<~s zfz; zzi5s)!ij>ZP$b$F$nS#NN#u(1hWI;iZEJ6FsH8VCSvFvFW`*Je<%u{W1atn^df7gQ z#={L&R+O4B{3D_G)5ZQUhSycq;4#^uyJ+vfE18zm~$>3K7{iET( zg*H0hgmrjEu^zpoNN;W}VmT6k-(=I{x!Q{)4CF~W9FSLO%C{aBr6!xITi9Pq1XoKG zyCF@)%F_9643fXjQJ0}q2u9*CI@f(GNjvn|^kBV|eeS2#{{XeW#f^94et|T<5w*BR zwXMrOx2L`A)`G5f>#M-=CjMBjDZGO)Aof1SwQRL2&lk&tjum`n+_ObZc;x7+) z1I4~1o5Nb83Ex`O?XB*<$*fo^+uT^dbvqYnR?#GqM)Kth=XaK}#~$`6;o~WLT2Gbl zf5`i6Elxy}KR|zBW`^7Lnn8Rouy{i(0o!{BAb$`w``g{p>gfEbA| zN$p!}{{UoEl7ANJH&Yfb{hhSu z&0;@%nHiK1{B#9B!;m=o zAIxan!Wg=)!YkSSCHQ~h&&3ZB=^6~0Vz-O*gqu{+F72-&vb(o{qT9iWlzrV`#_Wj%wwCT^Zd`r3F_YZVU(>#2TaLEsDME>lBO- z7XvvYfsQL)3&Ue_Sd4}k$gg2S`W-jY#LGSP$Ii0GSp2(XjHwv+Mj86zx{DcZc1Ih) z3ji`>Y>%h}3g;c*mE)A41wp|BrD$6SuDrOVRv@Vt!5uS| ztGPf%*HP`99-j5ctatWy>kKdEK6|AD_z426+uvIO^1BAkK4F^d!_oJX zGpRi+j*jYR&AY4M3=@-Fbgp5tDnf!f5#G12E`nQ_<|^HIIT;wPV(INJqVi;hRmi|G zGN1f(;PYOFE*#M@!|u_{-3isi$~O*xduKSWn}6V=@U74M5>NJrW&~%$H!~)abG_E-pi)k{{X=;{xn!f(cHr|rKC~?A(X6_PzF}| zgUB`Yei%Dv6z|$!`b#`cEO)@m)<61EKZwl*&)e84FbZ2chaR6!_1fxJ6F-pvI0W?X z$La^IeNWz97XK*(L zj&)>o{oj}aD}Y8<1RQb*JqXYAl?+rWE1!_!C^aL*ZPrVzUUZG}h;t+3sbY5H@c#ho z*Q)$EjTchAxm<|S)<_WZk{N+!KZ=1*@W;csCh$JB;_Iu0MAN2d21nj8V{sVc?=Ttf zO!#+mzV}cp%-awI+)RFbLGjhyr;T-Jbut(xp>(Mr7&}&0 zRaVDr4z=ZEbmb>DkCEVdytAmEk@R_r^GpUnM6mA3$KJ=f^slY|0A}lzmqOK=6VmQ6 zhU5NOgZlf|$W|JuyRli|DpkLTumO4uoM%4WGhc4}0q}mQqG-2~T-r}&FD@c6$#9IT zF6i^UTNohp7~_iW#z}Ho=jqgGv+Mr=0Pby8L;wIIKD}|%C$Q^Z2>#O&L;nB?*pF^Z zSC=Kp{{XJ5w;x|>`?KNN$sVP56~+=s<3B4LF~=Nw`t`5R{{Y%T>K6E+aTbN>+OOuz|r3ZiuhG2d#ylc zk0}TKK5Oy|{s~e_e#0IZXTU9}YfZ=fgk~Sozgd0~+%iw76W7X2eK#80E{&W|t;yqT z9XR%i{EvwL0BK!j_Rsc;)u9p*50!2Tjs`rH!Q-jTP_ep-?jJ5vu0Y{jZvJ%-?K5); z{h{@vIQy{MFXaLp)*gW=HiA|Arx-Q)hY-j8qNl6;&(X5k{_ZQu9sEjjwUxH&D4Ye(rA)J&($_J>s-aO$Zk`21ozG>O>bl~gplJW88z|sVPn}& z=Tf&4P4dYdyyJj;xtbRr@EH`0{{X050JX$75aS$_^9Mfw9zo&0?L=SsS1TE` zc#Kb$a(|Utw<@k4E-}XhgO9?zn$WvCVFhgt=F?Jw=5ZC3(lH0_?h48- zI6QxoTx3&=sM-t1^Aw|CJ`X@odX~^Z0_?O#h@vK07>tsDo->2_R|J)w=8{Klx00f; zJa?)x+~pYL1J^ZLM!OIrH04Pj^3g;804yKYnCQ$ECK10QCNjuz?hfJl*3|bQqtX0L zJddM9#~B7N*zd<`;ABB8!s8&7UAY}_3F}=JqdQ({!rD;Dg8=UI4&+8G|o!L6mHCT00ZBO{gl_@m`xi;3pySN1CPvC;2-=J7s48+ zi+(EnIMZxmEduG+5X#J(;#ZpDHBIMmE5Tw84@_~!e#3#)@wD6N?|s^yMt5(k%Kdcv z&*q2wKJLOJ9p$(PtQ?m+fNyQyy}mq&v4euEHj8hlr5G4zzVohPEX1^*8!>cN5e8mBxz<+@-#|h zj5KACkW?^HxNt~2xdR~Mn*NH0YZ+-cpNm%E+Kc#AKd-!@;eXm&<7b2!UejF~v`PRF z>TxB^VONeCKPq5x(3uauYu@}n`+odSu+#_IbX!-pWW=%CTuy|r=*lJ%fsgLw{cGl{ zZ-#nSyLzx_Z)-6M%%&*_Q@11*8DI~-Yt8Rt(ClZD8=0lCb&#Vxt0Jc0afKulIl;jl zb6U9J>geKgO9PluR)oJF^FLN}5BMn-k`$j7X_vM*d=D)|=tt2Uk@VU4*Rp6IwlD1I ze|R;kZxBbK-E?;)?B>rO-4V#Z9yfaTuf(k{OZ$9&Tso8&I0Tl54t`VoN*PZb2_5@Y zW?TIojM7CXnHv(jMmk`tVB?_~!Q-`KJeGvCvC)Ojaj;*^o!3 z52b%CFnn6^WtOOQD=l*4O}O68z8Cv3V#gx{Mx``XBo|ax3$U(dSW!$(Dja{yCnSI0Es^?HTVZn~kfhK|jyXGIOlKn;lp`ONc(|yfj+$hb>W0CD zqbIM>)^)TJM{e>1fXG4j6|+9+ZO@r+k^cZp+%Wdwg*6(8)d~_oKBN5q06NDE{o<-h z>c-HuNNhw3cvzKBzR}Hmy>V-CV=F>q$OPes3Nz2sj+ON^7ZFV=`D_>faxys_^{)}v zwDSe7^(O@F+x6*>TKXPjq>eP1;+`r*OM5$uQRTE(3zl9EA1Z65wZCG(DfIkD*0OcI zEZk|=5XZx+?p@tCA>2AsZ5JqvqXdD)deWktEO5@FQaa?lLd6DDf_n6*HTi9hkLSz= z7HCLe!OxhZuh5*<3rvDWj#kDA2B_|Q^F+mHdvQq zBT}?&Hre$ZILMMjrT2sLE(Ul#e_FOYW#TPs!;xtk>y1{%NWaux#+NDcE*l_jCbfitq`R+zGyoQ)lFVV&(NT63+eD^x*e>x zmQVr}V!F5uj=5$tK2N4orykW(;e26lc$$xayiY!faKsy%eM;3eOF5VhTkpXvWPk_C zIO~p0b+AQmATp=kD~_EgXjOMKfMdUrWUs~|&x^|@!GP01ta~WX3f*GSI zNcJR*_OEvbnn^UL6N0`DEojc);oroM59{`@CWqonYuH(qEG@!IA(vrdj=+|{^k5r3 z>*+m8&hlLr=I2h;JjfNqA|?O@1auN9Bsjs{o}Dl{*XMj62EH5WwvbzRlSj0_du;i0 z&uUP-p#lauE|dY92T?Yk=2 z02ADO4IJAJw4KrB)WJ&kiS^Eb;p@#~#UI;R^0b0D@dT zYEKn>Irv}VZF1X8)lx|s#S-L+id{5Pvyl-3WM>#TAch#~Fh*;@d^6+Scf;DO8jp;F z>Q@kn6YWjv|E8QE8krrBzE~(C>Z>c){(i$A$b?E@g^HJh3c@`y>Rk zd6Ui=ImjS_4>jXbfuO8Sp<;vYSiscCMNXlemaMuG>wd%4JY^Lv*nRa5)T&eL=-~O@@)7Po_bp=n}2q zf)HbWJ*H(wIbL_MBd!V0r%ox6oR)mg=R$K_G|jeldw^hq@RFl(Ac^E!xG^^2>Y zBVgE2LL|Wq0o)oRgU)O7tK*ma6Sm_~iM%uLFIwB-O)h^n)ZC!EkX>97*J30xTD+07 z1q=fntfv_x75xPKRkqhOLmbd)_j;X-jEE8l#P8-DhY~61n{rEI7|waGA@SG4odO%{ zJ6k(ge4?N%aA8yv(`t-yk4nZlg?u}`IcjNED!R2zPvtAe{{XT70LRY-YA>g*m1U@y zmnX`*lFH&n136oJ*yDucb{8xG$30DXUx<83`#}6N(i2DcgQsi$9@E%#xzh=@Rbqbq z)Dg*Q2mQ3~)K~SGz6;fK8+(|nUPnQl(9ek@Ax>}-P|7-TI#-Ho{utABC|M%VtnDvd zw;6QtErD(AlEorsh0k1)nEvoRE7ijBwMS^yifj7$eg~UVhv}vJw`2Ka@PCi2Z;|dT zM7pSziAn8m@f!R2cx6ob0)Ca#S$t5|?X?*qyLc__n+gTq`3^m$NW%XB-#~iT@0Y~? z0NBU&U-*p#x}U^f4`}JBvc5&`lMT(p(Sh>$H?}`wxJFhZmLx87o^xL@>R<3jzXod- z(Q97?bvf-BK!3AaY7lBDT#xx{ZL1_VTyz01F_F_5&(Yzy%L}S^kK%Y3yo$4T(Vv-3 z@l!#TMX{E~JDD3e7L4vlQ_c$~{{ULI<+kt_h?gE9vyJ@N_JzE+AP@*39l#8O)4prg zz9oObIsX7^Z-G)(c5sp{?9BQO&knXwh?oNPtj@f(Fzh*OOg0!~XyQXnr5@EH?fqj@}Izv&}U2 z$t-hu{!_+0rhs{ov+h&8IM~sKYndA2A~*s|c^tA8 z=tB{MkVb2=(LN?m8XJETYB8*s!lZVN0~I`+qeuxI{_hp*e`rlRNjn_$vYJu5r@KDY zwDEV0{6Q4o416W3Y4^IJFx~0;wcY3TV@nJxE?F-VA7#`nT%)rt&ps{QSa04&Grj`r zuxdI+x#Da8014Pyj}*fli_aabR+jT>dW6v}{k5{8W}4pZqJh|47m@}O;kHPqzd7i> zBf0TZl3QsyE#9F5uJ*XZYM36XdwN66rkq@FR_wWo-uC1h0SR9mCid_dn1{v!Cx;rHzE;H%6303F+SGJR$%jV{LC z{{T(8f#ijyn%>_0D#3cOT}K@8!L~**qej8_E9_qbd_AxHLj8jD-`Vp@(L5XCd)w>l z`LyfOKeaXQ?Hh~pb8%~BrLDS2Vm^5xlIH621+`b-CI)NZy+=pWt~E*ZO{UiRH@>2e4}!k|JQt<-li}>wx);SAGVbo~ z^HJ1cHx^p_7grB)CYPvPKqZ(!cmCqyNWl!|e4aSwD>N-ppS1oB} zHS^DLEGy;9Ylbnxs;}iltf=YzPRFh^vK5{oA5u1C0r(pJN`GJtJ5|@eZ_fyPQ`RPo zTx&PpD)9uGd>GkfB#SoGkSWA&=I%vIjn0vIoga_tNNTr|=4mhr{D&TZbg#;MM2u-- zY12#W-RZl#zvO+N49V1^Cf{H3JLqp=Y4=KY=Qt#Pg?X3l7va1A02%xw8b*x1Y>%kg zIT;+RWy!`>hJK&}>MNkK*L?ZHTrR*k1oX{$&+RX(Nn!Ao-b+PeA=Rx&2_wcx*)B+E z&h9$#_}9QfvT{C+!KRtU{>)w();vGseOvoR#M%$+=+n#$j2Zmw?fF@bR1WwafMcJd zJ|Gr|>A;-syNsy(MSREmE_m0)9wYI-uVb#r(!iQ}NgM8GP2y0bBRYm4401`&y?r_3 zd3^Jxss0Ox-a&N{90-JPYxj2!dSnye)~z#ok=;!l~e)3@hY=0?x|0A9MB z(Zs1-|JD7~{iiiUdHXvzhkhk^j%^3V8moDF71WNG5MNzAxOKfqwD{IZVuCpgml4}d zm^L;Gg$BMY*8U6lw))#n@eZwPsA)Euv|fGP{mgOP-rfwsDVO_DGkvDs=HNEOmomnL z#xR6|>pl_q39o!Dv%m3Q#A{h*(zS@L^uG>h`n{TJ_WBL9kVf~rE$y9cEvA^Fo2Xhq zvIPPb9J76gYvT_9!}~DB;eQZEY2j;)QpFnDz*tEVLem95=(fZe{Dwv==f?7RFM*w^ zEBJQ4Y3wAZVPnj$zaN+5q5Dl0Yd!V;t^G6e+r)k&@Ylni3u=BO)$T4Xt#sw{ZdO?f zdGcf|Fzoqbc_opRA&(di!)OMZ@XtcjVDRmxv*I*EtlS5CrG<;ak~tg1OhWG*1riW9 z1pr}s`L~Q^kH-En(|#F7$#0`*jTQc(6DVbp@l;C`M3C8Nk%XtXa~pLP^wz84Zye~~ z4Pfz&wug4wc4Cg+Xrn$;?qY}wjQMH^#^>A9jFL0;bF?X0?)5znY42)s`m5Nm)P@W~%9HEO~}k$z~_mdT~k-w5&Jqwr%T;YyFcm$UrgR?yh5M&1V$W zq~W{h2LdgTRv+ z&`ah^KUD-}r6A`%%U>Y=*q2c+$K3|jIG4?q31yYQ%Vb26p>jz$3NSr}I#=6&v$dKJ z0FZ!_=DMlR`)&NqetG`Y_n&M1tYXsMQdY}Mj45TgRRr^)~Pufaer{`Fh z-ZnfM`%1a6{h0-rc8uyejE+j+7E^G?ub~J5uNC-F6T^9^JfAdX00Q;ro-xO#y?yiI zPYC=me}no)o2qN9sBI#W1&Cq%z?;qn4?x76jycY2pjV1Pg*jx*aIj}__57ko70e3ukKEG+#C5p!u^H)k&q2OM+!tKmNpWxQ3aW4-;z z@0^*MzT`>R=~fMp321Te!LN#SnEc-pYLkvX((VYr=N@Z*d}Es5gU{B=DS z>TmcP##f#gyEfOXC9snuDC(oBBeQeP59eRApRxzUty@RyEtF=0EKR zIg{f4uF6TQ3Nw<8Rlke77Uz)bRMr{nalte@{Q~Om^A(4)`$_BBNUAP#f zB+sbW7FFeFYPgH9mnk> zs!J&#I!3%><|zzd{$7;2TsGc7Sk5w|jAPorCE_ZPc#4+(XZfF_WzqK|=#HA&W?7>M zI*<-MD@t)WB}Pc=&{j+l6=f%&tuHmPgU)N?>PqrGm~3cR!yuOk=O50xcwRW;!}11y zD&YLanCwQ|n&>p*h^iZ|Mt%KjgA$IaM{{Fjk0;t35w%V}gjU>fn2Q|p%D4xx3I%f( zdepN{ZOZM+x8;iHq0~{Xglsy6=eOMzIgEszfq(sU zOMH>8&@<^&C_`n}PfcJ1SbZ+X*Fjm6+>fi92)#E{{Vu*-Rcl{ z@4#`bwD%Uzrd&IuJ4g;4033lt+oh#xLI>asv z3#ZQzXFP^?M#cv?Aa$-wTJYwgEu_3ntptUbfWsg$Bmxgycdfm@HAg4aBPqeC*&NmP zfczEW2o~yFiy0f8qIi6V1Re-=+J8#&-xhdG=p~m$yp}a{zxNj=c(OU^CU6EiSJB@H zJ_vZr<2Q`7{{RB`E&%e~INu?f2x5X2KRnVk#L%ynnHd3#bHF^;ar;&N$~yl5?CB1Y z`p=0py?OOe>QhdCI9o}#a7L9LNaa=lkntV37&tfwvX)~?HB}__XEVV~!O6!%^SSQz zSSE@+NiFXpoH{X*5|5Afl;hW&4z=wcvfh-}9~?Xyq6Y~kvVq<`+eSAL*RBWv_s3fD z?NT)tmJtDrthxHFeX;u`S-k!{_(ihc%JaNae9}wHZwqg5paM9++s=FBb>_W1#$UOT z@_&;)dRTa@Oj_MP)cvdY84cu`Ec%SbOPQ8g7?KMy0PIB_&p@g^mG&Qk?i$ltx6ySk z^h?b~-tl9%o8)JdO$?0~Q_q+YSnWLEA46Vc`yyGUo#Cs!dUe5BgmE#)1}AArACBM1 zSI{0G(csm558+LI?2omth%O!FJhKBk3}>j|bE9LIc)l2> zjW>Smf)sJoH*ER?&3+a9Rrrr5g}-R;9>d{nbXi$eLs!Eo`)tf5o~5xSn3;ei;( z!(*emIhbF#+&_8H@f--D|#(xudLS_Up zd}@}z_smV^^w01fI@hm>$gxzOkh=0b`Wc-WYUN|`9?;6O6u+P0#laj z&o$vzUlDX!6K{sRZJ=vW#)XV`h}IK@1LX=O*h?Om2Q}%p7y90l9CB;^AGz|RUBYcT zRJ)HJgswn6@J)Q~D-SgN=(KuJsTTd&TO!oJxn$^URM1e;c`Gz`z4Rq3n zyo~2pOBniX^FE&xg!0+`b$Kr4LxG$gm6K-_g2p)`Q|0411A+DQ>5A62XxwhwmFPdl zwZusyDrvI4?5wfG>T*aUo@s-_cDmk~Zu*o_OKy%?QdC(NAx}7PM?Xr_mKdgk%k<7Y zkLz7NpJ=B+SQDb|{Z|#{RmIt8bT9QKht%gZcoi4;=BBDrP8@AQZgNcQ$3cI5Om zm7si6gIkF&^{HTCZn3P1tuX1!VG0g9^u+XMdq*IqBF zSz1_Xch@$@4xI~sBS;BV$>Ld^5qNQ2EiZR z7&!LLZ+~ypn(vM+ zJ--8);hc9qJNTFI6ZU5K#bF$NF4C=a61+y5v=;iB8CxL=J<2V{;tmEU&N>?TC&s_< zOTU3)DK5SrPj{frB#9J~=~|7&lxSEkRV?*Hy|XGY^A%(sjn5yh`d@(jAtbF8f<<<_ zR!=mvR}2)A2G&px(0XV2*3{kzw$n7z29X?lNd%-r%VJL%BcbDt_1%Zh>(TdNHoj*q zTm?D&NgvG*#EEQiZ<57dcseWv~c|5vwkZU&Jpd{~yY{{X}7Q^kf)_)AvbS6!_WE&MTD2%EOflLmjnV;t7s;JcTn?Ww8R=gA z;Xe*|hvF`hr{Q0P{AJ=@LSMSgp)%PvKJv$O-b^E&nVEqc=D(-gui2}>dVEppdWVB; zwKD(`eVjSCiA#(CBDgLWj<^7RwaV+?2EG{H{fJrke^~HjmKkQa@V(F2nadBkBpFy{ z+IdpZupErw;=P5&j#gCTZ|l(cXUAU!{8;h-0D=#O^eq!u zTbty*)NFj`gHDm!`ypz>QSr8~BF3;|PhH5SRs%B(eA2Nb5C$=T2T@;K z_{rV14J0p5wkAOI1XssDvMq()o8ygfCs7{zKmvo1F(}SDkU1O^*Raiff8xe?^&K3r z`C9b{=bGw*NgVb5Jr2(8iiY$Y);XSB@PAMKy=Yy{8IL2ltX<#d{EDLk|JME0zxW+= z{{XYuUxwP|r><&VJ4+2x;J&E@_LusdwaB<;jvJ?rSq!nMlgmjR;|L*=HdWfcGydCu z7(N;3UMJ+GX9GuN9@+T-vIOiK4Oui*Oqe%oKqcba%r!71VyzAGB|S z{5}1Wnl6zQjGh(KG_7Y<7UJS1vb+~IvuRfU0BN2kjvI-mk+w{*NxfW$d^twPz5SMS zPmkIyzsK(i*}-U@B=H2=uCcD_JFb^dNbSthOy6jl6|u2`Le|n-8Aka-u?qY@@itFZ zijuF0e6fdZoz?Zz&gbkDD@r;^_x`^R>!*qQA+XVWBjL?QSDG_C*N`z6ZxNVJCBrIt ziiN=qwB{p&ljWe!E7m?1Yl)$2k?R*Ndp*Xh1;ke?V=Qs4q!=n=yJrms#bO#&kz0Eo zmkbwff5QDI;OEEh9e7(?_OTq{d@1maZm%_c7Dy-5w5ihG*jOl$E^XyEQX;~(vu$NFh|4Pi09HGP zuaCu2#M+YkM`FaqE>*egelxsnTfi!HEh1Ph^vk&I;MDF!FwJvgDzwIF64-eY`IDC0 zFcG*14-%3MM{Dpy;wyY5pW`65jA-%gnZc9y-a|1@v;@XB1_U3QaK?7xzH0IJ{1i{Y zx>kXw{5trBHjm+-9-TrhQ(A8c3w>4z9`;z4J4yb~uNA?#6Wf{Ph)U`mq}dv>n)imXnS)_swuxafadnEZ{k_3f`Z=Wo| ziQ%h8cpN0=)znqiw*32c(B`L)sFIDk`sw|3JV#HvNp7APm5M|_mA0L!o|$Z5XX{-Q zQL;0Xa5nz6?w&32r;Pp)d_cFl_#LGkTf)9E)vj&su5YYvAkuAML%-~DTgvc+oxxxv zc-<8gs{Zpa*LCru{uG~rb)OITTjEXbhv2Jgk#zPkHN3(*ksM0P4aT6d#>{1B9$N)F z&k93f8-oqT6jft`%Y<} z3;Z?Wn0!0o-3s%=m+@Jx!!oFAD~~m@c7ofKE)FGJu6|ZgfnLT@#58ej#wn(fdiB3! zmNhQaQr`W#A3oc>XrvY67&TYz(OSrxLFA5}^;#W9$IUhka~GOZf0kWC?anuCCpgc1 zVyV1aqzcQOrz0GWE8XjJqWhz))FemKWgu?jgWMjO{41K7$goJ<MZ%+$K6^y+%2aRo#JQ7$AEZ@h{rWcOS4v#w$;~ohfj0jPANj1L?(mUx$j>wLeM! z0O=AuJ}7s=%ho^oQa>(#XP5r~iV_(C`|RXrrhLK=r?q}se$iKx+Dc`<{HyZc<2Y;2+h4^>gLq4@suXlqw6$67W}xpwm1fWVUhKX_Ct~EX9wWpl)|1v9~)$PAm3DU$BwQ9PZg|bSp{(@8Q`O<* zdTl?3dF@{oYXwJ$H545_*FH1vD}I&sjfh3C(*gmIVb3@@+J8UG-o7-|20TN2!~y=7 zIripOAFX?wF8%5JkIZ;Qd(|JO>LRXC5s(P$o~@prTKxe1idkXs?zpdwji`;FcmDuK zLO&dnpKAC7;6fN2?dk8{zW)7+_|bKCA24h67{Ke~FHgh0ds(c%a?$!81zuJ^P5vLT zT~Mhx1~S9G0rVOA*XG~HyS@Ja7;mnWfKBX~A639S^yy!vJ`_}r`cW9#3Uwb(&c87J zI7B-@;x9V%m@~fumh|u9Ke1AqkfeTHi?9q;KR4C z+9Hx3JBdk@kI45?@)tE_<|VpKqHP{c6>Xq){K-NpWrf=^W3u z{`qpt`d6J+tm)ZB@JT`X(MFh9uGRccg9GQ%J*@)gJ1>IONyw!RL_)J+QK{`yV7 z%C{oAifPv2BS{&5i6p_}*@yM6cFiwyte#X@e8d>XYNRGYz=Ag(N8v+UAdk#=ri_2N zzyAP5mZ(x)KH^0%&lqu@80V<`REVCnvpNkOd zw@_qEE?_Ow%!E1|pYKH3s&{vqMeLvHD`_jQ7(csl?emO}&b>Rru$i>vk{zUc!PN7{ zNWmSr^{zh3)9sVVW>h+{+NmKsd-mg0+d-&qdd5E+#@%;q&8Z5wbxHgrh0QAdGdcg4axg!~XymAhEZA z+C?stGx=ct>5rI$a2+=ehmQTbz@OpeO%ENp%9*)^>k6n&y_+`_ISnl?PAT6v(BuEx35W zWMT4{*_da)UMuW>*%HN{;|GC82O6BskaPb4EawgRezou7)AmmNpX7XwI;(h^N)PQT z{pkLczi00%)8WL*yGQz5xCG!Fs2~dV-3i{u;O2uC@7}Je?4?HGm${RYPalWjUo-q3 za|gpzY@Lg8&C~&o7ySMe^p1(B{iosYgRM?NUF%YeAG&N{*;;N$W&PRmKt+F7L`7-Wz* z<2`Glw$d%(D;!6LZ<2$I*jyKjIqRHMn2u@OIcZ4 zWVMH;N-^?WOBu&TZR7G_f1PyJ*E0zfaU8qyOJi@{9>kB*yo&lWwhXrnrBh%WgI8h6 zQ}4$;#X$ElAk6m6IQz)KhJ8C4d{@?Z`HsMUya?y(j8-sjn&_$QXxUFM_L>(Q4B&nh zm3WM@nTQ8#5`LMjyLB^MB=D~ELUPIT41wH_O6M>47>kab#dLFtQ5W2MpuONZe21jD7vSLWs z5WwMvJ6HCD@ox6|*Wo{m@4Pv1Z1p&_4N2{;1DO1|MYH*DynW5BgS)p{{y%jO*&E{C zo24bxdU({9RS&)!sP0}&^z1@bZ5@ABzTLqESa^Ex$5Do9JLmo~bEw$#+L_9XIIaUfqpI`TU4 z(y^`E+Fen$0%sD(@^U2M8#&N=Q zBL`EJSo(~sise?aqOO}iaWwDS7S~AAptJFAh}UlfYL7j=!xQs=%1&nCWu z_$Tqp!k-sK>-bjM{{T)pd5dwpNSryFa<0J~hm5JoZKu#z=Q8+G-^FdQATvg>+|fxFQpi{Q&v9n3y74o zmgngGyn6DgLdM?jxn(U2^Vq8fUO(QgZA9KZqMO*Igah}m9CSU@w-v|fT9w4dFUbUC zWAv`OQ@du?jHH=RA;xGJFo%N{{VFV07{N>gtLX` z)_jqLPbu7ip4k0yUTZdm9FW|~F|}BM!6c{y73%sh#;b)GEfcOjqwuarOONdl$q_qF z;xK(MILEDN&f=AfCa&y!{i;~b(GCVxl$OVChmWONo)_B~0ALTwy#rFvC6XD^GI9Z7 z#|OSEkGM%z4ZsW@=Yd|NB^fKDlbxdtGRG^_e_G~kyvOlG!7vBe)t5XT802%;pXXZU zd#Pps0)x~ZGyXL7v1@&BUz3xtt{0rIAIhS1(a{>Osc(JMd@GD#{IAOM>~cR!=xjAR ztp>_-Xt_zi6Gns%V8O_a>FT>Yi5X?b;AY+)4h6qz1FO%AWc^5Ysay}$8B(+oO+y^+lq3mPBH<>;<Wz58qUW{ z)GQJ+mwD&0x@N{cSm#MC_j`#zKGo`86}^jB)BfFerQGSxfR>JN9EDkVKQs*@ZO0() z=cWfr;%B@JBD7u%@}gk(Qb}+8bxl#cf*Ey7OK_W7c>pY#N92w-E*r1CdU#ysqWI%; zj;;$8^!DBy0^TnK6z()?^L=6{@}|@%D;8 zw5;_zn8Zn+F52QmL)STm=w8LndMHeK`hAWAJj2@4Ry@2;Uprnb^$kDacf;QiTh?E~dkCj}V%!FGX@(DkfMH}-7PMUx){c$V)=j_fkr!@FuSC-4xb zVFWe~=*h@fVK# zQ{Y`oO`BMAaI+G5!{5{9yQ4mS;t}@g1#=`3_Rf{hG$- zfwzcmgl>9v2dzQz2T=$1Yw_jeQ^!A*px-HsGQTF?LC_P&Am_bwtTmP6E`+>H({bCT zhrxfbc9-yvL%!B6eidp{>W;Q{M*4e+9!C2^ZZo0=3561O!6*C705$d}i!K?p4I(Bv z0`(7G_3$t3F>3~s;(cb=NEw?=flfDY>nSB$EJi|sf)sJw*WErXnZL9&au3RH6Z0VA zzDBgKHqUi@TAlUN5!Zo=&GK`|&-wf+^vo_HQV8UBtm6!S=M@)SjCDW%*Zmv#!>D*$ z_Pz1`zP}3oC1|(45YcpI2`$~tsaX80R1h{OMy4v^;E}zQ*~KwS0~L00jR4 z()5oW{3q3XPw_j$@pxZYu+jBBbHlpKlSM9}rQ1OOYej}xBV?29cN0%}G_+#N- zxvv;>saE<)d`o$MtY7ND#>&$}n^^HS zv81J@oj#X+^X!Uy_lxZlEz$_vQWmbV+a!xJEx-+5e|#YLkK>=(GvVF-q48r{(QfqX zo83m!LA|%%JW^Rl2_sRwn&KEHmtDU!&7pXYUhw=M3huP&HIK2p z_Yz(RWW0tJ8S`ClE!4sUgZGXi!tcpA2QBIz2Dtb&;ZF$IXlvlx2^lXf}=6dTz=kh%FN7D4a4(W>y zlWv5`kj%_wk&d0qc%xC${5Lw@>9AeSmX{a< z&4nC?j^smt=tLo0hC2N~f>LS`5!I-Jz8P8ta z)@|0f<~Y|Ww4Sb{f&Tz}kyW(`zR6|gFU+9)-|qFuuO!qwRU~%lYpTZ9=;wEz_hMB! zE4j(S!+VCrW!k-vgB{)qUCL`Bv z!V~mN;*Yjc;?UeTGSnBNt7lMLJuSZ?~I9P%Qy2S1iZEA;2!-TkJk@b=0rF6JAC`#gNZB1&@0yPzbI{VVd*#`jW5 z`)+u$%z%Vyh^0qR3fm&_?djhFzgWUXUjG20@ild_KVv*Z^0&j6wT0F-{Kglut9Y zF7~=!8|EzwDeKU2_;X(#YZf|z)vvG7H$wMU7}yIDiCfd_Ut;N7WB$)GVBkh@dK0(w zuac(LC4&82!l9PpUb6M4feZXd8(T$R`ac{B!zcg!x&^zKw!*(K36&2xaX*^ zvwvlF(=7i0wj$H5NoX}6DV7OI1cU$(vV+Gt%C|YMTOCo#4lMl-4-}-2(@zS>t9#CP zi;u^pepq~ByM+8#)VqU!a7f4ZS%4?s+P_V_8GRka$w+rA7}R{rj{dnd`CIW?U_LGC zO5Bg_3$u?=z5f8EeB9NQ&j*Yr`-t1RKdUd;0$sncXTw14yp0kvTc$~o$v=&Ll6)SR zHO8Px{{WUqG4*=6?Ql7t_UJt}RJ zLZwbOe=$iE^6%UaBRMsNZFbQ!^&09;3kg;6*srL~WCD!Z8x@zUd zRy`>$B}G`3h!`C`s#x!CzS3>yCPZL(a2S5IjSTF=o}!#2mF{;&z;J&-)~YnS7tW5- zT|NcTUh)XjJaC|=L+ZGy&mzWfC6od=A1sPL@1TF3L~qT%h9aGFwoDv!$z$nU@7vnu zH-~VK%mf~J=Yj=v)}v;Osn2s#S#9oN6FiWW2P@_y#78*$%EW);weSb+bZ(o*nq|eh z>=4=EnH-UVc*g_Z^RKffxr!K?bexN5(D1_k(EQ@>^(eU0q2Y{m~L`5l~0C{G*1*=sRM)j_=@Sg1lp=?D%8jy(d!hBMNP(m_f>eNd zZs2^G_Nag0p&E2Ce$qMx>?#8v4_(Cg;O$$OT|nvFlla%gUMleN!=Xv6s~zbGkSPI% zasJkOoSawaam7mMuE_bhVdFR}pHEzV$KMt7*vh_}YbvSX;JQ`a$55}Z{+08G?NRV{ ze}&%+{5Pn0hzPZPTT;0#515Qn-ULNW;WD`{#eoN=0rahphMyI_Cuv$pmqypGtrjN{ zgpOe>e<&ZpD0Lr93g$m)-xzqa;$MX{y-UTn{{U@$Do0pb$ciR!v}_N&#uqr+3h`Z( zF*mCizUC9cwFfkQVBBwu#z?lqVyrTegMpp=dh^f^PHXfB_BM&6__5&LNq4@wXC(D; z1<5C#c_W(q{JUufj3mh;5f>eZC5IXB#~rFW@o9DlBc&V{JPjWXOZW1Zf-bj5e#VY4cJ zTx4VPuNBZ4?j|Bu+MsfH&N;5`#%L{-mvK8sRyil!pXXngVxxC+=;JGBcOPm007U>C zuq2%QDI&g8DB~(l2d#3gda^{VpOyIo+NZO;fo?zS}k;NC4ctWY<%s zNZVrL3Kmb-73Z*DTZKE1e_!ccv<^0+Mt_B|^cb%$w!5A1eNCIY){TB%)mFy#b_a0d zZfiPgg%OhciXpg#V;DK$RxoR!Qg=Tt{AUlv3$KgbGx7I@?yaNK^iPP|CWANGQfMwx z4NBs{Br{w?B4x*(Sr$cPF$T^;?LVRa0JUGm?}~p2{w4T&-{HT*>wg$%H}>~$tGrUr zY;9UO6>a>ZZ5yjIp=Myf5)=`f;QF`Qe`ntV{5SEAy{ve4=1be%I$QhG2?cI0Y@l~X z+vX9toy0!C!*)6f`BTB)7-zrNt+j6(d_VC`thZCzdDCc9jY`p8<^^cft((Pi`$kuY zTjvZ3;Ai&U8F(txB@9YV*64Z}<~FW9(v}vMl+#wxU)_q?G~JqQJM~vSjs2(mJD}-5 z1pG;<*#}5&v~5Kuibvi`EE#>0F2cDCM>*%xzn*V}*B%1+Kk)AA>%%%tv^G}uT6Sz= zB@yQ_uwbgpLT=7mXy+N`ziNMJ{{V_!4e?LxyW`dH8r;FCX#N=0{QE1bON&d3fAn%9 z43{W?v1D#RZf4u86@}juX8EDQ}(r^ zh0+&VEw)GH8Op6sk~oCO)+6BZ<<0IG3-8Lf-}YmJq2(70K&@B?<^k8V})Fd z1xbE@U_U>HUrNE1W70ecrTA9w%SmQca@>FdZ@|d|yX5^V^a=^{E1w@Ka*B37b@4WV z;q6NDX>2~uVlDF9-F(c6{ogUyErHyUPpx?K_;<>GFGgp!h-4LlR|)|f;dxQ`4uci+ zz3+x>%&yjqT%3X5Ae;_IU$1)dy>~^rw7zLDHYolc3t;p5SJBa^z7)>}v|lykk2Q+Y z{7)P@g#L3fgP8*7JF)kaE^F;i*r!0&ZN4hOb*C&%0Ml+IoU>&8mSU^&gVp@doCCnm zHS$HgGEc6@1bnf8{d2gIM?X$0^k?>8vy$hX;O5+SjIU|#f zyw{JN&HY|T-!0`>xXAr%(X1rW{0VV;ZvncVJ9aA?XD-`t0*}Y}*P(nSv%b@H%a0Cf zR?}KtYPxl{r)w)le5n!*$ufuOy?3rb>&_1bxxWWXsA_sj+F7WJY6Z#Uecp5USI}A= zhuC#(6D&4cXlxD?j%>3AmGtb3+}Ria50VYnyjT6cGfZQ3Op zi8<@&2{nhJ+*^oJH2cg6Ao0&?(Z02oYf~Z>`LlozTKJlh_q9ERvC#N>c;wR|mK^c} zAI7`uClW?jADbLkllX1fV|6Eyfq~k+8&8@fh-93cy!Ow`_81C5g;KH ze8ETqI6um-8(DtW=LBe-v%DU0jk)5#AvCYr7U#rTm96KE=AXou7Ru1u-V4bruAoxe zn_{9YZgbv6$>S%1Urcyk_MGs*jyro@f5cjIM0T(Cgf`lYVLsu80Q@J&uY=0+T2-5CO7R;z!Tha*1loTe`%=o8Cu%jZ4+HSS&%Gp zHlKFHe-D#3)4^-5C#o^6<3E@ca zt16hJcTKUnm2ftJA=)x>Ps@OKu9DMNjwwn?rFSk)n{{Sgm``1~g z>hRmVmy!pOVpejIqdT_1P8C-qo;@lfN0&#q?peRU((FPgVYWa|{0oYUbf039c$4K= z@^jPB)9qPqbenk;e69-sdW@6BIhor?mewvyBQF)DgQ?NzP{SNwY=SVZ(IM-|`LSH~ zwcy=1R`B!_YLG*B=knfkVfYS4K45To;~1{9M1@A5hv1?{PvXbh)Yh2Gcu!AKdGo^K z+_}YYx_r%@@RgC_J~8-pV=dCyM)LihB?ytT`SC8`H)0vYkpe*jJ1gd&6?_o*fAE_` z@n*T8jlAdYt#mENiK2m&?M#oUUTfxm z+Qtxm4{OsX*nGPMKBWDoECbVmJvvvnfyY*alvJIYvCUeHWUkJ8_FwRKg*2^WQ1M5H zbvW+fiM0^PK9*zhq>5Q% z3lWkA<|5;JBNWf^5976(R-w&Kl{-fMt>3P{*K2kCtNA@J+?-se)(Zza>=eR_Q^;!A5L5hdl-+Nzjr-J+64g#>EK8|7l)x95k0e`;Tg zI_JbK8{)T*HLYvJ{spquUf05Q-aL*LvAHc_a~<}vrCdfJloyU?i$|6jOluP}yhV!9 zmb$pBRpxw&U9ZV}?0DFUmR3uD_yHfpo6n9v9Q-k&{?FbD*Yr<`lV4tGnvSKeY8rKp zmwF4yv+cH8Lzlc%wntHH#@ZPl%nlZ1h0pCqoACbtz}6oN^nVq2iswP`t=^lfKAUN& z>bjHZvs*r(ie|AoWBj^pmDGz7k{E=Za;$PrNy}IB>7;9)0b$p)C;gmnH9Nbze-UXq zE&a1ex5l5^D;$ztYC3({x`Rx-i&DEWBf@^p(9FQIL{ut&N1w9i_CAvwvHsA$5!8G; ztT?jKt*$&x;s(7s!g-Bt2A_Dkq;a&8UA*!~aJGvw2@36+b-@P{Q^Yybofpd(-$i@B zUbg#_TAcW5i95gR`uULf=fr*w(|>2EJ{kOSZBtd(^y`TGwCbC84#r zYoH~WV)Ki`o_dd>RXGsV0)QvHtI6S zOEuM;aNDfXEF)@pjCL^UI{kaV-VXR9;mvpUt@wfD_!c{>%Ny&xCtB3TpP*RHBSUU{ zySe4cK_sfPTe$hsZ{0~A$5pSQJ_}ra&EE}vAKrMk;l0(Lias6q>dRBrZgnjRNNy5I zH8~o2C)90h%flg)F<^^wOpNM-A1!>HToqhoB(T~myZgFd)vey2RJw#eXhjl-W6)&nXHa<=9dU~KH}-b;wBH##3jY8VJ}2oKzl8K#-|X16 zNW4d8EH!BF5m6#8wv1;@=4B z3O;1MK42ZeQ~vTQ=2nRgJ94Hj^7|njq{?!_iSbxDizAsqDJi9x84aA;G3`Vi3h;0e?a1>xJ zF^((pGsRaHmcQ^vFN}{n1XsNA>M17xvv`v9C>iQW;8*F-?OlJ5*aPE2L_rPVa^UBT zA1?~we5*c)hNAwoo`Rbyt} z(ev{Sqyg9p{RjPu;E;SS@k}xj7C#v3>7JZ1jfnko_2Rw?@a603AMjEM9$lqVS-jeR zi{%QVc6~4yeQWP1)RU=Kne305sdr|6fqv6EcaMG<{>z&8#NP;bmhVuTN``$zY1-6Q z*H)|{Z{CgZ#Rb59iosjWAQQ@({Mq=0ZQ=-Z+k>yIjM{)iM>gJNx&!4YD!aV(!0D1e zIL{g5^)L44ceVY6d|r_f8skR1CmfxH9h)QE^Iw(!02RDKZ+Di#E5Sg*$a?zwD!K_N=On70U?7a}Yl_SlM?gsri|6jQwN#J$N5O z)ULH{IvrnBMomIUU7uSprAe_x79QqIE`F&seHMhhqf8oxR;vW$B!s>lW?l)&5 z-qJ`IM=6O|qr$E?ZQMYVV>lSEhdg<$>Ng(~wXG{vvxZ45ujNs546h1H=5At4Y_jb+ z&QueQ#=I)nN!g{)@HnCz)K{tfV*dcaJ+7mZ_9*yeB4Z7rBYeF%mCtW_{T}#FC>o{e zXRHAA{{U!L;GgWP`&VjT2R;P+2lzjE@b6jI8^_i&+UOTrE#-{fV_AltVRSCq&O2xF z_sO%&U{x5Y$i;q^XF-*at~u% z)`vaTnF((tu0iNybt6Aq^ImiDkM@=LQQ{AYx0l`p@vfP14yx%q2-k7M96V@Ltalz; z9pjbb1adeztv2|F2B0mVI(zYWeX)m|c|nsYzQ{YwWm?BlOQ zd!3EgM-nnRh8S!P2iM-VV~JYiDzmaeSrW6*?65f$u+(m?9u90AiA^{cxjT*gOH*0G4zZBF}P zZFPf#7Pg##83;U9eWZ^94?Xu47#Z(fb)$P2#n??kTQ-7Q zN!uILp!E5;>t7-M)VfR;vS`LP+4d`TWd(DUD5H-C;cRV6E z&N4yDxZBsz*UjIwhM}+N`VNnEqQqWU^=;CUh4$h(7?KB0M_V;9GWzc%?-=j`DzXw3vJpyVaVsySKVK?laKgbS`*+( zsp089T#=vOTM`KzjP&c%^R6f2CyYO4jc3C*M_Bkzq=^;{$u))ctEbGM0C<)wmTdLG z!64V@Sbj^TBjqPHqwb#__)oM@vM)D-ZAjjwczom zX_p2`q?H&08>qlya!DOMx|;S~M^5+y;_W8d=fJ)mywfB`^6ntKzn4&66-t=pmKbIj zIR|q-;m&i{#y%%_OT(HUg?`(m+K42EXw^t|VTU*%hHgRq4R=wEKWv|wlc^_3-1tIG zt>besB|w0yEUvRz7n`BJW|04 zC1T~e4q17xvwvi1WnYYb6_F3!p5=%f91B~Fp4c_@7}fikdLIFet!#A|+jf6NzX{_h z;X7+9L-Xw}SZ9{N&;I~kzUR>G6@CWzDk#V@CZRN&N%={VX^vYQjEuXKGRk~Wu6 z@co^%YWI++MG8E}*nXIN8k!wPN^7q&Dd9Vpw2RmCs`r|_(|CT`YnBWo+m(JsGhED9 zT1s3R7XozKt%Tyo0l)R*qleui2Mb1De+OD8HVcos24Hn02(T z>96E}LoKH1TI?r6*43WI_#0{P#SfD>^5nyR3E0P;-D~Hc2!78W4(&BzNaS@*KK}B^~L*0cz@!anfo?;L%H~i ztR}B{b>R!mHpI*Yw2oKKIK{++mok8$%;$o44C23@PapovpS2gke}H};{i%Fo;LEFj zhdwKo7YpO#esQPd3QZSeUF|#Xtz~|)zaC6h2 zQ~06qf8jsu_n~Q8*NQcrU&NY@qkAR17FtKyG`l%SLgD8##dmcKc>zmA2`=3iAP+F` z$L&cMgMK3Db~F4m@HVO8y$JoDT^q!97hWcg>e+ye#mx72Fv?jD9XyqUq>8P%hC%cr zgy)Yw9bId_hn4ltODgWn;AHWi!{=6#{{T^&No4!~0NNJvz^m+ClVRr_7}iFru<9^9 z-h->!UPH!wi6WR_@%$u_$rbhI?FHfg0EK@Lyk!T(4}zW*hroUu*Yu4(?kyiop6**a zdxwbm=GiY=M~XK22av2|jDxf98vO6^PlPl!)^BXDpo&Q?WMuQ2G!2;gVU#H!hd8d9 z7dw(-i@mU0a$OgfN zJ?@+N;C@y4v*Ey+PN-nCk-`WvBQm>2GOPhm#{l*?>tCjyuryMAtH#o8KfBU&m@+zi z*E17?><7JJn!!2}{<@w8n$2l?KgplAUJ+<5G?LJP$ZX@e+sEl%%LR$k?{qzCMj%O| zET~b~Hry`+bjDAoKhNfE2U0e=YTL%dI>dPPAoTaIrmkX{Q$>Ufq*@Y$8Og@jPyWqc zl5q@IvQNzYgDZO2^nDKfCz2v%Y#s=7R@_c{k+lA`>5~5dXMKgGaRJz*;Te$o@WH>G zc>e&0)EjX8lbnH!e!c5!RJd~n6!$8AgjeU-on>Utej09i4XwSbWX&47hE;5+3@}DV z-6Z~%F}{IMl@!U5+c^Cg173Zj-At-iCjnRV;C_|VTgy3O%-fXqBQ@q?Y3r%cPN!qJ zwsvqBE`FSk&aO1vb>qEq@w`4-pDYeZ`EtC2?TXm8okU=XWWmo4clvj1p7r6*DC~@u z#g>&^oSt~=NgEx-M^8$61S^w+kHVxD@UxjY2iS^GK=(Q=1LeAtwm<{(&2xw{vqla; zcLW}T<~=JCX1`{)hT=GqWPr>fl0dB3EX<_|8&0{TNsoR(?p;<=g z{K)R%TY~{C_B#Q&jpC@UQZ`!ODV5=Z+#%~n-Hu#U> zba^#x0f<+?COIO(91Y>Cm}JHj=?YGq4A0l0dJ$z8UzDz9M`@)IJw{ zD%Gx+SJLcLSb6P3tZ8cC2$AB8=blL=8Hr#82zr@{MZ}>@kJ+J7p-C574+{cI2Lx-Bqd8RQaod|i5Okppy z?T|Ohe5M~6c+*e#jimSh>6`6E!duU$T?K-90=a2QKK-QTG8inRA2SL`*blw_!5{Ea zZ-AZ}{eqzIPsg8$wmu*6W~Z#`s3F$x;e9^J+BrPvZiVKd6xPN`Hmlmj6k9;exQ&A3 z{W0;M<3EAFWX}%8@KfTihZDya7dDsL{kE@b7+JLY`-xM@p3>uEr1nz7BX3ke_$r7$sy6$YlpD zf)07-C!TAZvp`&or_k#FnryOvdZD`J9kf zBn+;4rCytskA06xRf}m~bM0*zl3gNV%rYX6oc9iM`PPlP?$R{GW67H%>aG5I*CFA{ zI3m*Si?&@ORxGS}Ju{Jxq-MHHRx#68e!I!de9ik?DTCnGh$V1US*&tTLyMS{j)%83^xf*+rnZ-I1bLMK1Z8*`9Xa=} z0{x}*ZC}OT2=zS!PKY(-$+eO(DGG!|IyuWIY<#M5$K}?#aPHMr`=LLC`S129m(0HL zDzPmR+gd6(BLsPpWB{Oj-Uf0(&j*8FSZgxf+Uojr7X%fO?L546Cu?K>0Iy#!e#o8^ z@gIh@zY-m0DFR(v*~0|0NQGR9q}r>s*ea-QOE*q<9X`bI$AkP?qKz-a8jYg;j@w(4 zPivTEig>=&6fzQmNS)N2at}f}5^LPWuij{I#%kwUTVjQX_ z95KYf9!c&Q-dG`4A)0dr&@sjcHR*ak>{;Xqg)8 zV#6R(G3Ou=wm3LBub;db@uT9Gi?l0Cn=KFfH^G{v(9^BFO>$$gOXUVRZ%j}Q>6(qQX_{57-MzeemBh2O?R6E(+s7zb z;faB@NcOr$8Df=#BmV#nLl;?8lxjQnwZF^q@;_l%nQ3N^gFkC;*>~dXzqB5a`#*eG z@T{reYj|!ozY}KqOA8+nTEL%XhfiDBL3CMJ{L6{@g{aWrzrq}Ogo(m~%^t;HTx3_%~$~KG}nOAG?kPy2OoA)37 z3Kjb`U;f*_2rTb(>+ccRTj{#h!+7^kyVU|qb)v&=?=6M9Z+NY(rn*Re)FP4tENLQy zA;v!~J|t-x2ZcOYq5L@T-+{HQ6H@Y|(=>TC+i9ZGQquGz-v0n+N2sgY?~OdQ{m9-6 z4C9P_UNbeIsdC}n-pc(G>ify<*)Kh~s>#)YlheP;^y+-C@hjm!?CEMeN${(|zA>}8 z(lre(@nO|8IBoSA^)C=u7*-1_)YD^vIH8VdCW;B}2yKyN$C?Q}&+Oa#IQXCTk@)-J zYdu$9y|&Y=E%f_e4|swrNHp846$Hm`{g{^)=F(eRTbppcQaZ#O$nY7aUn%@E;|8^= zcpt^qng!kVnWXsFPVnSUD4y3$zkP z{{WAFX?=UaUMbUl9r%uWEh5uXwX|(!#$7v3w~of&eZ{oWi`IE#nqo!4k|vDrX#_`e zZ9X0))r|*UH8p)2+wlIiIH^j`$Lk-)uND6Q!9@Q6XPp+~;z!5-02+-#+rv`F;v4uE zMZUOe5pg0~A+e73NiB7GUS(^KE(nm@7g}k zb3?ZRIo|~m2xo#}xP1f-Sl&Yv%GwVJI{i(ld%`Q(5{5tWjng0L@ z#og_je`afXUaXI&YI=3e$+VImv_{ffNp`b&ZZ8tlJSd<@R&U-n>lM7ZPs1Az8+=F9 zv>V3wnSW@WBJnM>7VwjCs@UnDU{sPtlgK-4lGsaczF5Rhzd2@c^Gk^*N~9Jd7E^q; zy57pyS}m{U{EnE@QGCnr(BM7>=)MluJ_Yzs_HU2GdajYG>AD`BZ>e8+k~4VPt6DOj zK1rHXWV=~ei%j3^)$q;Od7h)-8@~>GNdEwZdE-m#eKH94OC{)vX;om6Lq#J+60@%0 zS8!|?&UbJ!ex25TY9EQ($HA=@+UDOv@bASb#kQZRY4>uxpJNkS3!}bMaCbukw#HeR zhWri8cs_XkFS7ABhqO{0-x#1U+T3tp>t;92-;gwx; z@5XBf#a6OIYo}ZYn3(S(ImaPf>|^u*^ZHfYN6enpqn;ITGP@*DNIk&}I3u_`)+ODw zq}qH?h(ZLOAOvKl0AbT<7_YsRMay%UM(q4%*8C;l{Wtyy?eUw#Hr6poWo@eX+GwMn zGX$2;Uzu&BPc*Sth(?82au{y=#G3s+{j%+CG>`Zu_r|>^K!G8K)4?wlyb&(dMTrzf zh!?g101EuY@l3N?{{X=zeld|GyJWxd)IMH8QqSTWpfk7=ocbKs>BsGnC{Ne}QPC)z#}_%_h~(YkCJRFha3_1_f4jcF@G z<)0$e*KXWnf8e{W9<>j%WX579NNgw^ApN!M^St7Srs~2S&S@Pu>{D zcsqw)nI69N^WKrI>iTA*d#l^KF^WQ}SZ)|HHhIXw`W*fh?7k10SJ7-5Cd)0NB z>*{k}6@T{oO;+)>t21pQvj>F6*2>Bp{N7l=&N14)lPt2GMcnxA9gXW%`F^LN-uS0R z)aTGd+G|?R8`>70IWpU0lx5>=E9YqF#jtr(n){#jR+CJ%)Vx7^;f+Oa3BQr%X?Dt7 zln0R@1&IV6MnNR?ug(iwk0$F(o+(-xP^uz~<&+Y{2EhZbBad!t?7!Kb((gvM)#s2y zA(rK4jiPj#GrBy3x-xJTxL|O@7{INubeAGgKFWn%edFxE*<}&F7-;N(s7qE{oab{~ zN!OhA$E|*P`08;t#xD^&a2rx}J%5LPx6|^kyZ-=WyZhPx4e1Sh5hXi;F@_1WOD5bF z_sOr%Ul!^RABz4X7+^oOB*`QYRSS*7zdbN(hDzM)`rPsOrT+l4AJFIQ=P@4+zhh;| z{nG1hGJ1c+n(=e>!+urv-|e<6um1pGkB-)Nkw-Fr!VjrR71S^f5JruFJZ?q-l=TE; zSLb)^?c*&r*TTQCwv9A%nXYa1uP0UJ zu*b(D!6fUx*(Dd<*+N{4V{H{3hC8hirzWarWbJcMj#$nFZAAm7XR^ zU63eL1&+`^KiWUq9(!i^xqEGMHuWsM6&@dxb9;ah!5+9_PE z%&I0(`(N#o{KFvN^Iwlwt12}sCf$_(09hZRQinWg_vQZpBjBHdzaM@%cyGjBE7yDh zYvRpkM$|k}qIi2rveo=Wd1Tr=+PWgf(!yiAmUNa`H+;+jCIA7lwffiap#Iq&1pTWm zz6*R-_>HFc+rsjAX7gV&{{UpgXM8RqzFTCq1v8jXNVt`cnBX129=`_M#W#ljBKXw? z0Z)s64+3$8@}!E+Z{V?BNA4$62|2Dnzql^7q$jAFIoJJfDJe;>-Yq_sPs?v1z-6_08ujYl85 zcORi7Q_^cRLo-O=0twuqhu5I~b(Ddv;gvkXcahJa9@Nh=+}gdh)19aA$n^Z{dIoi5 zbdMl0`SIY4b_7;@zHG?T4hI7Y-Sg?iBE7xL0wo_Su0T`M9V^H@b>oc(!dg?!;ypqg zHYMlJxtcgv<-t6=EN2F^*FP!i zUq_K;6{**C9A4d-!HmU4#cFDJd%)k?7Eo$hb<)Q3pna?e+DSZaQ~Kh*d;O&Cwf_M4 zR&F$RXyk(WJ>_Hz(Pdm@{{Xe?Ub*n=<7dJxR?5#y(R^K`U0uZ*#PF@;Tf?CtfmR^o z9tiAsu2bUFKW_M0t=vMQL=s!LVaopiuIrJ^dP6Xs9bKH};AGwCqu z4$?@z9mqKc3nSaG=zm)I+u|0fZKrA0-XE7uSR}f&DLjFtQg~3UvB3c0yI>x*>_4^* z>~s83(XLUy)_6|Y!*Vhk%jA9|^sC<-t{Pblm?Z4e!hzqXQD36q?aM2!$HrEZN-24t z9!0Ok;VA{2FbRx~r#@$vK{y_VIs7@S-;Z81v%S+}(JZXvw-N-Dy1sA~v(p^z>7Lc= zKMtaf_g{|Pzjz+-WFB~pbAWNs;@*|UlD3D zo1YiSY|Oa`7mf}~0y)9s0N|g>zd-*0V=F0bel+-I>4x^3Oud+>U~s0|Koy5)85vW^ z&*k(OjVU|!J|8fm<&L8z(Vx*D!UUZ>6JcifMULGak6)W$1Ft8iB=h}#h0;7p8vG&f zu8}(BrjLA51CA2$_bxHdP)|JvO8JlBro7iW1+X}ZF$$fz$-@8uCyM$m(crfD1>t9S z%e*e1XJrJHW}Z;su*d+f9Q8QwUzhO(K5Q(}`kq@WPB%G=9}(VN%5AO4PFUlBd-bo7 ze{3JyD@py7bS(p0)%02Qr`K*SZ4k$C9kejKF$q6$S*ZUKlkw2N9-}5pyIwb z@vr<8{{Z&O@qNMB<6Dh3CIwm@HXC~=@J}FZib+7nAM3IYAlK*BwdQWzV|MJY^cXn< z@D&Ba<<1lrJpkvgap_;u_%0F4=u6zAADX}CbK@`=+O_F()ti4iepWp@#lIK+B>YD* zpB-zLy2>%$tEgRE-Xjuw$9Fuj=QzO%IrOeqO}@Xflr@x}WHR#0Y&?vdWMo7Xeqy;p zb0WSt7-7Kd2VefZY1-Uez^qpcGoPD`u5vM7MNWh%=@nz9l_)K2bsy`H@Et}O_X6#l z@JT=9{xzX)=&BHmu0C9J`VX(^SP;uR4(TTWhDisN&$;hbqI@C(Ku|tWjE;XE^^DfK z9Y*|jNhdIt3;-BDxc>kO(2naX$;Yp$$NBo#I}OXMjpQ**5I_WR+w1t%zd2SwWFY|P zPY0h<&T3jZVO@Qpm(S*1$iwck?ZH2o_wR$pYWnNouf<(c!5<5r!6!EQedWwsK_la zSarCP&ec5G#FsKh6736|g6uMP42)+dfz5Q+rwLL@dmotQc&UXRop`sInM*HHR6^lWp}GWy2aF1k;wR#>T<$3RQ>rR zD6YX&46zEw9N^ax;jf5#ori>Pycw(8i>n*!Q4A(gx?4tPNd%D=9J1i;QaH{51B%$t z?2n8jyt6FuTi$N6hzuDA+7Ff-k&s(!bQ}ZR^`%1&sY8-)spjIV`+BNTKWlX_*-PRv z@HdCNN#GlCdmYY~XDs#&Ad2D{%t0I~$iZ9@f(}Dt?+%suCF4sA_&i~)Sn3F2xr+AY zIZ=ouB#GS}PBJnXNI2u(zoO3r{?y(E)h10h#qckOFX3g0V{;_7ZX)>+yyF@eizre= zg9AJ<9M|)~<15K8yjid7Z-XV?qo`b*9m|<+Ws~JS)Nc7taf4qek>YF0RUuhg+a9hu zjWm;kJc-#M)!Dc1^5Bn#pVq%rKVdjAapM`7^@~f4vvnXzaKqbYYPR{>{|%o2b|w{vT8k^MSYQ<&H?kNjT$- z0bev}ax8uiym-}1q*4^)<>V94divMX{xP|lE87cBH{0m&s*LgrtR<9=naIt4Q^a~< zoAWQXWAVKb>%X8oRrVi(OkZK|sOp^flVt z>M`6x(UX9A=jmUR*KJhL`b4fa(#@kSw0yW9%C(|%235cM*{4|rXs{S?7mRyUY)0k- zv92oB+if#wt0P4&302Qw!Tf7l37M^O$DFrL+>BQ`T!ahgD%>|1v}Zo2kIJ~G6IDLv zOBCgh8l5HBt}=N!u5K%!sk@--S}8KRd5U`U=CU+Q>oj|f(#Qr$+n>Z&J00+m=8{EL zVD0l`t~-O)wDg8;(z5>mbaeVxAZL*+_5qB(bM-ae{{V!Ml%A&_@sm@wUxVH;lJv4d zo*&fN*s&@UTR!ZLPftqzc>Wo9U%(zI*6jZP;SbfPw4VwLx@70i5yQ3e5J?+t$QjE4 zoMaPUynnQP^l*O8UlJr{KfCaruFd}dfUncuznj0=UR2TiU9NbCP%_CLli|zT)?LXf zHZhuEEkKiEe`2O%K_}?_en4 z-oZJE5*VeHH(xAo`4>$1wAydPq2v4g2IK7?WBVL%+Dwe8E8$VYIgI2CoE+qV)0+D; z;$MU9yjP%G{9n*5b<2BXiuLZJzqPVDY>!~86vEm?a_mx1n8!26GfL;nQi=t0=QD(5 z1y@CF{QD7yhb$X!U(oso#yrP#ulBy3robk#yn@O`S9Oh|c;>4MhAMY@x1oR*0TrIa6IkdoaoWS%u^X8SnTAAIWH|^yShl- z?8F5^%z`&*%OLt&OR|IESByM6@o!GhZ}fc~tzpxyY=5+M388}W_T_G7w)<4paN0C$ z7T1ZUiB+5wB^a?k0emd|k^Un5F|_el?K$C{2Ts*|dGPnfvDoUGM~JVZ({41$Z!KC| zyUQrxmf{UfU~AbCB3Vd|Sjz&>76;fr1i#>?-wiw)@ZZB<3qNRmYew-twD$IzcCjv@ z9kNfS*vch`P8)PJxVXI3?9wRy-F*~K6i;&mipvy}1o({Wgz)vBHX5Gp@=tc2nsrHh zmt=H%s;X6WN6_u!?*m=xx85i54UVKDww~pz;M49crMF_k%Mee7iJnH>RUHccOOv5Lz@yuCv`i(JCTPPw~w2^jg7R#sS>1R?vNlg}O<`#tzuSnpvveg>ahwfKAEX{_}No0)e;&{|u?Bo?}@oN4Ewy@}(w6Gtjrz+Iz2{#Y;d$BVoP z;n{8M{wC`;w()7Ygu07a=~`Mvdp4b@X^SK>!0RNj&Z>%hyMhl!AVtjd8cqbVVWQ@n;TBreh1Tf+YUWfFrdO}}!w80)6^S0{*N_-CNsd_%sz)Oy2CC>7>lv2s{w%D(NjqVxINr&H9z zW4W4zwLQAxMm{dLO#RpiBh($=)K|*RE|2} z=N`56vd!jr!I=(9S}Nnyaj%kpYyE24{{Z0>x`m{xBu4v6RbYNh8-fAI?Vd^WuX_xa zGXDS)JFWQ~m+Z-^Y7_iExz;>Ut^g=O*S>#fO)bpPCZBGm>f+wo(g$a`n`zvsET6lH zBpcu6WZkr_dp?V4;0eW|cuvFn3H&<^%zBm7?AI{gA`~#(!26B5MI17yl5r%5dXU># z5I$%44WRg@+feaGiY#T8#$7oqV@#FAx9HN#`?6IT$_R@_%65SHW5@^5#9t5mH+N&< zFB;xlXtp}^)~{pY8<`!Y(ykh3xLe4ru291=O|_saA^=N=T#!R{<``E}rR;f?sJ+uY z55yiW)U-`qygtxeBf$i=4RWrR5ZOFWBRtYbM1D*~vmxBB7w(61jQvre{?MPXmY1Sg z>Ao!Zd*b`;R^ABWyt%cywlGa_iWXTS&C;JO+!jozJ4oPH$KMhDHvC!m(9&o=3cI(m z@n(^zMR3~X)y!9S8ia5_$#l}#m`pG|)rokbhA220aM3p-#iQ`Qg0;xkTF|^xV`pP{ z1ormh!ehT!)*L8+1!o z`M2{wPK4E!o~QIr;m;Pq;O~ho7sT6b89ZUGO>N=Zsb5#CMTSpZpWs;J55OH-WFrx^A!HZy4yaY3IbZ8m6~q*BU+C(i5jx++JBr z{hMgAtmYY37m?++RF2`KbyPo~e+&K-=pP6?C*#|Hg`W_uRV|Ir*z~Clw!5brU9n#n z^$6D12f&@Cw}p~j#ubYWHkild`G#3Wo+4F~vToMa_e;*pU+TuN#NXXtmY;`X^1s91 zu;!2B-xk<>K=`}jTdfPlo++9ed#x>eE2!H?bPLJEmzu1z-Ai#al79aAV=>4;I|7y4 z^{2sa_#`Zz5cox@_%p@&z2E#NFMAfJ9k-Of*zjC5Zzzh`;x@RtUGC;>vV|iYE&#XV zAKDL4);tB{ABSEDzPHpYwI8z09oLPOax##X3kMf_WLDR7rqfj1+DKYp3`( z@yo=%AhbGWkK!FK{t`VD$tIDg>Ts&;nh52@t>;Wxp?E@#rZt&KW-+-cqYPix=+9Nm zq_yA9_SfdR9XOh6?CEb$>&QQ34+3c432%H!@KeSbUX!e7Gh8kAhp%r|`$*MVE9Ph~ z9zhkeNA|Uv0C%jRq>*C)DF{!ee`ffr!rv0@Zu~HOMdLpUc&SCj!p&=WHSwNz`x5F4 zs$0sfhJ#|q9B}0pHd*6F#&LW*@x$QHhxNZ0-+XrQu9czQ-d;tn>T7Pc#>Qx&jun`#f{8kle)@ z$|6}V8*&LSn8=VMfosX3LcLi^6%yo&Z=!egeOF8OJoi_YUZ!7*d^hk<#2!Dsiu&H` z!&fV*#cGzb!h`!pHH;*S6^pXE%8mn06;dZn#R@k4zct(bz+V@B5`0Lt@kRHFb&nP5 z6Er5v#8Fw;E~BKi^T?Cww$^b>V$>azbkd2*7WUzGv2sIjci#~8PmG=x)E~#65A`1j zc-O+ZkkkIqBSfCo`VJY_@Ok5?Pubw zO|97?T&0E0mZM~ir)qk0Lv0*ZPkz`%o;d~bM=>tY=5^f~GLHN;EJW+5S65A~?wb5R zuf*BnC(C$C$o&BD{{V>oD*QX}>U=!Z$(zKNi#^@-#1`T0B72r;tu1BSC}s|2^5T%Q z02Ppy2m$LgW$`7b@aCmwq}%@hV)##2OZyA!a^m6bF5`8QJB>XgjiPrfT(m07GAVp; z!2WMq>%Xmdc zMOVfZzf9A|kXggxFNz*0zO>UdiwkeCNo`{=p7vFOJ@+OLHaN_aMIKCN$OmgOoV)Sz z8Kn$H3lk-Wk2U+PZLQaBG}_mCCU#Jt+Gob!wO_;CSHhk$pT>8(nAMKAs9r~S+XPZu zUq!M;w~;bIQ3R2+PCoQ_EETKd-DfYQV&7?)Nx3(Zvvv;M#eQsOt~vDPze&C=d>GQc zA=*!Wp!k`uZhTLw4Q(!^+|evCNVc-=LO@G_Zz`vjNE3)bM#lr;FN%K)d~5KQ#@6S= zk8N>hd#MXMd#gE2k-Gur#z-oxo@_wKokjx<+m(R*XNLHuhANVzBvBwx^c4N^%B-ifm#Qy-<3&wx6zx)$l z;?sNwyPHz+97>^>hKPOc_O zG3TkS$$N=>kCVq?r#vi_b#{;DeW?0@kCRFZ2qS#)s~tddNK-)g%l906aP679ma0~`$Vn*5B^ zto2PZ;Fp8nQMPE->Gey9W(*`*VS9!#vbg1e4U$Gw9Dq*-y>s@^@n((i?^f|=!Ec0D zUM8#-lA|_0*DGb6-);v%xSS7R$KzjK=pVGN z#3ucx{vUir*YB=pxA5Y;D{RKrM1tDknPx~XB+Q7Vog^Xpl-j#igkB@w3quBUr(#uKl2ZuFX6JD~s9%E`S&u)nmWVe=~S{ZW6R1kW6%1$tT zn*E>bbSR|NZoD6<$9ojBiCKw_2$RTARAIN}amZy2Uuk~de-yke@tgL3wAXwgs#;#( zX!`D*FP3B9aAT5XfNcoTNf=oqR&jzGXNwApZq4i1lF}b68IARFCc}cjqRH( zWLs>4sz5t~0l{KBXNueVQa;KqO4o1nKGosZGg14`nLlQ~8l+#N%!j zNtb{JQgC|X8L!O`8BCJ=P}e{s{y(2RI1bO${yF~uJpPt`&eL7pe#d_gyje6C3U971 zU{{tg9CDZ)@X=x2wvbMDZR8vba(^uRMSBF_5jEC=Sn|4*+tf%4U>53_3^2zT$sH@I z*K@z0=6tslxqAyE`V;+=wJ$Hgf3n0V=K?Pg=Nx~_ocQ{TSK8mUf?8_7u+PSuJL`Dv zmQM)Yd1)$cQ7CwfP6zjq@XALafFi#)zh^%fX}TVh`zd&GbdjgLx7VasK!s2}tjIjU zxH5)g!P~T-lpVw={SE!9HQg%D_D1-g+RX9WK-w3Vx@HKOlIGn&B;&4qYN^BBN7UsH z_jJ?pJrnkrTWv4nG?tOaw)R^YV|wisbD} z#H_N6U?1g7YssRMa5x-v`hSgilYLIgb7@%bqLqpSL%4Sbs2!=dvLeF4j~#jUu5KMl zGmsw)M;)q-#nkNKuN-`%9-pmTCS^9w+g(EPD`Dp@4&H;P9_PJ$WAT^a&w@N({w^ul{`5PI=ab8v8`)DV!5UPSA{o;l_ z!8!V5cCTv*iEx$TbIOu&XMq05ejn1jJ@L!qzNKLta9&+_&r62lScjHnL7?fg0Kjp9 zB(gtGyyW+<1NdPSOZzlfq-Y{!@o$Y=_hV-*uiXvbE;+y-m3yb`(XHt>KO4R}NoR8u zj~|Tm<07N06jkuc9^s3@#QV`v!f9WU7<0?X&6z9>u)cR-OU&3z! z>z*a}qjB~{X=L$f^NET%RG;j-84g5>;0^LKagJE$n)$0=@L$4BKjII>*d){;)9!9; zG-<4?#luGgR|HMx8(3jwRgzCBOn_q~bl3;?jyAmm`EohcMivZIqE(AYxD{^HE2pxl=?5i z-{g2TadfBfqtX5-_}<$4;;edCh4s7nEp+V$#>V0+q%wb|r_UjY?!>Rgav+H zGguxf_^bOsT=&Y2#R0GRJO?A>+Ypi4}Gbobe&!Bd-T@lFTyjA&!MOa-2Vbj#)``1E@9cV?zdzoFTQg_(-b5^srZxh^2Z+UAj=+yZ>XpUe5 zZet{JglqwiUOg-HqxM-Ed><8j6Mt^eIA0B28>wS;U?+H;b}ZkR70Y@=iL1Il&{augQ;&9vFwizq9jrli}pI9t+gQr=!}%ccI)* zZbDw_kXcHt7n$U^G0Z^RhiStB#e3)N1#jc;hkvltJ~a4c;%oD$Xv-Xy*4nX#_fKFh z+BXsgfKCfXxPr3`HZzfse>cu>9N1Y|@2BFA(em6+?-rXL>0z#TYS2$DiqC3}!QXm^ z1G&eRP6y#%6o0`-?X<5P{>HvBj>6_SCeqf|R7D|`kJ#>Gb^{=FLc@yWelvKR;-A8Q zhML#Jdkd=%5VnbGZf3ET)=M^&DNz1Y(azyO$tny)oMR(7uO|5C<3AbSd<*fNw~Tb? z*H8Y(xJ$__FMiU{iDUVgTiince=OjTB*_5abgz@h^0y34D)D!;m*COs<`vYbN}F8| z=S@+T>L}EMwEX=P=PvIw=*Y$DmkM?{0l>9oj zm!2B%CXIO{N{t?&s9GkWa>17(Lvc1)j?=?>U{~o+gtfgZNw$v1LWU~{f^Jc52)951 z#xRSF^x`V>kFR`X?mQV zRlV+`Z*dG#zRmk$GNfV3#tShVU>0K8^E zcc|Pxb>orGHT@5N!BnBuv@F^{3vQBgo;`u@{{ZXe zw=|tv(A&BN_4};*5Lt^si-rb?V z!ac2;w0^xVanqAucK93q2(j>M!!}ZC{{R#GU9D@k7O=dA+fI%hHr7T}Bz(kAb_ROl zU~3$IA7wNnZhmh~f8c&64K5$dD?7>C)t}E4ej)KhkYF-Ihk&ZQF5Zf`=brfLD(;cs z&k^{$P>NXWueD>#Mu`-%svq`Qw<=?|Omwg5C-#f|jz45CgP*cjjyy;3*TFUxTE?Ga z65DGR#$7`G-Q>85OUHF|(n_th$oVnBz`)|a8N4^IUFcImCZDd&1-x&#!v~xb9OM)@ zPQKN7HyP8%Qk7Z|vQFCiFY`Ri*Cn17Y7>>)S3Hz_8Sz(!^t+9J#1;a|J5((GXD@@m zA$$E!Kb?F2i{R*dF{8zK_LN4MVwF%LD*1#fugxK5EEo(B2hdj?9kLnjWMW*R`T|XUUuklo)pd7u1gu!g1C=30*zI94suj+l&kzmGs}l zxZttZY~cANnoTJsXDNYz@ggQkJBSDG3ixt=6l)h66~q?kRtV!}eY{{WJ$dKfkHWs# z_>(jj-Y}0$Mg{FOTSa0UKYB&uZy|o^&#!v?zlv_fIe->{I`E9(MitIUItcU)frfRTp!Z7h;R1r$FHw$)oCXn9gVxEBE0If z^)AOxCH1UtnWXZR{{Vd%dHU_g>sPJrZLVw^?9#_JT>ZvY>y9`z!@+rSmjFj80B0lD z6|o)G?oJr?CmF0|PeA25XuQCHKT*@@D$H_}gAIUvYbqtBz)9Q}~&wVxJ4qY;%L7&ZFk@uNyzW8lY(bSUm*fv@yyyFntZS~D_v zd)KEcoQnLZ@Mplk9{6Kz$A|o5b#HYG2C|eZpo#Z#To-(LEfs|3;6Th7&DhBz%Rh45#YZC+&%Uo4QiI*8=?L~@yF{>@(qyk+sX zP4T|5quF?J_52@UJ@wtrtf{AK5?jd;UoT0%xO7Izh?%Vl77WPy&^qqiI_@q=p>JzGa9qUzpJ(P zY15#;=ZSRL=8JPk z>2(~)Z=jLwHZ}xDnHkx*YXs*Lj zk)eiZS(Z_<-6hJ-+tD$#*?wi0z_&jdEOal5?|Y)@8inqiXQkM|Wo+s+>sw2yAw;y( z-r`qygl^(E!w`$L5N|u{>yL-p--Es$d^GVF#7zgpI;5IjscmuL>%BT#OO2Y{;zWY> z)>kcPVtauhUn=TXv_p`05rO%>Jy44)&H)r_h`afMRd6-!=^;@3f{{RH!_;2vO z+ep_w8MUpQs!8FkS3uD`MWMC*&x&jhiKUwU)_o?+O@>&V!I~)L+R?SDDoDj!a}PiG zY5O(!Ewqgn!Tuc4p8Li=4}wi+#;GQidpk#}JcZ_kwFH$O8-3BV`{uSGxeQCon(;4$ zciQK`%`$%!SzFp^%cqO!G#x7WUQJwI$nn_QHKnzt`}uD%6GWF0BgXtFYNR2BSrDwg6(`Y;%^md_EBkj^*^>gtYf>2Pt@;LS=|=eTU%#0a9NY~W}RJH zLnQCD-Dkp7*1JMQIoqot54QHXs_AV;tz_okJ)QkfqXUb2HMQ) z8b^p_^XAfYcp7V_7LnOp?UPh%p>Y+MxQq9e&lGntNb2uja$Wd0_HOvK9*N?sPYrzBnTO%Xb`MBatKi%@8VvSCToT zI5|>)Z2%vjb=kZFs(5!opTkma z9@Fi-ts_lunRQJ{L2qXvGB|zmxkW@xgs_uiGcux+kzYN7sY!0YWw&f+_ZQLJfpt#cb= z%XcO~T0Hr_SY66w+pz6!hXTE8Quu-3%?JA;@@-z;S+)5MoGUuXX0cp{bFqx7Sp&oo z*Um_avxwYp9Q-f)MErHLljF6&f%MCPHoxILBG|#A+}p;DBr70uslg;6VwcH7%(pB_ zaI2Dp8ulA;nzGjS9wQ$Z$3vv}bTn^Sl4cPN#E zf#sdhLno9zMk?PeeH-yV$KMJ56Y4e={{R_Y?^d|i?6mt|DcoEw>)R`{KX~@;(?rmzd{)=9MQ?oa!4&d9*HS1FTS}{!VAYeN{?;D@EOcv2D?K*X$%YuwKjM?n8|761 zb0(}D@!f!9+yh;XjarNMbljJw$DN49sJ&1B)8al4{3iHa`&0Pm#JWGnUx=E=hdd87 zMjajzVJbGM_E{R{Nux+@UeeJZ5xvxr+(OEm*qJuQT>QoHWA<3^N9@h}UTM*MOt#ZL zCHPZN(ypY@6Is&aSJW&nS&|!t33og&TwCo6X{lT05?g#xDo8;4XZF{n@vcJQX6eLR()LsyzXYUxANkK-ZReD0jv49 z4OX>EKGkU_eLCs$zubMjO4{3@^k?j!FM)q&-`Xh~5}M0(xsf@Ycqn^Kcb(_yn_6UK=>iZN;MWLdUubgz=l@;aExtxtBXrI$}s@BJsK z*Gi@1eb3N;jFWsD_}lQ>_rzZZwCyic{@MF2?Tysew~;N}hUL8Ev+*9CBU{a73rNjr z1eWYNt9foo9yJxFrQuHlX`dLrIp1npC&Q`O=o4GtYE!0*Zuax9mbzx6YcgM76^G4e zjGkdP3S(&SR~`-GZ`!}$d;ZMuXkWG6$njK==wE5l^mr_dy_TJ#CZz9S1DLabrPnTegT>mOx_%SsU$jS$eje$*4$!Qg?D%&2(mSJTr}|w#`p%`QUFe<{)FINmKW8it6n8fg#A6~GYl~^s=De{i_e0E* zD}Hvb>ZibU{{V(Y_+jHek3Y0jaW8*eA!Decp+sU3^W{JS(YK>IeIJ=4&W| znY>jcs#*zJ?&)6Q8^(KySw7a1C?#TmWG;S9{e-`0ZyRbaKCw25tWDwX9BcQmcwt;3}{yRSiVsPk9`KAEtEG`a5{ab z5XOr-A-59wkVy|aTW}*TenC~lQ^i6!rnSAAR$RR_=?L~&n3P7r>kD1rqTm(s4c`3VJ^d=c-8K0);~IVm5ahBMTwb#TKfmbdQXBp zLGbq8>*72Uzl9rF)5eQZAh=Pv(ZB@59$=YZDHIxuDp`^htwG(T+r03Ucq;;)CbpNHD7fo*kbEo;I0&btppJN5XZ_;cO#zJJK+1U^jK8&2lzK1hx$$>e@ZX5y@dlcF26YP=qLR;1 zvb=$$W}Go8QM0wh>@z{K|;aMof$@B4UM9a;G_66kSg)k;_2#iG z7dDo+N#rM!r`k2OgIPlm3%yy1eUcY5kzQXG@P;b_(~4=ur+e>fTg$flSss30n9{Fz z7PUSO{{Vu6{{X=ed~e`S3+w(n{g(Vss(eMCLeb*XJWr!)zi7Gf1hPmBmpYw{!5%$M z-f+U;Cgr%=lkGCdo=EYIt7j!=60w1-_$o;VJcZjw7$zf@8h1is*?hJc$D%Zjw0_({0pB<_0VBkL`u~ z7JtD$KWyt29wyRlz8d^(@d1wd%IigwPm@^Gq>9k3)Eb4HShdtwO@Tby%WFv%@T3w; z1Ywm4_xLZxoJ}rF24@E*-^7;MbzYr+1LnAwfM{Z)?Pxw%`JclEi10;WZ;Ys0xtI<{ za=+nF-rikH;wV;V8I`SNP>Z$lH>p2>J!>k)RbR2XJU1~#aWIbJ-^^I<0aLbk(on9c zD&f&tLj@`r2FR&`MM5*y;DT^!|@}-!bv0*qkbAcPyiW);VNkJQ4^h zN2sqW*Y7;7G7)eckWT?(fH43#Ambn$k&X>my3`e;Cgdw8Am<0=`WjYTkzQv!aTO`W za!I4xejof<_^+h=3%JsB{{Rq67K`B_VD=hUlW+EMCZ&4NO3St(C?FPd`_8_CzI40R zZ#7L`d;7-!0FRl3&Y)ofvg2?Zk<%E?Dy^*M-$1pp^2{D`M$#|L1At2*AP^e_<#U{M z7_5|UD)a?l2I2yec{n)d+w`wPnch}A@mYmhjlylaAGluv{2`-wg4^~|@lK@;tn%rR zYf*tcv~sx-!7T9Ykl~cZ>A6AN6rI2Z8}8d&s*t$f+P(xe4qc%x=A zW6dXSPVzUNxXG{0FW4VU)Vw|WA^2v~T3FIXmg~qwwie|R?JNqihH%VTSeD2KI0FMG z^tbku)M1laa@<&1%}or`$|G z&>+b_dtg@2UaA4FFSFIk!lF}xjetP}8tm5R%K3L^oQ!+)ugt4blUkoy2X%A5i!e#H zBXI%$0Agn-NBgW;RIpv$$#M2kblo~j5Q(V5Et6e>!te6g?1QH1!m25|-pR#ZO zkS;y5+PtaLX`5M{ToTv-Sp4#SOB6UwZZvm*=~HMoxJ>M9vz>q*pz~U{ z8iPv`$WL6?J$rt$={HeK>wy^~BO{MrQC;+-p)z(pW&N6Gwuj?q#>p+CKqm2ym${Gt zNpBD7ha`?q%t`u^dz$=o{i5{?d4FbK2t#ZHGAD@qT&|#FmDZ+KCm7Cg`QpD-{{Uv+ z5a>tacgE-QBiRRwG}e*6>c%6`Y!S&MV*tk4!Q&j);-ALP8tJ|!_y^#vS3sA~vAgjn zjHfILkXGYZmV0xREC6Vv3lYpx%!{2aNZQqY&Bg|SN_m>NP*$ls=xqBl5la*@xjeU z;9rRn=ns8sHHeg4U5N(QD&yr?_TiLd-~*1l)=!4~Nd}Fo>T=!3A@iOszFrs&s(3jg zk)CVysZyP!k@66fbiT(maTH1MTT$@+w$mTl;fu^{nHPL<3}-ka=E)fSYqZwC;!wez3anBnW!M7)yF79STJs$@R<^wOy{PK>D+#}XMlFCw#EXNBe)|5>8iPmsGg&|nGV6XC=N*?^vnTbhNd2Pqs3-k}ZM2cU&Azt`(jD6u z&AlK7#~C|;uL1p|*;n>&vaxBULi+EAKh?I@Io2&n=Leje@z8*4)W2y{J@4!-X=`J0 z_Dtr|Qn=b5xWaFcyk=8w~`6Kb+u&o%w7NVxr;d~B+$9O-v( zjCCsPs-OYXo^kyv$o~LnABEl+)ch&pZyDcPG`F_84ASODEfN@_Z@C%yLxYY+2tDi7 zziJlL;{BccMJ|?!b#Nnv2cIXF92=OGw-r|BJ$_IRwN(9}Cz3CKem1=^mT5Fwz*Zxk zN)=)W+DKA!kT(!GQgAEaGJI6&#+{z?fBR|l_~>2@`5q>~y%2D5vl= z_mae`!^?1nSCNJ=PV=8%O8r05{5xR_!7Zw+PRgnR@%I!020CLL4%PX=X{+hhKk!X9 zc`+b|z{_hRAZObrk|QcY0)AHEy*bIR)h`WO=~|-4J+|ajZ#NI}xZVNA(g7JAn~~nU z?j*$2iTIQZ-P)UWZV~$XGBpBpnaNWw4$nGmkPqMcVp?+iem5}-o zU93@9%#j(QM-dFiy;X|y)D{GT?hSm5vpefFdY@HmVa`tH>@Y0(%S9PwQ_kWtM<1nrUwlrt`0?@M zOYu*`*8a@Vba-xHz5f7CxVXBO_8Vqfu?USJjUbL!MU8-#Oej94zK5??b zjn~H4&dbpJ^NMP@*NeA9%&vT4;rsnFQ`42h$m=8uq#jv%u;YM-xyL@Wv++i2DEw_> zb*x)Cfh;2lIAEb(9qh@-U`sCnNXJ$g!LFC%XYA+k3*mo-tUOKQEh@qtK1<&qUFrA6 z=3@H??q-Tu5xlj*CHEeB@@va}F1%7-rQW3=vXt8@4ap?g0a)%P1CTICUI{0)ecW*C z+Cu$qeA7}`dB#gyqwB8)>H3$%PYYRip32#+ZEs>ri9JuzLc$3GExZuj7yi1dvv`r7L5OC|pRM~X!=5Yj0B09W_W(C{=j7TzJ$8ulrXn&RCT%tnuvyoZsy8$MR}w#$9j%?6x^^6Z=O0ql@b|#uxcY;G ziu|@vvGs3Fj_%)6dxS9al_ccz-n1gr?jul$N8?;|y9k|{R`1mG#cPLG!K^0BjIX)a zTVJy<1Fj8jMRZ6F)1`5Cj1c6WK)|ioE(0b>1L;{f+-g=j%X@;#9kMbr&uXwQBi=zg zfE(+Y=i;0kdv~W9e858tHvk`CX(%CF?ycriW^s%Ve_GC-SmKSG{n5@vGTP_;5tELb zkL6tcv#BMHopX5L%CM&5ft5}dB!FCn9CsuRE26Y76p=7?IA4nzMyujqg&rsH_K$eh z6W?jFMHRr@Of2lXWC2RHR~?A@*XO>O;6K`D!?7x8-XpfvRiEWcms6gnZYPU({zb2J zZ;6_}jI<9Dcz?qe5ZmA0M?BW+dwT_=CAgM0`&l!U%u%KaqJmr;E_ttA`X48sLC&;!dmaaI z@jK$>zl^+X;GI56uNTDHD`~pkoDA!3(2bHzZf$LyQtBAG7$*`(wU}d?=sq2O(ASne z1=TdG%|}qv;=YydQ$^G4qmsi>HZmeHk)Vc2(p!8nRxKPVqYg^~OZ<8JJ^1s$J{i+L zX&;DQAHTcM{84oC%pg0Vx`ODHGAk+h;#mjplOt$9FdPx*m-m_%#SaJ1uV^;9oLXOp zq=oKnwCk%Dw9~Ei8C*2*!E*6$`&^M2PnLzU`<^m6KF$*nhpeY4JK5R#o)na#`CC1U z;eU-U^?h3R;%2p|=~g}&@Q#xHJ85o}*HD8`*)5IdkrXPebeDmcK^s4na*E$K%23nz z-zUM}+f(7^g!~iW4;JZd@qbU$fbgb+v9){e5L#VZMGNUu>y~!578-u33q;dCn|BaT zGs_-!&9hg|UMhpYQCfJLR`B?_Tg_6+X)U1EwB&0@V)>pa?P6&OV;)Eil|Tk!IV@}J zUxPmp^!x9Cn(yrS@P|m#^!uNQ7g{#A4xe**EZ1XIMO!wK*HO8a(V*E3DhZ0+gsZn_ zmht(A&98~4g_N<5=`Xt0*ZvD4qfRf{?0mi9pV_O%mwyDMJ}}X5JT0n02uq3e%S(+W zSmj(QGhiNvMg5uVym9+PSbog2UyTF9mY-shO-fmA?r!z_ zk!>V`^2#fN_Dgvk(ba%Ok>Ef{JR)@;5B>*!(q9w&ZuZ_k)I3$=9cl!)&@}0_KN7>K zE}d~Kk=))!$5qrNF_|r(ks$@X&*w9l72Q=U+O0ol{{R*E&rZ4Uw6}ui#(oyp7HuIU zyS6ve>4GU=`#rAg1FD4q?c8H5M>34@Zmg$x%i*%oUfp{C0Kq)y(u0yq zW8wAiM~U=56lmH!n!UcO;VT_N(c52xPjP&gPYRp&D8@PEp4LPb&lr(o7~Z*7t9~8$ zm+-0=+GNzbt}PeRNugmTf-)&74eN1%rTEU&D!CI z$s*SeuIpYM@mqLn_JRGQwC@^fHroEJb)@RocD6PazuH=TjkBlNbg?uRHjuTX(Zz3c zjK$$!G!3(=lkH+DDaCt5b+Su;_y*XRIuDwY*8NZEcVF>e?3r=^yiMS1 z9Z1Em*y$522Eyv#D|wc(%?hJiM;P+154PUyon47wl>9-^Ecq=Ckn<>rvINY_#iL3&eBj@mYPA9W6Tdz#;?9SLGa$w_RH{3!LNiLwHJ)MEv0zR#@-#4@B3#_xwVf` zpH#GViYd$=XNbW*qi)wwJdE+PykS-)N7%l02Oy0MN7gA+=H}8-nB8;x@hEh$PW`ReZW7<(gF$d#4RvSytZZ4YW~t z1{38;r>S9 z%dZfO+r_5GclIEr=GrBe<|c+>%%(+_Mt__4XyI$(;08a3s zySBH`cHLf0VH8TRyf)FolEQ{~82PAWR+?zCfxdD)ALGZyop0eG)3twyz5vzjd|iEg zYkT2$y0&|Zn+uyLq#GiKK0{njZ!ecEu4I*?B*r(XboGB1d|CL7@c#f@(vwl}<@bzr ze~FsQ>pHfXZFzNVqB~lw7U^ueWtuf>G<6#!FvW2pKf=V1zonH_ohVdIZQp+!%f&H$24*W*(Ro0QFcs|!&ZBNNt-A4Z2_QO`Ncs$sQl0$7Qkk0BHFkd0JDzqSC zUCB3ud>!zsTJWZqtm+h<_dnZj7sj<(s`lQZwW{{sQdPSsYM0n6Vvitbsl6#`x3zbv zy|>ohDt4^ei4j5gK6(Cu{P20@KIgvAxvuxcM%vw;e=(;@hr4t!)Xmt^qnYkHb7c*4sopg6NsKkg zv?N|f{2~?-{6J+F3F5tXI=gVknP|}x1?mm?x)B8={_-z|aizGb>g+w>B1lh$LK$c6 zLg1wS;D}>OG(9=Ptfp{iC|0E3Ouy^u9;XGE!>n7xZPACpC`--v?}57{t`t+Ta~Z_UaMA8ew}v`~_{y1XNtSh)X?s3OqY>WO88RN@#Q3A= zdvNO|-A!#~lh6b<=BW-GUN`Je7l}3U_}3)G#2*6*D7SQ3B!WClY8*wIT4ijF2H$Qf zbH3s^3q_|}TB;k&+~K7CO`TG|LfVdk;bCqw{*FVkq5Zma7SQFM8~-8ZBw}4$oC`g; z(67`M>|ZUJdBXx z_5GVz%FSh}lt1c32yK-y^6aTFiH-8EC0RFeJ*gYgsbz_SDlMC% z=nCB(J-3o*oidnwv|W7LxtC}os)h`4PPCRhD&DE0I{8-T8a^j<^VpjD)vF&vis% zHP;1h{Tmy*x4i3Vp69LYxW^YTgxXI&g1QfOJ;(G{wT!i7BCoRUzF4H+KnlO6TOi)eBjh z%ZGqS?ic75n8r{Tjm%fUObdIa z;R6{=9rW)soP*KgZEbBr6jW;ib-fS8&2vOlH(XWsy0ywCh+4R z#=Bzu@P<2$V@;)WKV=Z&TV^(le^>l1%T-_?t^^=4GK7-8J2?#^`k32!L!Q2F;S=Z+ zEV-;J=W|~1NApUt95aY49qcB@?Ng=qyD6}Y70y6XqZ!1B0JU**9eAfL!zttsS7*Jp zK-D6G;G>F@AeA$5KhY|?_atuRr5%-s<=s5}oDsqn-p|Iw#yTeeQ`p2{m4hp}qpDX| z#y$7cN|dPAiz^1jry?VHvi$iy2Dhmxu(z0~O`)eD699VhSQx%@F=nrqDSl9or#oXy zcdqpOSxbb3lNm_SDAGTbJ5j?#F5y_%(H>|D)e};b$0!?PjyD7Vy?U5K6>~6%8sn%t z?d}uNs1>*DS-Spjdf$6oR@;w7MPlrfCS_&lm?HBDUm+ko;RT;dO|#p&*U{~INLwP) zAAnCtfAy^3JU+JoCORSh{mP>*SXx^B9r zJ>aTd5%NWOqZ3f;!veO$vV?<4cq>lLY!dST?O=QjXMPF3XbWP{?yXVN&zVzYRTFW& zRdYsrS3tJt@L>|$KO7Me&<6&=pPmjG*CJNE`;7fVSY}{)vQk(Mrr~6icqPKXqpVC_ zSVgwsti?by>W|)a8d69NvyK49-vjb|D}@p*DHjXhcMBg#$$lxmbe9bn&z%VnfQYA> zat#<$bZ|S8jg1UxVi>pf7i;w!WZa3>Mh@W4C98^o?RgE$adkJ-LAY=$;|O^!WPdM+ zsRMAAQwpuWkf^CeFo@qu7i!6Qvo!5X+#yB@>iG2r{g9k?A6~X36#vEC5)XvZc9Fse zrY_>%BNa2kDB$de zuIu=?!DB6XFCtC@_NI-5v|4BJ#-Nvll%HO>`p@laGi_#x>Vc|AEpX~U)-&cwgw!!1 zjs^VYF793)8h%fkkyK*(&#ZHNwX-{Br=vSfXQ~ROEOJX*lE?3*B+&RLgI8~jFrVAe zD0B2|%ldroCxffKyS?6_ePJMwYw2F{uyvVN(Hw_mt34QAlWiUliU&3fhxrsFJzSXC zz3LBP7v9g-8q1l1&xEe&Rz%kbY;>s$&Qsf}PAQ4~-rYo3&UZ)V3*BYa3J7&E-Otd& ztM>*Hybh2q*?yh2w8afnxh5650sWIp z##=lwKIGBxVDx7=X<{ouBZV?VVO#2_hG&J6>+S*+;KKyt-lc0xFN*q;D)Y2fhl4v7 zWK^#6q3-tGf|mcU#+&`%U{WK&r?S6(12?`EuvRIu!elT}h$IFKSZ(TFeBVr^sYD!1 zs`DS-d*oG&yhzPZ+Z)*d84+Bra;m6?y!i~}j;NbJYKZ8pq7;l9x7rdjL0;y6hXlAI zgh*FR)=dTG-s${9e`1kS)zF|EVva9A1~rG0&~Ho*0+tDf*6BMQlyrUH0+C@QW?_nJ zEc5XW=0H*<#qad+ZDhOolEA zwYRw*3$g$C3-#IT1KN06tKf(pS)o)RsOWM?Oqtj`2HVZo;NodH8W$i~o$3dBrmg=@v`G~VXPuFXjQ5$w`5x8UA43FFaD#rdlx-JHpf>2bRM z2*9slEDaECp0f!Gmk+9RGDf$J2|Kq%#X_E&kjEhEW1WwVjzgyRT3Q zTT!^|sRW2~wO}DZ>GyHEIIBhaSyLqzdbp3a%`v}7KkLe($FSh#15S}H-2C0}t`+Fc zDxoM03&L>SN;!LtT0TI~km?mcJHz7lt~7f4 zr(fvkWFn9V20_XpPNVN1Vh#?q^RJZCtyA~Z3yw3q>6^ngIh$Iv!_|4T#m?!T%V_bdyUMx#Tjy52HOv(>mJcdeP7(zIa_A z@YT~nGLO9c4_vEuEy`I5q5drTN%9BLp;)}9m28q_3@o-k@%#|AtD4=*lxkt!|7OSt zyogJ2jHuSe&M;nq>MSnfX!)XuN32^l%2vsNDZ|gp%rK>W^=r)KjZX@lHu;#<$G*o{ z=_u+;vtE&P9$u1{{b`y;`g+~ee<`L}n#Y7$`OK|H4j?_C6D*$B-bpUae4J&ht z)C!DArH4emK)U~V-#DrJj9L*q9nSR~(m2E(G=fb?WO-P7=<8@KOeT{ zTG8uuhfvGEtoiYdp~KFNhp)wPN1F4a7n!2o-PdYY0Y##J#>TR7<|5{QnPe!c+(MRw zpPGE%AXA-9S?lH;=PlN64;|}9Z(X>uESLTtUvS1zY32fKP>bPbWo6QhWn;8CSDg+k zRAzoHXU&q4B=& zzZ=ME@z2bt5p`?Y+|}6B)!2Y3-VER{l^ml|`Zj=Hvf=%kB}v!i){U^G>8{>qb$=8p zdQr=h+?L*@yBo@~C}>qUf(y5J!@wZHb9nE#c%vIN!)N+vy!%n@PrOjGg_hY9E+1Tv zIMo{>I-dDuo&EkJ)p+KFF{*ms&5g7zI4hPy*bAr!$}ucEEx)YO9B6|SWz4FdzqlZg z&0dJR(<4r4$=CQts#n+EB4+*bDLF28w${x4ry()vp5oGtF@Jtr&akb+jE27S*q(=H z?~Py4twray=ZDt=Rpp(!0H2F}(~`Gpi;rr(u@#bn-yZY~Ngs*8>-Cgu_}ar+aJG8o zl*8d_hp`-=a+BO&+%D@3;*w$y2xvBuu(Nw2?yjHBzl9~eX~|9qWTkqv78 z{_)W+io1pJy&91lOEvgf)%K(2c> z8@q$kclve%{sBol#0pJw=8_B-PHSwbp7g(N8yh|q!QUmU37C*&Sv*|fk?l6gVUHKNRIC+vVvTW4ZT?B&ZSDOA?`Z|^4Xwk3C^hZ!fp`?X>J;o&JN9s;R%7sS81 zru(c@eF$!Y?>ZsURT?{#Up#@V^9{#ad-t+mo?hxk;>MTN_neqtsxueFajpE931p@- zc=fa@bq=d%&11r9#0AB{yU)A0X;}*Ag~mfN6@y==nOy+6s<1$kbp~9Lpa>1Y85ca| zH^Z+VUNt7F!emd6SM$}nJ-s&$ahwDL6L?0DXHn1(G*ZK@dS)%c`e^+J*Z#(g8 zufzxpH+IO{?53_?qWx?kK2x#z&NK3$Lh1G|M|gqDC3%GGXGs8!k9l0Zrs{0kr<5`N z2aLm;mhPw=i-!3QrR{x)IG80>{9|Oi^^}Kd0LxYNe$!!5Q{AMV%B;Vlt(Jc`N&(Yo zF@|F8_%v@mpV{4|8fZKVBs-~^wV0vXAG5REe?Y@)HQ++G+}pDe5R}7{C=A}!vWpF+ z09}7)gk@TBzFVbiVI(XORgm57ib7_5Dn2|@A}A%Hc1?Q=611bLWULj~w8O)#dRZz1 z!_?SQEIaZLX^L{9#>4vmS82L{zbVp~C?{8RJwu*d;xqdO*#zt_UmRL~N8K3f{*6<6 zWp3R%J68jRw1uS%f4G-8^FLNAJ-ET~0A-1_%B=y*aOtO!?7*NP7QTUU_Q#CVK>uRhSvPQn|u95a%o4lWjE}8GHxV?FpGUF))8^{ z!q5bQFIVX|6jtEk`Mxvj`^hpXBkt?L53RlyCO;hihauvSrDq27@oJ-gd-Z9rLi-**g;Qt%e}b__+IM z8OhO8fY-q+6{Se45^M{uoM=lQ@YlbqF>1|Vn9m#UBQSdUjkaf-O-w5uDNZY;Pu=dF zkuh^)0s0*v`5aQ#vSf10F+x$ANr$gFA|vm)cu_MlB)vFFB1#jIh3G(!H$KcM$>14g zZ+NWB_78K%*BZNP{j@4JzQ@Ydi$JXRIwAyQB&nFYWD`4ZH7uc$XM6qry?}?7_z*7{ zLadn-n;5EDjhl+KFU8xo+FK{h{P>dX_uQ|JttiHy>N_W%ce01)Wo1ecisF_rt3b=1W3B`?ilf_h7ys{*PTTGeGIy)^{rR`eu zeo0yON&%UsBUl{LCpnn4R41bpIoG*xtTXEVHW>S(IeCRm@B=2>^!6ZWSfc{L@vYKv zF6zVjAZe)b4AAkEuP?^K=hmGP>ZiZR_w;u?Y2XxjKAqKcOVUT3&G2PBHB^#HT0c(q z0V0IH4vL90HOI)VjI<;w$1RIR{C)M zkM`AyM`@)Ca!@rum{QHZKOu(5jE7wIepNGzuTZns2YtV;rO;4qFt;Fa$Gpfd_6$XK z_bqs7?Ti9q1nrz?3!eOfPyWn}cXP%6s6PCGjqhA{%d)ui!%_j4z1H@prS+Y&HL53X zLn*C-p~ITD{pbmlQ2~Pdb3UnW)59GJI`hcU5cHeomUpkV9zR%0+{*`VkjFCR2J)x5 zP>`(T5XSKk2-}f-_3-?t%;r7BcXACeBc8i%)fl0QJ5xMGvScIx`)@s;k<)%Y=YPdg zL9CwK65<)Q={dJo!&h)uda){rW**l+;Y>XvC+px9BdL!|c>U-we>B&3SYif>gTKnV zftiKe#gZOzm<-91Wk@O7rB))N0Nu z0t3WpwC@Xo`LP1v=7i@>l8_pC5M(A6UlJ{<}HOvD`PP`sdxa zk;S7sMt=wPS~sqHX6uO4HG7c+D}FHE>;4!*g5^Lu9&6!}1dYS6K>I?A-c92G(sfKH#wbIh+r@b>s0f^z7xB?Syq%heyb85&Fzeg2>Gb61I~xV)aC> z>3D&wUPbsvN`Pvop=q))@{o8S3DbE_K#{3r+F8nQnXQF(M?EJo%eT6xqM<-!t<+ME zWR2fD^>{^t#lJ~=W|Ung8^#=u`^z7A5q@Wx|5tktNtjTG3Uo<86HG6y3<&zf<%o!9 z3_0b=|E=5Q-r)$eqipeIoRqkZyH&cM0#e+wVyqE!FeR|PU#gmG8IqwednKV}TrTVR znty42S=C5K;y1u_@jOqYkf;QSF%L^(fhP9wt||loQ^G`i+h4=O082X;Dn-Y+3u{22 z#>TeDm!qzAp9HsjP6nQ<>QRa1vKHM5>b)i_GSp5IW0c{%CePov`=Lm9K;I_R*UhF- z5OaKCr=zRQh-)x^{#>a&dZ{QpIS!!t5!M|m4l7H2eDnzl5KfUpARyF*k$Fxj5pHIYg;E<|<5aq+w zDm={WW@I65A_99~k9>T)ujn2l9_P+^p)`47(^t`D9o+4ez_E~3iD6i+K4oFL!8YP=_ zzBaW;L2(6;=1+)$ep^b&I@)*uxm;$E(L8o--^*>XKs!WP|MKS&m%M`W`_CZp=EZyR zZO;y4@z$j2P<$B6T^10WK#VAe31QI*_G zwo4Re4?prUZ=>-ah%FGu9QGN3ei`>o;$aO?mU$bebZeOYK1xP@taHh@FEH}5lk9z4 znr>6%(r4pgJq=At{{20E0z=9i7$6XXf2)fr98i>*>Co5`qCW6vV;^jaDsf_R&gPtH zh;w?A025!8eNx3UO9$tze%v`ynzO5En}!@h*L>2=P3`}3!@05f6U2i+n0j_o1@?8TzOiWr8+b=T(i9> zYCq!?c&bLE#MR$kvD3wk9PV|;eJRE+T+sP{cr-3X@QD2CN>k@OP-oN?(&X=&V9wol z8-Wa7j>Z_b85t7Fon8~u>1L5T5NdC5{S%-O(3cmqQPZ@j$TZW%<`P)(0h%_@mYbP& z;3uhrT2e+kai#h=v^EFD*d_{xG1ATJtB5|vByv?Bxy+WU1`5It;Ck1GXL147`hdCqk}}tJN-pS;9On$66#3 zG5_^5zv}6d?qjSRrWAL26J*iVD6!#CLN{LvuZc;MX`Ai_Q~8$}H{%xtOLa@8W|NP1 zy1s`#ojA{F&(gDlb7ZI(6C-z-uvKjtq&sWDn+Y5iR+jqQnxeN0DeECRlFjn1w@ zM(C4)k-8$!4{xT#8dA-*xEex?N7;ogPLKAG>zkoR3HC*p<}HEz#|OXpB$Yd_?bvz!{9oJ>JZ+(G!_C zJdn?gT&t#yH*-SqFQAdD;!G>w_|~32H=T!-e{uJd=41CtWkG*6bURq3B}mLCEhkb#J%a*O71e-AyLGx> z2%*7%5lIp41{z56sNGD1S&kqL0YS&QmcJS3?BKyRZSN9YdBFs6q(!CL%j=ZnBW+mW z#yD>grl4|FX_FMFJH*hu zWY7N)%yE~dFM*>)o{ufPV-m`#nS%xSR;mPlLu`$<&5A=|TBSIEa9+(1H}bE4%FGV3 z+qnx~t~@!B>UC!x5M+FkHKi!7NQ0Zr;R?7MtJ{`Z*sRXN5hqjn$&a)BOooWJ$^31> zrOBzvoa5-PyKQ_PiauY(?aEQdp<8d)e@*uA6+{DQIuvU0ZF4{*`+n}$Xu|A?jcen~ zDT~ellxYkT=of=+(!{y)SI~2%!MJ`cdiu|0HEVHur76WAr_58Iw|Bg4Q>8d^N;UKQ z`G#Y$ep&yB!IWvpUiOWEc)w82cE`Jwjz17_l%~FF^EIc{vY*=ukZAkYU3CX-B{Y8& zcPQg88Y~Rnr{YrkbQ5I+*LF{Q8k(EKb@}*H204iy_KVlGP$oobf0Mbj9RN}(3gD*a zs*1E2^R2~W?UKL$?g=r@o0v;hvA%5W2G7R3@eV%8Z0(+>P01Q=5xBt+kK=orIF`qt zP(!2*`GZx=^K*0=Mj+t(0ULMhiGbODf^}wQ$fcs(4lO^rv3InJ2#o-+$crFA%OI2P&`7vZinVVIE_`FqM@XM@}g$sFL_xDqpvlRf7?$^+<=alabnLv`{q~}2hx8K zo{?7mCez5+6ZA$Ov1U(BQ`z-Ta`82T8{Wp57|24W>VR);^7lmmd-de^|K_nNNP=nv zIXTo|DASS`Xab;-uLX{yy%>3`B5Cz{MKxrLybvAjDm9A3L@2_|61F ziLxDDCsCv|bH0)KbmVjT5ZVU6)m`uj)-uUH_z!Q0dTAhKrN2~7{F&XHeG64*OWG)# z!)~h4QfCJTj=*1g0i<>U5)1iGJ~15?nRBl?fjNKYPTOpEuK5X4nYB;8OFXz56D^Vw zfutI#q#$g=LZG;1O0O=Z29Dh!n|Yj9is_T9HXCQw$zFp{jhamR7ONKsd6rYi=Y)N7 zIQx<6oU3Z>Yn^cq&eyoCNSN19E4eAQ>n1N)NkLy6wx}szNs|Blg?7T^X8<0BaARdq zQQ?=>)S3<2HVbGl@vXAJ$RJB?2vYne3q*-{Rh4L>FcECkgksd}<_l z+;acDFw(Z`yZ2@Rbw6uDS7Yg*!}RNHw^C0xoW$^O%Ig6m+j|58>Avef;W3Zf@Pat2w#VFHuYU+JAH4Vtp#5n^<&ktWa z0|4aVQc?H)eh>*pt5uQgvl$tx)ma@4ZYOGerzshJ4%d;QJJMTioOWhCy4xX?%t_=R zrULvQ9vNc3k?7(MIMC)-`l5{w;%{HP>>TpuHC+4Shm(&S^+PA@8Tz+C85jt~9g2(8 zm>p!JM9~hT7c#+~o8W4?-4p*LRpeb>a+Hr>rj(YyySzcg0OXa^l&fRHss&-!{ckpo zyP`&N7H1H3L)@yI(Jb~su}d3RsaB?OW_#w6p(H4ptCg<)EVF*4h`ph62Buz zTnTS}R9&XuDnh$AGm>~B6)AD;jw@rfUp8+gkoB%R2ik1*sl&J{`{ZD{Q1}nE+1Q&W zER#$nk{Nl6S_nYAyv!{>%p6k)LB8zMQqUb(6T*dhvQCs*)wDTOrP1I}0CP22KN4qb z*qjg%Z|!$l8mV4SbFa_HWnm>y)ctErm^x&Rn#LtKPZP)ho^cU#MptMpaeazCK5~04 zf!M!pZ1RHD4+$T){F2(cz6TN!mFkeb_G_-)YaThQJDBy?Ro(?MLPs|)_o89ch5y>n z-ND7su8Ca*({D|!O}IcAs2Lw$+;Bsxw9OZBHtm%-?$NKdF%M4wzy&tNyW$X9tOru4 zUpGp2?;}hO!D>55Kk84L8S^9TR5_N9*hg$VYeNs(tUM0=zRZ}CEUNj{*n0avyoTn; z5K^4<7jI5c82hxnqhadUb!}l?102m$pMk4gMRn`@S4PP(&;RngF@Ix9@|`xb1bf{T z7aWR{lpJTImn}PYk)!ZkX8+y1hexLKw+T$OG{(IPhl~*`^aW>Vmr&B|jD36Y$6<2B zTG%E1h3$U(!R)nraDG8Xmr9>#pcbx;y1Orx_Jsawj_sgF6hwxQv>g%Dp(vwLg~8#d zlcS5@h^e)iU^iBqg#q~#xcRFY#nLt&0N7OIp?yS)6h%aBO}DFiOj~pBS0CL^z($^r zSaK9!uWUjgCTzH$`A6e|z@?1od}9!}@&zKu$?+QP8vc0u$&ez&^?EF>wOTNvHKU=K zed{!2aU1R=hq6)wofCNG|QrHtX?suLJ(JaZoi5gd*CACD*3{GbU<}cI}W55+G~nCK%}~xi^DOl&=tZ3fl0?T9%tDkD(*LDaKKa#3ZJWYfiN;9YBOp8_t(Sa*g*Q z$1iToR;}J9J4o-o6mFlx*~?Tp6n=dIE#?wsS|9Gvl1mfJrAijRAy zMVP7{tJ#_X9zw%4TY{POXHQ*An?8&6%0E~#j3RE5ba1_R_ic}M(FynDROEvxB8`3_ zYy0z`&HF&`P1@?)=PsHu%o&PXX8~isQo8iy#a@juva-pDG>AC$y(r&wj}HMN-~akk zr0^>)VxEn1|smGY8{*q`l z_dK8P*hoY2SdY8gV9Uv>d>S&9_V!Pl{!D{1cMex^@fW`rHLRM!xmBMx!k+H!P8^jD zdnRi0EzhNy;On)-|Jl*_96%#UCQ|3!!;?rLW?70H5hBKD4K$hE7lQ?h(3@EgoCZ?b zb5CNHpzM2$Jj^+SgtUi#Y@%})aktgL9zPMa%|lGziu0tc!&v>p2if#a#SgGQmCq`l z_4eg9dfWW;`dn*k>ap!fmz~k6n;@_3Sngs*6T{!-_T$@^W~G)e+rML9A8Q`FsRrUG z1!=odzxg+s5-y2@2<+KDIKeKX+$h3E^Su-_hDrF2fD|y^skzZo%xk_sN6TE=BjhrM zpN#a2KI-&;I?o$ZRcCd(XJ@CsWew=zX)c5Ysm}>or&oW^ zXB5=jv9)E0uZSRJ#&2;OI(aqx);1fRpp4c=fg;2op#(<$uA6t-loV)5~s zG?ezaR!8%UO)( z#TXp$d(>FXAq9%`dVjC_TUls#$aS1-VX=apFpDwXK$qQ&Cv zv&HnFm@RvGan{NkHViIE;1K~%x$2B%Ok6u3zEQ120SzNXlW;HUe%9QNsed7oJ?3!P z_zgzGWjy;B##*8+cv^ky_{YrvUT-5mtnOCfrhDYQ3>ZXq&kU`^1&+kH!8lmzoS7sm zj;$5$))r!qm16*7DsPM4?lae|nY>KaU4id?{twT2GO2(>2q*X3QI> zLH1ld&jBQ0SF-DDE2UdTj#J#h{}|{E`j-Sjj23Z>wt5g2HfydoqHwN#a4R8}w0o$_ zs7Z^6yS@EBB*d|f1x?wwzrKeXCjN&fl_2yQ)i^w0VLK3b6(v#0PzwwAc1S7tz1Y$$ z&C$`ODOG*Vhp2&L$h4jSpZ65>Y6)7~dmi#vR-xjG&Cg8DvYWtC@sctNN zQ-1GTxVusmefFv+>2-C|obg3m@a>b3q2>cWsF}H;hZ5b@mAWwQ5!ZiF^ZGx$d!pgn z7i%(LiG?bB|3r);`HVV_0fF(1)722_zJIXDCe)`e3lhTcoR79-0!1c2Py`46(pTso zB_De)DYjyKa(9JG7TQqQ%ACO()ZoB!88dqS;r*CfB(1=`l3vL3c`}FEVqayE={%mu zt)A@86ul2?1>tS|KBo3T4~ZC-dN`Yl*uw4m=wN-4pTNzE)-?1y~}F_gLK9TB8*ILZ95Z zWD?n0cL1M-x@~e*lyw8nfO+m=W`5!JsL3tbepSO7tf`HngM$&iTD>2$NwSn!v(kiT zE*@dejcRK!vU{-LDNJgER8A*s(r`|Kfjee8DwE}AET!%<$IGy?B9d*A7lWZXOgvc1 zH1q&6Gyw)efWg|-@1h0*^O}A#F2({kBDgdC9*(D2fe>)iZ{^ZiVk3USW${;L7@2F^9f+O4=>CWI zm><(^i&C~kHST>D$DKJ1#87iq75~Y2^6YrLmQx>*886B5Fi5&?|`( z`Pu-Zhj}K;m8n%^P56Zu*2>xC_0po4!fUYcd4GgM#!xCXA?gCoRmv+xKzDYwDsAsq&|N zswq%-k3VbD+pFi@I^B-AATIEhZiG)#T%b@qR-t-7lZ|I))5)Q0)}3AGx%k9;@u`xn z41A~CZ=XJjfVdAr39&Y7FM?m6YMs5w2vcaDYI`PX^|SY#zo0NV>$$l>u9eKjc8&%> z3?|Xg42osa9D9V!a(MWjorjW2${nueChVntlgNiyiz-n@5abVW#%jR*Af?Pd)!rfl z?=33-ty1@xl}2FAX(!VeMT#nDEu`q|;h(*)w->)++WVOn>H@FmKbd`K1KSp^w&ky+glp8bTRtSx))lw^2POa#o1WoXy(Ls)d#P=o~Sb z#6t}cH%r6B?@w%qwqaQla%|9AQ-pqG>F}2D_}0E|-Hq<;tFdb2qV6Qa;sy83bU`XRoC_d--5f z2q7&O9%vUx!MCI+muG&@==dgbu1KX3!Y$>fjRC+t%KB1ePE&)7e>?n&`%cqd5c;v? z?uRzT{j9bOM#~>UaCz2P{j7P5akUbD+Iu*Q&F;Q6HXkNDD&5*4Ze3REy)s$v&UXU<%jK6CB z~3sRkb>JG{vAlj`t?!Vj)<@4%O$T6BUc718l_B_e-W$ft~V1k(_CitA{)DabM8 z#4&PVv7heKM)Vlt!p*YZe3HON<3}m+TWL73HbjcL!>Ps4i!4~}vQ{qISz|&!bP_mq zFuFs_X1GKJXXHqf&hn2xzG6JJApX!d%lq<)-lt322*5~BWY6%Sex(TeHWistZAztz zg07M1KOdFU*ExaV?cvJpcBwi65?}068NT3YFwokUGJEOnnY_*xPpjv1q3cr}`z*{O z|Irg9eLvGy-5sS!jY%62J<_xPaUd#I2glQLCi)}IdiXD{6taPzyEq)$tWYJjUY|7Z zJ_jfiB|-Q>jCUnjjJ!pOJy}$SM4CNJ>FTVk)aa?XWVhI#At{IH^Tp%O?9)J73rXT3 zf+w{w&RDGi{tAm_kzTxX2IaK<-?hI<)zArZUA$OZWW&#YQ%>2INfip09d&;MMx{T? z+y93*Upk@}gU2f~ERTzw&B5eIya}9nRL=DY?NU`&J`9bu z+IW$Xn_=G-^#6xPA|RpgjGv%s>PMS}u%z4?N9rk4y=+Vz?FKfS5s$_8hP=Ni1H|1n zfzyaM(v|bSh5WtIMc1PC`rWDjnejwniK@(87AX$gJw0j5*U`|ouEoXd)(q_r8aCT< zJ-{0U@dN_8W^?@3>CeWiQ#(>+-s?{+$8bnn6gkFT;B2tYmZZ}{?CnT}g;GxluX?tC z4WGBTUCPL)+iS|nl8pe)3pV4eeQ~f{p39YSGd-?3I_Po)^>fD3*n+Pi$sw-CR%!}? z#;a5JJW}QWAVJ#|P)D5a159OF@X1nh|JlcI8GZYY6#ia{8zcK zQuZ|&5gu`SOx>qR?q~7-wQr}~-R&`T&nIH7!j+qzYNcCQ8r`#g<=_j$-}^KObwB3i z&-{hS()@X5tn`VYyq8G4KCNAPUubl_26~v!CY4eA?p*O5a#Y&15oM)S@8Kwf=5sj;Dmal4GsaNzOdNaSQzeU0YHJ<|>`hq^sVoEfs_^wPYd66`sc zdBzp#_bGjn@qq7A49!2)0HlD@@jhzb{W)K=V5os_)v7`5RqjfWFtw52QZ}{&m1ENZ zX?OH_9d-Y{;oI0P^(_gRoMiFW*Yfl&2WM7_msNRtmRY$EV#y18@HE~HXD2rKH@tRx<8YF!CK#voKI-G!QN$W!`v{c87{2f z7bSv9-fgNY6;Cq4*whW!S^U|J4NzXvZuZFe+^K>!+!Fr-ohqO$y?ElmB%5WRUsWh$ zNxIZABXUEVDkq{2MYGSx&8TP7sC&z3e#QQUYy6dEx!ei|**RA^^lkh7Mh}6vk{(3kLImZVPEdbn z*cgx_?vEg==mgVP9L}qSC1~+4T11;szd-TyBp0b>b{)UZC7x-2%`x9Q2g}Z`p*3VK z*B1?CYpFgA0Q$5XyxIer?XnTgcL$}jP>C$m{(*`g(>5~V%PJFV8q#~`#g2cJtHa0h zmq2j&+g9fYuhHdP2%d}>r`S51Nd!;~ncAw_o|Ak^$-dfxju6s)K9D&N-pH}%+`x9M9R7nARRn%FKBeZM=L*aY*batXsc(3CBhBU_prL!{xx znIV~|+xQ`p|KYReqV(JHw@K5wv9DWbEJ7KLHQ;{3QF0UxEF|z!0E;If7ezGE$G~i6WONID{0H z---}W?)P>QJ}5dcqOyGP$*MkUEJY`U_i5$x@X?5#e~*3t!+Y96iIVHjC)bTvWS9{= z6@a*nEuKHI36inzE&$gUsyJCc3AAs=e!i z7?{j+buE;1OQ!Dgaa2iWL%NrYU8fltOH!?D2uaDml=aj~ZQ$NZpLr%xTuqq}N8jS$ z)D@>x{wrROBhb;=az#$?kLg(-L4O9ut&}%k`7FgNM~#=QhaZ1G0M%zhZ9+*8@5wOB z!QjoN>4xRe?njt9y}e+9`+u6hmKxs4i@Nwp#V^sF@oA9ssgX0wt2em$0x0ii?@5u9 zN4VNganVhLp&w`){~{SRaBmpTq#Nx9VKe`?W@1|Osip*Ha`@2^DrBYV%!tj$X%ZHP zOc1m3nzkKvvsew5VHF(aqS|%)eG63G^#QYL8ZI|$Xa^wpzUjd&M=wr}<>MY{$=SM~ zNOE{J^&$G3QERGnO;Ky==A&lG1V`6Yyo2Airi{2`3wB?jrVdw#EM+Z=if)W6wwacn zy6n_;HcSS1>dN?|M?ZJi zaha#ucRqMyfi5mzGT)bF$*5DUbua#hd?~v+zVlr5Ovxk8$tG(`+#oCe^&zNoO$-;L zO+O^No3eGvGA5w_81Oe12wAr7^8ebQ7Y5m)_-)6WPWgoRg1t7+KC-)Ih>k}jMrTvwLbZBrD@uSQIwfH z)ZenRaatVJAg4riZ#GAD$ie`dr%|#NgKObi+Qd32O)#E@cg1RN;8|40#cRtKdfY~- zT+VvPQm@MGg4%8uP>k&3y`%Mw0x}GhfUQQH&I}wJEpGsErjKPKb+Ds1k}@#ie7CaP zbLJLi`dkBDAgJmv5%O{x=8OPaXsNX$C!?!;a)cg=`?dLfaxYhNgGY?*s&#iYH^tC^ z_-82=?dXJssVJnf_>vxf+6~_}YP>%$Wqi`kgS(R>>uO+JNL@{g0Rns{r#mOGSp-e?2pzi_3@4^(MhYgMlZbhyM@nnHkF)#jzjLmFwnd&IsPiff8a(k7abE&s=ev{ z6Ph8t9X%|M{C(5kGus#LsSnwM*=aE~uvC{Q;9?EnZ8GDkjLa zjYbx1KEj!LfA}_4PP%_(vPpHOC-VuLI&6wlZ^!y)>i2+!#J$jkeN7^ze<5!$nOlTO z?)*RI6xm&_GAS(Mar93lz6vO+@`T(&al{9e1Q;#Zn9Dp3j0>JNfR>8Ep7pqur zKm|jmeaCg*2Ad6`s<*Os`bQ3* z)QURB&y=~|>x@`CMT!0YQFPW}O}%{>M@0eYZctL`2B`^3gD6PDCIZ5kbdB16(%m5- zOiDnyyGuYqx?@P!Mh+PB-TOD!buKuZ^PK1T-uL~v)8yBp@@5Ltgt?Rq)lu7#A8LWG zq*r;SMHFMD3rr`~{Jqn3N=VdVBoat7@q+XsExxjbl1>Y$zu38fh)Et;1Vwq?E_Kpi z47CCc_3O~UgtE!KO9FG^zbP6n+!|s@S{8oW^bsRZDwXlGF3}ioAn)83nn%nOd0u{dxUf{>;XQx z+!SIbwY+#{#Z?Jb$4UE!IEb2J59uRNT2es3CW&{qR60W{;`0efepu#`fCC%KU01fgUD zLb_*(e*vf>xjZDbs*8U9h)*urS*09Z_sk$EmH0~UV$V6pDIjq!-MYOaUWbzrjAy_~ zmv*w_<@Jr+Df}k_D>l_Jn!hHK?!KyP!qNs|0a>*@8TSx45?|!eqpuzBc>aVWG`J@m5g`*lRf;u9t1ZH*CNUZeSCl;E6WO6j6(;~M7{%cj& zmb5K-615wW+Wu?p$I<7;g=0u$-z`JF=o#SOET-u4E1n1h+7M*~)8d92bs+02HWCMZ zZ=%W}d8Srr{${pSLHo>HGL7$E^JWa!4cC0VS`PtwAQ7?3(%n8qxG^{(_FWNh|0b}o zerwu~z0Ulcv611=!&=-@Tl%lG*;C?$kpJ{(zT48|2mJ@#PA|byr}r-MHpW~8vS5L~ zZbD`=8S0&gYo`nEmV*|%3v(tFe6kG!mb+tO@!i+5UukjF5IBcUakMB~Wh7Qj_#EO@ zMCxlZ^C;VeP}w%B=DNey;dPMUe@A5V;<-IxoKbdd)9w8}&XpD`IR3?LQ!+K8Ta4UI zDKmgf5#)lF)Skl8L(jU>8*qYC^~6ux;0@)oMGu!7IsHiZpcCB@!-LMJ5jMO36`NlQ ziqfuN3^$D%Z&jhJg94d5VfKwvQgyQ*m)!i)_i6&)eKa**UHJVqA|i>WvB^Y`VI_$J zvU$}jN{)YLK{3X98=W)`e%0w(Ig7ZZcTJ6<+hkc?ZQ^B(+Oj;hR8=|J+Klw4;X!Ry zWgydk1U-V_VoDMy2-w6TlaCbq!@X$XyJ|W4%C(n-%r-@$`ANUCZC0v=VS}J(z)6Pl zKY~)2`g^AWJ7E76}UXOUCM; z)$03kv2g!K$=~-ntVZh*+5WtD*trfcyyn&Rur2p*au04zBmI>5J8KL-sywaSIDvq!?|mUmgga*HhY{r=uK zl~m2dFMdusV7Q5H$V*t}Z8kV?t||-DVx)drjuZ`9RX~MDi&RYMPRxkpL+eHkNQ3=_ zZd?;0W+6kF&x_vXgYBKhG=4{aGciQX;U{oR9puoCIV4}Q<5eSrZkUeiDQCR~bbbGy;5vA0&@wxz^TDR zF>99@m~ekE<-narU(U1WH+9MSjfwnCpJYa>>RWXbr)XXH`fXUU%U*sg_!577uyJyz zc54rfHg#cnWo@(W(qK07;mFtH^3g5(v>@Rh;Q|0S(b%->>;a*-Vseg#PtpHn8Mu8Q zm)sA`_~!f6+y`Rdv;_5^kCfjjttZB_*^dy;=5TJ&OZ*n-H4aK`07ELSTyCmggfbeB zNi-(nJknXxV>nfmDW^WpmPr{Fv%k7k)<}ccKy6O()woM_Z$2uUC9^KPnH3_q4)o0d8 zF!O5Wn?`;jQi!dkZ#W3XRo?H&o={fC7;bc=wNQ59ZRJqsN%01$s=9`$&9_8ss$Z%x zb9*V((mpXlz9-K@>XUIDXBZX?ekVux6@TesEVKsi66wXClybq`p%(3qJ;F3gY}m2I z)c%57dF!;5D&<{g$(vTx8jigD?q_A54xahnA%kM`Ul6Z1_=?|u7^!5xFqQ_!t3~qBzdqhQ z4%SJU$06=KdzX=18jHXFG*PyZ7ZI{78o9-Kc$oPTy1oZ@hQf8mZ>F(G4EsNVM{}?) z7_sJsel!e#2E@fI*-|YiHN@M0+j-7g_UQn8^$k5(4oio=!jY)p|H@z#3_+2_*qzO4 za!NG9SIaw4B-+Z@oOr75G205dZ{?V{Tm7n7-KlU$nB>Xz=Qn2VBzWd}aMXoAv#E?l zA~34N%g6ubtG{2OcTDE_)Tv@6jy(f^UHCL;qy38~#?PP!Zt2Wuh<(^_Ykhp0#54nh#WV`&C!QR;gQAD}y$@}CB`|ie0 z)T>GiPiMH)(hMeaeU0A1hxL_N0SEuAk5#F?tLAn~VAh)lY1giS(DKfO;7i53IJiyV z$AkZWcdkNVL*@D%auT3GHI8c~;{Hd#W?ZQ^DZTc1SlkV{BEto5OSdQ5e1A z&o|B+U%`I_xm>dnni;*#N8UiL)$Y`eyw+#Zog$RtSq~2Sec#}@SL&8a5P8r1${`;{ zK-S|diz#VLIhmjTpnT##0-+!H{na^xG(uqhZ;FTky0K4Ib5)M@%iwGqTx;&+!b(LQ zx-#~BZj}-(Qt7poWroff*54Ow{DZw0n5WKS%0u2gWX!HyKlOP3@hov~;W#9{-v`|g zUjce!d&cD?e&uDz@Tu{|F3i~TV$bRxI>qUX8}|Zd^w&H_))1u)tUVj{st<``zwaq$x9=Tnj!+djb+7Ejc-+kOiz1y=Uhp9Lrpb}FZ5z=r&*fjH z(oW-zc;i|1`PWmxxML?}AJpS^5D1XRDBx&G$8yIvv{x0U-@6Y!n5Nu`qU-(u@;HA) z87Us#og8n22OF0cEAxHsfeOT{Px~KxmMZwg8Z&vgt z3o$8Irvu?2{HlYgg3+czJ}wa1`dIftFWwU;4t{OMOQU3`ds# z2JMgdvrT!?m^eZzU+ADM*8F>=k z!8Gl3{3k!dZn7NaWgR3dBnEJVYqg%w18J%mmxX#n-Cf^X6Z*dKS8~U%yPVZaUV3}^PjVcFnPYf_#;IR z@oYFk6o&Iw)p3nnkso!UJ8eCcGd|KjqnjB!(ez%Yyj*C^N61}F93_)Dr~<0sq{6zU z16iXkHF!OvLz`@cG-b|GgLagv9`JptM#^{u42v+f;%e5q_PhD#Fq-4z99B_k23&_oGjd=#>J~@8r;bPteXd z+jM{FNEct&asd3ZI>tEIiPAoq zUqZeKTVxE!2|H#w>bSYMWsL}e-`s*HH|`ClcY6|SHtKRfQZe}d6rpo9rU=z&hVj^- zjN#z$oY@o)Z>!Cg*;Y&V%X5_Ay;)pobxNsglp2}hM2Kv=NVc9%N%5X>!P&nWG-({b zf#Hd#%7ysryuSNh1|!--kOnYUKut{2^{s?;(Q8AWYwm)!NwS05_#e)4(b4#F^D|2y z$d6-bqc8CdFJLDuUt_S_i<-9jln+Vnk^;2XCM`=kfed6cZLs}*c&g_tOcp(x5r24R zyf5UZ`;TBmdMI+Eg&rm<+X+vW$<0JQt<~Sc>0;M0;V4;TXZLNB5bpV|Nha;ea|hTQ zM_&SEd(BI2arUl{8m^K+52yPw>y}haFMa`LJHN*-^de89F7!G0L)#z$H*wiMPJm+( z?@^`8OydxMct zSSO0&q7OH=jG7z5#bd{}=z1jX{Te#wmuCB_#uM{l6N?5-tF0Zj-aduKHZDV1)nePN zL%QInw8C!Y=~&v^ONChz((fY%_8- z|2X2rCYe)Th5_fb-boKwj|35R2w?c%p_ERdq<@uNFa>GGlYeUyd_-vdc89CJ*ENnW z+qBtysG`GmJVLBS-{cD3ix^rvy4#3eqk)Zwcekjh_!L7HUo#-hWCZ1hEy*srmXiO| zWSiS!$hv(J;FJaYpNy11Um1;4oW<{2-)70IQLf9gQ|>eS?CIyjN-P9TLvc zX*<5Cd|=M;2+>n(wZilfYBLoh%ei}5zBSHG=W6je>Dh~W9o7RcI9Jyq@h^PVS`EGj zNUH$a*$ZNo4iq@h(23ZHLiQ{=*jf~-!KGz>xYB(nF;LR;Fvh>7a57MUO88&{g56Ug zDNVxE*I#P8F+HD5jS+j-*Z&$4vup6T%Z6mv9+FYV$CtFwqW6tezW>y#i*rA3JakLBdKOHWmjn?C^Q>UtAQ?^eN;r-1r|sutldMobm`x0gdlv znTU03((|&L@=*XXYru{S95_+0{}C*uZvaA^>WBlxAwP#4Ypb0U-qojjXdZ4lByRB7 zUMwo7X6quebq4U5gV3qs{UQt^A~P<7BlMO)qX0ZfEdc#!kt-k-;qx`;trpq%9^z{l z8+CBhP5xyX%5DTU9=}dbgM>aTMqPeOQYfnt(}~usv28K)NwQG*#=t}2x1+yVdDx7G ztcQ19;x!8$)s~6+Qg(7-p8EEeF&b&y!wLO7>VI9-?XJe@_dy(<pm3LazYP!jlHW zf057lod$<(A3S9(7SLiE5c9xll&v-%eOYH$9^gAIuM>7R(;_2?^A(<6KYUvC=ZV}0 zMs7F4?=F^@>W$lLMW^-Gwq43xzN}kHwEFEIJTADJM{QnZ031#z9kpoM8afzfs5&Mn zc-w0Q$AJ;^H<{O;c}ly3rCq)_VU6X!C#oQ0bfV$boo&G`Q-uM0sM9es7%z14Zz+LDUDdKH@zexP zv2g=kUMB!E?+bd1TassL(P}3oz%yfo;5Irkt9~!{|XpA~+v!GuuW)>g@9~lT#M+lpobN zcxQZO)fh8~jE>-`%?^H~Rqpg#j}lEb#B%cL8=J(%!9~`J22bl7A|+Xvl^akC2`8|` z^y)WF<|ZFbg=slC%Lza16JlY0sT@bZ-IAm*=eEA03oY%X+D}K5`p(lsL zoqV{yer3rKVgeJt)l@I=bHG`!owEej$`iJTFX)hHjGzRY|8ZmLC?E)?IYuri?$HG838Zw(~mR|l_IBGu^Etjtg&RP{Y=l1 zygj8-`asd}enf|7Tr6Lg&EgKBH&KtdRTrG$AV{#>@V%;z3Jv$m9#v*)>UZ4Z+#kuK z%4Bl%{6fROc*fHQaN;>`;8;DE6Kj(ZfkQ?~NdjVKid^V`37bbEupd>sjtFzcN;2dl zmDD#zb0zd_Y=72BJTf}n%=ut*#`^RdfwnpcYZfW!hhar#jbAxs#?MSUQDOus#-wW^ z8+asg0^FgQFYp*h?0rns*eYvEY-jfJ5zot)1jqgUTGJQou>BE>KS$-~Ntj{>W5-Oyrwa`4uD6TmKyFsa;^bQ(as+z1ztRO@*oDCyazIX z>aQp5xeEU#{rjH>Ysgz28}j63+G);8`q79|rv0%pUVga+PV({fGh~*6sr0CUbzp=4 zRtt0|O+C}>(TCH%Fs3mAQZ6CuU%xXHAp4Y|2+9B#(a1L0^|eO}oK*T{d*JegQQpxu zMxT!|o=suHl9JbxeG^FJsaT-`%us6{3lo2?tJdECRj5CjL_9`9M8Andd7|ICrCuC*%=WljS?=Rl zvDTP-9;m@V7QW{1J6C2}X}_R(L4U-X1CSAnu4|7TZVm0L{%uOj zeP6S46Y)h0$;r8l9}5GldCD(8c=?CrC|vW^etO^8reXcr`LgmvG3D;KyI$o%j-cfP zmKpV+PkuZWN42mF&IAM>@;ZM9`gZ<^tWGvB2q+Rri)Ls=72#+elw(|Q>d}$BT8A!WCnDeGyTL9L3&MZv zRL!?v5Jn)Lt0bP{GwC39O#Gm6rb!P;U9`)Qu;jbeSx>y64wRZ}6%r|%Wt^-yUl51z zR%WzZvi^tu)p$>RGWRR47i)k%iLX?Aq6n&l26Q7`s1BzCwCsEwt&?@)-<`@_OZv*3 za9J#E7;EJ#^mn>sEltK+0L zw}PPf((|=jQuF)emzT1Z1b>)ro!y%j`h!dinBQKh&(;*Gw*cY$OUD-^H{sE3+N)Gf zqJt)6r$ZaJ{(>()d-}d3N-R=)IMmK4SLcvu7xbJ!^eN42rqFBF_2bZWLey*te^>(8 z#P9Q5OP;yuic(fxT9uf^cHEHMqB zwZ{%xm0%)^X=D?kk5QGx+K??t0cfR2T7&#>!Z~Kst5p~Jcg)dfCuL@3wZvQ6)O4!o z7;gTJcjvGW9~e@8fcI2fZ7L>?DBkohi;*}R85Nk)YwGyR5>rRkNSM;f#t z#{UsS0m!gmRB}HHfA1Q(!`!3sm-RE8uNvkUUd*DU!SNYxN>M|#kY(=JA95V3xtAgH zkQ4ZVJG)Jnxiq4eC{Eaapcc4)X(H=qcP_KIQrQ1Z%ZF?cahdTxT`!8(H7+3(Y?cSj z%kJ`+#(um@X>&rSBkR!Yp8{c%gZd`zP1B!Pq_KP?8FsaG^{4o3QTlIQq`R(N{%0}z zk@S!DA$^@?@2}*A;gp8MeN!>J9UA51Bz@Pg_FmAYa&aLvkAzWC#=UOA}^}85?p$nu3w|0~*nI1r*ZDU~}ZBSJ5im ze)l&k>0yqsn1(!YRootT33S75?RF+%E*kq5Be)(&bom+G8F`fdr@sYbywk_97#Ci& z?`@NBL~$e2W!qtr>t8wj7AJ|1~e^zL0b ze*+Ho5i5(<1<|F}twlEYB!lHsTcu<6c&PQHU~m2gqEtIRMB7HiS-6oi1(um2-cI8b z&7sLXEhiDeFrlId=_V1Uhd&U$8*_CH4O6qLnv0~_=6{}q(9`H_OAHCt<_(^yXO$}d zBQQWQq;Il|OafT<mymw89wa$|d069L~^o zN~FN*kLK|EEguF;+5weK&aqa|J$zF{BdW+O{w>P5HLze+jcnGbW%NN1(1t?j-K(@KD_zk99glSXem{-&QYr0pIMM#sHS7fHl zBfXf&AX|q)e!*UxNaM^C9jkZm?A)0@y*iaSR+kUmWa{p|%#Xyy`&L+Ko3_=|ROFVv zo7BOr%>EP2XYu#W%K5E(-q2j@iRjFBIq1Jn1@OKx~K@ZMS21fssMzZHW# zKIp?Sp$b>Zu5y?fI+`r9zW&Zg+jOcEm2z=q!t0*uJ)>K`E@_noQt{y&jOiQob3JQp z2Y%kc(v=a%QduV?Rzl+Ch+a>wHL7lGzkNPU2S~N? z@Fjz(+_FdMNdaw5s9T446ecYhEAe(Sdr;6KD6*d2W>X_H(TkIG#W8)6BGtz&W23FE zO0=l|(vwDy*^GtvD?A;H&M!Udxl=9?KH>3GwRzAjApTd!5M>EUy5sYKRu? zgHs)J+(U)d&xLzMD+pV3qgGcG%|iemTSJEE1`aL4e04_H1k69>aK+t%!j6@RFs%KI zpgKIDS+eMWWF5L)n=|3y2bE3NA&c$UXL7lGxm)7eqww{FI)nE#MbV$a4CxYe5{Pn~ zx(t}9wCe=&#u{uNvl?(gzS<=jft~3yx&BBM=#vVRqEMt@!)=WMVneOQbh~pk4|giM>{O^ga2lman;Jg4SyXNX8^!KIC23`SHi} zS?01eQ+m>~x5sH=GlG-1x?Qsu*^g%x;vZ5{^ZmS0&J|+#S@XJKgJP>C(Z*v(O(==S zKkxRZQH?Cww?~cc646cml&DI!=>W;iK<88ccGVrNnZoA8?Zo5H!asbHj71Iun!k)l z(!5bX30wui9%F+0b3`YxPZ>fBm&jE&xO~eWu4JikcnE#hNF9FiiJ*$`KS|IZU%yZ? zMR0w2{Zu8W?|ScTN)fbn9}x;7ap`!vT7O(N$Iv3&nwn|qeO=Rp&+5JN&3FrMJ0u3F zLM#8Q5kb2G`2WDGl*t`rI9NivXvU(PMfo>be>4XuRG%crzkK-qkD_QU)I_BYH4sc4 zHZZ^&M0~~Ylf%TjS{`J)pvm# zH=G*(eEL%(cRvp+-D-Ld*0}fUIQrZ?@Q>5Gx9%p}xZn%v@L3B^<**-LzL(O)(uOdN zmmI_8FGz)nRqv4o0R#>MWH;2}#Im5Aaga&jHVKbVV7nynFaM9C=!onsZ(m%NodA`Q z9Q~5n%PzappAD$Fv`6S0qa}<~`vdLf*^5Yq)yXlwOfYL+9$E6mA(2PFvZsfVw;}>5 z+UKOO5gwLh+NvV`cOhki=PJIlbAT=lojB%v+Vr&C?j$< z*Y38&^2@Go#px9CVDwbx9{-oP`ZJw*IR$vOj9k%|Fa;CQ*?Q3hkjDup5mq$`j}}cR z;%1V&IhVee3x|n1rg=va$|kwZH{%NCwunM6qO*+I5~D7U0R5ALG$$@^^sLQm6G!Dq z6gUeK0xWX&8Kd87R?w(DiR6xLd@$!Tl6*a5K`+@vK71MA zcYyzD%m&%!C0;=GI4LS#SR8!)g*zu+8|wem8PfLf#ocOk72;$`_S^;}iii-8Z+|LH zjR|Qt(C+Y7R-3r7#VVrO%5SsXr-LWz+$(XzdiKi!A~iO+*q`$qfwrDFFbK~spjj?! z!a6cwV!2#-J_NSBV_w*|7jS_fQ~=!vk)6{ZQ}J^~!Mm3~T~IETocjg!E^Z!_7uG-m zi>K7wfeaq81k{pMW&s#JEITUgG1qDCb+%+V zF`C^sNrLTJy*^Po@@$mvKqg?I6sjVj#^`1#LqtilRO_v0AlwdjbC&>S&y-G6*K~6$o%%*3gimV5D9gcP{pO{ zoo7tgme61RUF__E z-&AgHMNW5{BaE8dw&EVVvBGjcv=dB{yZ)0-bXMt~h14zHvR!8KV*cq3e_S_$Z$nyPqKW&ME)FmX-iP_%rW#hcQFBN1m6#kfbKZ;#!l3@^Y4R&^!XxFM z?lJN~FVWfvFtZI^cw z{`s!;h#(IvUQ$23f`a#nQe#Z|GGQei#mjeTtY7u|V34CEs)KEAvW?UPDwXx9JL5)l zdM`X24{c!Vsp{U|ZCNH>?4WTQZ6xp2S&!+hBE*Xz97=to&vgi{h9C%bWbncN+UHc!M@>6hQCU zcWT<@}*F@bUJn4wx*F$%`~~eoH&;UV(LXL0mSu?`1h&b&9?VtGB{`UUjFv zN(FQ8HPDb27vAT(9Y$(e#P&+>{=p2$FiS+WQpN;{7z33gC#!wc9+f3d)Uv(ztKBZZ zPy8idW5$^|xusCZU}kLlKp~foDB@vTbj!04t)7Un`^6__i{V)6yxEta-Zd)W0;2Dk zzhdsR$Ka`fc8qbB)byAl$Oh)d_UU$F4a4~%_mfU?kceymSZ09ZssHYmS|bW zeN{BkU>qR-%$atn5`S45iz?(mWVAtpI!%jC0c@i1^$d7DLmCzuU?~dcI$o$Dnwg$U z$4?2n@dtNsCj~TV=Y1o}JJ0FjA*FjFm!=|Spg-mFryoHVs2!9k) z8NQ3!{B}bJ#PylvK;awqefptSq(ytHH0|I9%!eN4`=yEc#>%h=8cgm+g4MTGb17E#b^ZTlVhK^)+9cqR{Jj(`h=t4 zd}pp6Zl!@l1lK>^0-5H=Z3I>A>>0>!ed0RL(@ug2CP zCOeI>@&Qksh!QuNj*b%W59G6I*nf}rn<_}`>IG!>$5s(2XBs5&)Rk-9B+#!GEsLBx zjIYGVzVSU4cY`o>Wy-03^BE(xNuKSLfjtR$5h&Io5c!gj080Li`qzM%-x+Au?k+T^ z(4IH;MkrzUaZS~FFbDy_|CS#_DKL{a{wI_NAnR2vQ@ci*pUF$^Er`i@YxCrKB@-mw znSeBMCXn+7BCHr^7vQw6$Jw13uxKCi`M59rG_fd}bOe;<`7(c;a@+Y!W9W-##5}n| z6pD@7cR!UC0tCN7K^qrw(QJ#cQ))lifqg6O!>V6sX7@t=6eJ4RB|m;#&AX6->@!uhx7c`^(gtFo;3*KZAF6_*g zwjW6T{Wh@v_pJMpg(9^jl6+&H$k}3quwtdWHUB|kfT8{b%04#9+n+-ewb^Z67#`$; z2ktMC*j~xD+$=8F$)+#reo+ZyzOT2{V-}gDnYY5M@O1++*YN}bAX0o_y)hoeK{k2p zgD&2H#izc|zx2%sS%Q>t@}KcWxu+x+=fO7w$wUlDfET z-Z_t9mDAznS=mp^%DJO7x4rA4Y__9A(|={o{`D1UYKiRSadcJGK(nsLM}8I-QhwI8 zSn?-7Ngtze;;sCfhT3Sp?9&VDr?x?=OEq!tzg1u1N|?3y8ID>^X1rk_>3L*UQn0hd zmEy={D84Z1@D@EWvQZM3u6PsJP?u16-V)Pr5=i<8)G$itPJ;YhtxdebDl+jq#6%BO z`yQ1Zf)m*7WPyoCcT)Ewm1xGzEgIX}8q&s)TSG=I78;SF9K0c)m~BnzL8G2219aHi zO?W&|2$QjyrB90q>KQveQMHy1;hM3F6MkVw(@*|KC}2%A^o;oZlk8zxmMc!qeHKKI z|Jq}$$s|^^aFvo3&iOp0LH0oU=x|PE zKoeVNtoipkj&v*J0}aePQ&3oRGbd|Z*riVrf?r?S@I@uBqu)ztIY|ODlB(lOqqxX>=(R2q*uk-jW9^g zoVh6MMS?zA*{D!riGu~*2<2k_rv88+fC_6iE&3xJ86M(u!15Ax857H99w4zaR66eG zt;b~{mdM^`(8r+}q#o?|<#a~QgIW(>Q;?c-ySjSdZQE5ge+>dI#~4NRaJcftXF=k zG4vcx5#Wj5iP);eq}xGTtOg?U^^r>Qi>C^qjm^mxwn5;$+NAdp5t}l{!?M3_i8!=w zr2|Zxat4vi(_R*L5BgSlYNjhiq<^tz>D>3_^%a?YNH(!%p(lPC0X0|pN6_Nh;ZT{| zhBuZy_>4A|Gm7Yi3I+JzT8gJXtGOfkKPoN^lxls zWb99)6vN{w%0F+XHzj4orj(^MHk7n3I3JxVwFy5K$t;QRc7ZfrQ=p4_%U@lZpdRHG zjH6|HVO;2eA|!LYLLfxb--^}r&zrE1HEEeK;ThdSfBojB@l;H^4FnNzWZfkshjda=X}yyg zW8e zT~u68T$G3uKMNJ=O8mB-KprS!k*->%)Nby0dTIjHU;v1hlHAJayU5}+{!L)lP()ZZ z{ac$|Wscb+T^POX4syYvIfY&``c!NBye%=1)2}xIQM5Bd^W=7q|h4j`s+2V{!^8v zpWL;{L8`$~IF1_)S6vCqQ|2&wbNc$mH+Ja2EB%?;JCd6hZidKJ(mvaT;O&xED+Ye8 z=BU8qOvoSZtLSy2L6^t|yy<6cgD5OOwh;3l{=R1Qb0M~>=1F}WOhD3sEkRUga)hD)V!ZLkL0eMESzvT zTqNcW+P>$^_wiMIpY0Je<89LD=~2KlOfw$FBYzyb4hqISpZ4s$k5z7IX`S=3u&^jD zn3i(jf;uG|qc%_CENVaP#yX1RC+V1R4(a&PFmR63jO@QFSYdxEv*y17nY*}M#Y>9M zpN$r^0CNBD$YQ`gL`(MIBxma?36{ZzKppvtO7`})@82X?G?NycWci0R7g z?j>#X(nO<`mh`D?TvtCn>m~YoKTu`-`oM}a6b@bsrwY){x(q{T6F)RZ5&umM`gs@+&tQuNX$08oe}OTh&NQnXaJ22H;2#<6 zAwEv-BIgHQ-~vbu&P=wtmrJuyglDafzD&ClJ?M12ox9Y-0DL#{W$|sK^A>>>OGDK$ z$CNdyvco2FZ@a9aWGXxtJJT>y+}IkWqLtmJkvuVj4{hJ9$j?ZAJ1i-pe)GgZsGBlo<${UQR!p6(Pcv**u3p{u00}^EHwXKpK~xyQ zUOYVDxKMs2U$4XxZ0^7|I83uNW)Wc#QpgitMj(O0Z!>n$Q9|Ie74@qc?>{0HVAuo3#}eBq zFWU9Z3zZ$*%N57Roha9f2K`5U-p|%U*!3-|n;ZQ-z(_0bZ&?p5-ukWj$-YsEuC&fQ zDzQ@Z7ya=TRA-Q%syf-Y#Zs^AvItq% z3n%sWeZhG)JZuSbrZU-iEg!hGdF{PxaSabuDRE$A7@l;dFidKCg73IIj^j!SsIe#4 zrBTMV-Bsi0%)~6%C5_sM6tCa*$^9dMS>jU9ijJsnbt$6&l(r~%@4<^Y|6U2b?S1WZ zhRM&jr9$4PgU^isVvje47!@C1XKa3K^DvX_I&@d_n+6-&vz53X(3fncKe+ra&3G^Z zD@_GhnQ&qKN1!?r`pRhoz8DQm7|g{Ww6bfv8c}UE6oF&K@NB=tne);6J4czB1(8fb zMq!8d#@w*Az^Pk7+%)2$vQU8FO#K>%sb1ke)xK;)rtTugcWGAAU&6}*dfV6H-#;ht zsuA*h*(6&cR)x5@cK&HEJI%J{=gkWKZpMFmUKucG=s=j7p zrk~n6AanaGW?vS9pxEm~y#X!BUv4-{b1pLWimAG_(4!Q7E#NT%d6Y#XzziZRZPDgY zo8}05Q!nY=RWwwBn_Bsc`j0ZAnbJduD4Y4RI?D@U6gTvQyd#3O8($JDi z`pxh})H$LGexw5?C70b44dKw=fOHaCQDe%?PyNriDhF@bU{f-Eq-&f|nrrc~HFC#> z1+03x41gQt3W&%sd5j+-e{Yz`iaBd;3LOQ5M-EoHL~-okW=bEXJzzCUTl9E2P*R zColhaCpO(1gJhOVWhV{^Icc`P!-U|;gn%PTynAhxmoJ?c|F}co zq~`g&CqodYFKCN%<#0Nhq9(|@R?uYw) zjGOG$T7AE^+IdZj@tlOOmmvtf>WwX)OQN`V}q_-)6>BbyWocs88g&rfM9v6u*0O!MXe2EzJUF0 z_UlTCX~6m#Cx+MStk&%RJNszQTfX+<0ip@!udt5@DbezwON{Y(=L-G`Snr0rI)Hs^ z5_Q|#gdAL8`7)?1A%3?02;exruw(4h&`E_=B7G3sXr9_$!aqThEde>-KERQW;74qp zgRCN7iZM$%D@g$eD!u1K2NKO#NuU{$5+j4SG?^}x*agtTp6oR%Kx^4J{imaD!oZjyUErqud&b6%q~!uqgXYhF+qONj`}=7oddp z!01noGX8W)c5k5+Zz8@WCHAZ79iP@@2E2WVfrkCVd9lpQpK$nGgvMP8w3(o1aYw;D zg#w60Si{ypf5v|}+=x5_EQtpSa4IX_U!%dWOKpXSM-ubp5Bbv!$_k?9^4EZc&`^MQ zFUyqG!Gomf5iYZc9pl|94Qxrg{1MMkZ5?Vp1mKV0bGaU7&cDxyD?z2|AVj(&>sb+5 z_gwEB2qCYD;yI0|yw9ugH3#s#?sFN?@(~iCBoX#kB8e5a^a08t@B1wyZct!oDjIFVH>7~(rHl!LQ$Syu| z|D)6Sc5n}XMVcq#sInfck|J?G*G$bWmSzMJ=F_@#283_>5Vd&GO7!?4Qn5sK#TD?` zV0zzD0-Qw(K56wlQkus($B10Y!_M0QMPu#PJ4e5|E`xug+%|a^bpca_8vxRfG$jM;m*T_WF0s>$QdjrYU9ZGA$2ia*ouEpRx8I@;v z$3Av5)^oXskmiZOJZdhT{P4zt>)A3vR}T>>!LK=;0A|- znpKh9BZH<%LGjX9YW7s7UQ4o8SnmbUB?_Y%03IXZZVwGZ*CgM4>)!P5shQx0GZdg# z>?U%8*E{{>tKg(HGw*I;Er0pZC2lvh)w}dA;91Cm>wXV1V zZl}hIQ(2)LqD@__@Q-xw zf69vG^&Jg8yVXjK3jh@sU5Ct;bho#twjQv#!?9mVOQx{=2h<*P>(ifRe7usgT#6Ko zBy`RnpP-npZkN4|mhoKHnU|4wI`~mF$aBeISrX@x-FF+LLkAZ86zwMwY> zav^4N@TF!yS;T2eK_9i`Kb%sJWo1mU=xwdlmKs{=K{RWpR})skTQe{B(TSed@`zUS zjgD{52H%jd@28<9L31@g_{l?*OigUtce-g_(?ZhsVh<8-1iQ%?gc&ap?Lp`p+o^lVcREN!)Cw6Ys2%BmZ!| zwZgI*&Z6p;t5yH~+4T5Fo$phCR(z6iI68p+FC08F83o3VLq?XEqT zCSrP1wCrw$EsCj!57`mxTArG}bL!GCN0&>)5irYP3i86~H}`RpqO0Egq25i1kP+`9P! z2LA2Fxh;)q@1+{!Rrdq-1~SaF2v>u{iOoE6rlkW&a!ssOyGDm#IIm##iJM;G6PO?e zW9u(oD|WOK_vh5GviLw6_b|_mdy&1GnH2Cn-dkM-&=!Y-QId_uY6tvZ^;GR%e;p=RlSL!Nf>g!F( zNcNVSYT7_$@T$I!`Hwb8&&V9JehE>48d2h3MWPLXaG{4|UgiA$!KZlGcid)v-&~mt z*8R5SINOG;!Y-@F+-Tpz16b^i%-#g!p)3T5DwhYa_vPR_qUppw*dxj)8HLZ@LMHFv zcLa_ZF)4HJw}I__BTKFTNw`V%nD(=*Q&TOfr^19wuhj;mgBAUK0D2dM3S`hT%{3Xvd4y6#MJPBVqa}TBOBeAH$LuiLq9=^c!8G zSbz{*4?nRgQZL4+Pw9CP6xtWfxMaQ#q7DA`uIMIqS`bJmb~jCf_X7Tv(T2(njYaGx zH??*ix#TxTmD7RvNDdmJYt6E=O#sInsw_mm6!2q)vVFfTWW5D$%Fr*+p8vYUu%rHg zt89o{z>5J-^;ddZ?thj}KGx@$$}ZJJo@vSW99+g+gN@(yqWw1?9ol05GqNP|jDF}= zK>F;|^egQL(sh3zvGWp9hZu@tLrfp6iq%D&k6yEq40LR;tG;j#O(`6U$QXP=S!8j|3+=NZ3eSVC|$l0ET@h})B>Kypv^ zqqU%H2kPS=D&Y^?`*wbWzucuB(Yc6k;iTHRyXp5kTTfJKQRNg@w3`#ALXiw)Ni^hon)*LNX-&SY)nlU(Y3mKGV> z*T&`z2QBB1SsySgc%7)Na~@Zh9zdUddQ@~I$ChR-Zmf~azN}dWwfh-E&U9SG+7d9$ zYHC{fqj;CyTJVS1S!$yuA;x5;gP{yVkg8&0rZdnM-?^Lm!->S5sACrS{adKR_6sgH zA)a@ZBSGi~oK&X}I=Br3G~8Tr>MJVrsDHeC$ao)wQ70=B9|BJu?|_br&nn&7pn}iEsTNRv0mKJZH1v zjledSEjp@0w`?M&jb>b*oppw(kB`0v@@*AqTN?;>iuAKW1($uv79l{ESdH9E6IAw@ zZm8OkJ>Ti4u#G6TDVK(duQ@4v;a>~4S{quSs9SxCWg*tu_vIBgiM;{~&970=4J|Ju zjZ$QPcFD4TaG8}&rJ7GkL|b!Smh*NtUN#+{v5Jm|svD~{(y*~i*2Os-Gb4X(sQ)Ccbj5%{;fw7U|5hnpHCLI|P&)b1>WI_5 zU(V)=GIjBl*&QffOpf4h7q3IGG^!H;F^)irMafzjd{V49 zn(uTaurY@rXIS5h#e(e3d6D=7PCP{SIvS1btxbxP*L@;L-V`O7h4goB zorI}HY{iF0pZ=T|y&SS%{x^x$|Foj~ZnmLSlu45ABsV;tfVTdXX&h@V$-d@Y^Lpy% zkA3-0`lx^kdt+-k;h8}Zr`U|phY8g@5}N-#H-CZ)Wfse|s&@A$EiCs9zW3t$FH!oj zn9!*zd5d*2G@8~6l!$B4i8ARo2icR%h(YaVH1=JqPKy%k_5XgYq|Ho0M@yK7?I0G| zEaBfS-W7nl4O?yf-kH6_T|0Yu6`6gxkJa9vXd6Z4a3fy}B&g@AYX4fV)+k+_M#JWQ z+z$Pa0X7DHF9RbJ>H|`l2F*09?yW@g?R@os`-s>|RE#XA>g3mva-pZ0bc=nQe)z0} z8oLwbdFp=4UHIaQmqIApmk{-IYg*}Irw_jshJ?3dlBlem-DK{MW9}(YBmx!DBn+rbecE11IF61r0_*lKE+#|G{BfyoMdvohtke{f@dbL)# z7D<<`$oreFL|px)lGO37ZjQs1eu~fRUhP5KUIf@C9@5{mM(|>)>o5EG_^(ywuH2R7 zAwR2YON7ahdw;yJIZ+C-KWfU2*IeiyPQZQvF~$$2gMMd>iaGoZtvk3|Td-)WC9H)U zh>LMZ4`yHgst|g!tAF8o-3fn^qWUb13RxBkxch%ZPR?B=K8pnt*mueN6%awI+nKh~ zo28{!Sol*P?v2((r#KCNCtrMPg_aqfv^M1#NWb?$?Ue-b&d5-$yU_=ab-nl_y0NJa zK9*Ab!~ZL;>cZP*O1~^as$@M@*z{UBPzLM`!eZzJq5x0lsc~A549etJRg& zIc!(6QI(_cT8F(fmf}^>4j;9-5@&5%D?pHbRq5OIabYE8i{o0N_B-T5P@zm3cz_dB zDy(1K5R0F=G?-$lBHg8jbA{u5mpyMd-<?s&PBbht0gYJ*j=~m8}yDvhfB`@XDCf>Gpm7G^HYLvM+1<4JMR#o|a_>bWa2 z$|9JqibvMA$*Orw@dGU$;gvWLCjB1`1pDZH=-3@gsL%ef0ndVy<*EgVvA9deoCQf zF;Z1oDV@gl%sF9x?;PexdsQYZVW*^O?|H7tOVahifJxA}iCLnn?g77HTVl1X>A_7@ zmz5IGVAt;cY|G$dw5i<}HcRU7sh$I>1T*IVjtnIC+DH8i`^yQ@Ln*h!t(5;N1y5$? z257{5-V!{~oNhF3nDru&dHi9X%BAP{7Zc|sC}0t_vV7|xVD4l-F(~|+e zB(C#?kdP$)oxx6U@@6@Ea$xjU;GI;#wCE+YZ;jftYXACJ!{a|+nSdZC!iLoOy$Uro zP`RLQ#5?M^*)S*P89$YAMrnA^GNS(^O`ZZWfAVX8 z($XsBUnIVf7&Z+&&#{g02f3Bx}R#d=m9lDqZU<}M`84R)fWUW{| z79?wGZp@1mB8ca_;x*px-m@$CJAe|;mDr#A;pg6rZHd%=V7dj$UGdlIUIob;QmVH1 zqLShtD1pCh$Ick^iTv(Q+w64S@*iqh!$Wb<>b*nc`z$OH7aem!hEmA|2#;I(I{g~T ztBgL+o?IFUm@Cj-ZwKgymtfMzb_aFsFxN0X>7 zxvm!XjvTW!WW=gN=r?mYWl_)R9=BYL8_{p`4X%R$mm3lK`J+=4^$yU1chWELQvCmi zAyi|B%y=lPs)ubu5uYl*_2G}#HU)!e$3>jD$T&+hL?trG{3SD)&0xV};5tYQt03US z8X4{uvVFU#m}B=3=Q$w)Mh?J&{P6XxV7`oTB!0Ks!{yd(;r*ti5Lp$Lhh$S-!|Dnv zzaLceBgLPGYge$>iqz?n{uB7{=f-*OKM4WEo$*xMZboG=0Zlr9ONe zK8&>%0D7)9*y4-c{bgV9b$F=gex+PmMa98jV7!?D3@w@?j(K{9C4Bs$=khG!j&HM^ zL3gOygfY?6=cD8uXb@hmWx4G-yjwz@!i7tIRg9!qh_}_#ZYuBq+(d89I)UF)5x7pH zONn`3=q6c2qc%RXYlS=6G_!XW7%QW>r3>Czg&>TXD3%p05DDLP5nK9=Ut3P{?rdr? zLV`z-ztK&(pTN~DkvdvwwnO#w&&^jJ*isxOV{$antiU_lTuLmj1WHbNIWv|ww$uj8 z!l$cIcX9VT*)1q_0F!k1(>G9wdDz;?vKQ2D{1pEi`QbTeTJG)X7x0=kB4k6Qv2RL+XZ zng%-YZu0po$L8wdfxQaAV~0ZTYv9 zi6MW61gW*&V#+^He?7q{LRD$Z8Bqb{J6>)i4hD~E*c!s&mu#Yaj*v%iI}#_K>$$!H zQqzd_jA<8DR%AkjZid#qN7$;s**~1m!5{|+V;9|_g^31|Kf^sAZ*17_|JcNz<3PI- zbB-gPF~82ue2%+MmQI&w?fYoBMdDMY#i3qJ65uGe)A$U+eByKWSwm=mP@Fa2R;@66 z(=KbJM-VCMM@P>pb$+8sMfXKr43+~$9e^M{5O1bz_XlWl^?b)`{*NJ<=$EiEPGIdB z1$XOMFQD}Ax|;)nH11XIg71HDLD$nEb)sDiN;5KAWohVAKFhsC^qH4{gWV4ioxVmS zxD}dpvD4KmjYDT7e-<;HXv(_7FMk7b1=`;afh-A{!f%i_v3$d_^w*O3SXr8@>Pj*c zr^NIUyf|?Zh-V0ooVComGds@}nQqPa2#)HAzfs@x` z>%zMfWQ|H*(k(rNYe}$9zzM0C9iB>Z%NJ+!P4ggT2RKn|mg-8X8o7wPZZ}wh<4upq zi+r8fS{|BO`XK*r4st z@j9FgQNn0NqnE5b;yPQYJEpJJ(&^itO&i!ZYpmgBtipf6#rt98Km0r(8cKqRf-Uz> zk2{<^cm2kN-DI{?v#r4-7zgU~$bM!b$MWb2@+eH4)Hr1S_; zb%4?6Y3?_IuHX=!+lx%lKpaXJ0YChP0LdGn<%My|2QR$nw4ZFHNYRf0zAp`+gJ@p6 z`$ik=2ZKO>6zmyMO7VVbXp&xFAjq}%T*N7w5T?ji&=&?V_EjFD_laq!_*AXO+TPq2 zU6@bsRZ5*P%xUn@*T$lU>Q>FIs_Mh-g~>oPUFz=pjgNSAv7Wis#O|p;=9C~Ryl6?C zLap35a*Nv*@a)>xn^T-h`Lyeo*&*HBkg$(6gs=Z7_&J0TA{#PO?qkNU50PQmvQana zyNf%D=6%te+%;*>vEQYmLVl5gY_NMA!g^sf!2ECuFvAf0**WJ{C2$oxmn$zCMo`+6 zExuUc9pVvNPcb;&YA(b`exrN?v;dC@?z0 zzIdbx6Hl#Wf=-Y0WcVba$RBz!n9)UM9)S(vdKf6ekknV)@HQc{BrjhE%-%VaY$V4 zGM?1EUkZuPmX|)IK}dWM}kH4nMH)wF29EkY3;H_y(O9glx>gx+!7fPseB#&}I|8Pzu)mQJqh~_V9ssZ3X5Xl0TLv#Kz%j zKMuxY9EuooXB({*s%R2y$hRb^0r_OYUsevN>*m^!7EGH*KIW}^d#csC2c-~E7dJ)- zd~P%uf28~KkJqOF7rY78mJ3(D#wRFM%ckVSG?!141kX*pM+3vB1DN1;0JKVTynt-q zFo+CfLfZ976?_hq-#iHp66zZ|S9kwmI9_gly}KJgVAEPJ=zvgXR9cmH6Mo0$$*Qo# zET6G15DlWikZgiU1K*-zH+b5)K86PVVwag?2gu z_LNUN7Yn~#5tF8fw$r>T*AXK~z@g7*-2RAKAyU7-7(+4JLF2G7T#_-?*p_x#md`Li zdb}&SU2C5%$T;ztVv~YVOgxv`cpABOB8bT7E2uviY;^yj>hitFh<|)l>VYg0@*cM>3 zr|MGKxL+8VmxQG^{KJV85+C^g^w^vwe@+>dX2G0QuFfBASFA4eYO+T(t+-Axo`Z}7 zmtw^2nSqedLk?9Rl{FnD3SYwkKG5awz;~A0XRX?9j0gFkKUZ$gFuF*l_*`ST#B#Oo zi4=>X11o)PGTDutap~vM#)sl5zfxF|6!}qc@h=rWFqC1(@5Gud@ane`7vKZ*5sg0d zZxcCtXoiY4W@*58ntvG~`HOOy{llAjd98sFB4JVt8FqNzn z%?r0h`Z0~qEzUe9AEC3~R%Y+E6kJd(!}!c>yQnG1v7`^KXSpmpX2;+~-zre?i}CA* zg|B~N)W7`rU+nY<6ymqt%s=#R3agy)x@+p>S3ITg-s9S%L7b5qcsJw;mcw_+B+i%n zSHtA)t6jw?7V~4nI5N{jXZm9|(G{V)e>me)m*rJk{6nB8?)2F*_Z3Pvd9gf#i`VY8Fi}V^+vvivblF#ElEM*kRZ-`fDOTPp#CO0r=%BDbt-9$@e*%K2g4b^}N`z-kIG>uJh)AhHBfT)49tyP^8=g zWJV~4$YFCWPRcuVW90*h_r=C;yWI<@jbhb-hl~z?eSnM9n6t=8^7ELbOcl2#o;vCaul$}3pnr2tZ*&@$eqCXQn~duavlYjUuznYq zf#BRU=pWZV>jqJ=qfP^wLme-TIB7jZ)yIQ= zCAlcpnRJS5_6)^o*8_THj8Q^`pyg2*d%En!-3d91n&=1X-` zlMT&Tka!^*vVx6J%GEv#XHxNB^UFb0!51JKT(p4&UV)ELmOkgalVa$(Hqqmn zY?8P7wq8s{o`>)UfqICt{u7TQdBmbwrBQbV@kQTGRm>D^lzeLr_x!L8<9|be0gDbu z2lt?|_gvGLYL;YU*qPkV68oJ&PPa?erjG0a?5!p@s|g{s|=> zHAZFm3bQ{uv=mwrL>$=Ek857_Zr|O+g4Y|Esp`{r!N36v)I~oU=+$St&K;*5+7y_C z#%HB-;QS_8uHJ8ssUp+hGLJ=lYn&GIV`nnkU4v)L)UUIQjkdK_(EZ57S2*26jn>PK zb&5}7)_5f!aoRuX!iRg#!t)A%2Drxx>O_RR=lk)pOd z18>Gt<;PKqJS-9`?d5}tfqG~_Pt^l7SnJ;*WLOqdSP%W#0FTQ{8S%2EHM8a~m$7QJ zo?eurEc%@>=N|jzWgor%;TTHPyrs31(}g z;+0RPCf?pOWiLKB3Xgkq0O9d1+AOWDSW&T^ZU~XaOb8n)@)mEQid!zST820}=Mjx+ zhkj|kq4^zc1^eTh0w^xdW4`sHXg-W064o<2EuVx`tn88b5hV7?bZ@bPrXPYZ{J&^1 zA-?lCFvz{8!HVU8UaTze%N{O)_OfE~=_WkXV(W~Fy&#XjR_c4`J9nyRg^=9G@w)yA zeT!XRW-lusn-jnY^^qgRuM_`lULBHSqK@YG8Q$AFQG6r6c^nXz4sW?N4el5LREFQj zUHPgAi0$DoJBEI5#dagk;u>L@@tdPl1HM=2pk;p9E>pBYyb~Ppk+WyNIzLEw!!D{c z1n*t_p|P%tWDk2$x_sP~qy%gy@p^B@1n?Eam^{&HaH2?_xoU2D#CEZS*=$V{&!<3n z5{oybUu0c+XHjg(%=fNDyySZS>f))OHRp5Fhw|O9>S2^v1P4Wd@D$(uP|_lisc|j!&$k4&=hbcsLK3iH*?;PQZfGLV zt!s8~s!9%jp>e1FKfr}RQKD(Gjmo`jKicwSM*7M7a^bd#1vQJ>~4b&-aq^RIiN^F~oio_wY)_KrR51yee zenTT@uc9BF9v+{GYL1Nch~Q#>0owM5EY8EG!l*>4M1iXS+P3rh6uX1Rkr?Xlvf2w* zL=$1gb4hi)9BM8~1T+D5L>bYfTM3yHSA#iNq9BCBDg7#XFkG zMd(^;lKRGQ{UTuR%3?Yci6p^BdJ$MhrI79==nrGrwWSRX3XWsfPsDru%R&O((JCnB zAq_yw7u9Gz#piXjB0|ueFG>!?Nl?eIPA+F>W~WHdF@JU?G5>hr@lk0E0kpHDp%kmO=$c5 z;htZRXe$Wfau^RGgZ5|!I#u?N%q(6SJB_svxhOa(rKQorvU&&3TN!blFvgG1c4`b^ z37L@eK~>o`SrS9V??|C5{#Z6Fm!mM|0}6x)+rsaMkonP<7Fty=jJZPzdv7CXxC4vSTM@hYRAk{8L13ac9Yjw+us1_hL3Na~xR5s~O|q)h=;sX4kB;=09)YBk@q2CUfB_kyfKi*c0Ek z94GZ^=`61%I7ix}usCyN_5}@W{`Lz&>t|%YC~46m**nRXN6c087;b7a9!So`^DV^9 z5(3;I@ejuh_KZ73EY{I|#as(>VT{1887B=bib!ap>w!J-xDmtMgVx%Q{%DAip-3S3 z&v$E}$-I+7x`E{wVZ$<~7t2xq z<$UTjlI(rUor_5)Z-Uf^MomAo^8`D3eB+EfP<4t*FUzuVwjp5sSn^54(OZ1ZT3#CT zFYc}it;!21#9i}nZ4M)O|MxZ8(*6zhLbdHt*Mb#z!^FFwNg$MrhFTm6FfOf}gEb+& ztk9lUSk{9GLssO(fIUgehgC;kIcKV!5LLb+=dEun6uz+XU`FgRq>-tZzAO@>x>*@y ze4#kAR~r!h>B#G1qgT^~mDh>ZAi&RxpJskE z=Zsc^bl$3Ach}r}g9i>>%@<~g$JnLoMwEjj5nDzw<^4WzF6^_-E<+n`dJi&UZ)xuL ztwCr$=<`L$$pP{lzkf(M(~6}rjJP1E`C+>>2*m{%q(xqz2qovWi#(y}l7Ho&XW@1Dsc>iwTYu05x7+GxsjWT|h%YIcmVZjT<=s`LB z!6$p`D~?CUZFW&uCI}q2x-8Dw+M>Yj&47P?0s5_@aouL`SYLBwjf;$C@yO7hVW4OC zn}kAqgWeXX3g0Cq4*o|iXV&kAJ±R>MFOp7JM4pD#r*Z9T}X!IUIE8D;^=LFcP@ zZk>W}tyGO}0FtfP@UbpX0!d-T)*x;i_!+v^zKs1EAXPB>x_hWN@!$?j^*jFJVZddj z)xgxChY(pQxCAN)2>2T9V!ugqAPnO9hX3K@+t(#x!ax2kg*VwnHB4k*C?a>h0d?B2 z*!4T;THQ=5H zcY}A8FX{7TKG3k8;#ce|8f!AQ+wtAbtq){ip2-D7rEf0v`^zBVO8o&E5{HZi>Q?Ar z(_ahfuN8s$Tir4?MdxdVrCgz0JQTuu!-=KTBdbt++RYqLzZ=`rV@cR( z6x8YH7P?FJ2s$TSbg$p|{J!d0;$=1Y5V{$dUWim`6nHR`e%}uUaq5J;{{0>Y;OktU1d&;jpsigIIW2TCsn#*|E?12KQsyGzrH2I{^DvBuV>$|Z3|XD zBE#@9aj|@$)9~v7jYf5`&J+9Pk;xSpV}2%Hmm^*|3z^qI53&ok|5_ehx1&4(_i}Sb z&sFvguq3lQTWYu7ZPx`Oj&!!`(}j9Ti$ia)-t;C%Jm&}t<0U68R?i>AgfiCaj|pkB zdX$X~#=HdXnGYy&n~glkAOs{*2_v;%eD7!SewtowFoqo`jJ=5emQ-U|VIX?Hw5z-R zz5+{eD3)|c?+p`QI4x5B?oe$Lp^(Dm#mDgQeUhKH2lq1zzW9>R>c>8mSMq^c8+5)z zbaI(=m9NbZPEAD0xz_Fj!R98|sRCl_TWHfW%4xy>Z$BpQv40?>IS)jKmHZ!dK#v=w zUh)lp`oXt3F>TX{_!_q3>sOH`Bg{2Q>yEtv?BC#GE?5~9kf7!?8wR?Z2K~}M2GL{K zM(Jf8r?LNV1W+?*(0|(*h6Ue=O7}h7X z)a+H+K1uU&fjvJElooC1W zguCF#Y_4#gQMp;ip?%iadmty;+@WXkby3qxlIpMG55sptDTI-D z4P1Y|m#bFHD`M;PsC2Lte6);oGuk^JCL2Hd|Mi>l08fsFo$`o5$?dNXjrPp(e5(lL zSnLH;=wChA!p#rP)p_q`F~4tk#J%=nQ2F*wr<(mQeNc1MkIeNN|9cOoWDvzCxefS3 zlDBYjjNFPE*=U0Dyrf3DNxV5{ceqiIyy)A2quW&Lbm{v0;{w3<6uA~zA|2REDkLO@ z4szHUVrFuEGjv{JOD%cwTC=fp^GvzX6T1wklYZ$FAOzObLzL!6iE_GcYKn9$64k6f?4)$`M6;KUJHX=lPaE+ zj9~MHF);7i;Qgbi`VZ$dz_umO2ZSTCL)#8-#rNNuW$k+U@+iVJFUKhc0|6ejngqk` zkZDsr8XU-seC9$EB|}hX&a>r{Q@WJmKD^B$Lw^H7y={0s1p(-bMJkV^CwAT+tQ8hJ ziZ&#M1_l8=-~N3Z#;?%Pm~XDHQ&=HKFLpHjAC5|h^J%0utx6bfsOoLV8DtE60tcS^ z8`TFqUSvR!YklH%f8`jR7S(4{q1EQfGAZ&_pB|3exl^_FMt^-&TV%hN%*rsC=7f%P zqjLB{)#GMmy4#2!-l0=s<_k4$iZ}(TmWa*S4Qdse_(3`*+NYz+@10UgjWmo0#q9KC ztTb=3VMkV&u`c33eUzx{qFtXuhoTE!^^YT;hf{1ukJMy7XDU4X_UNVhH@h>j$<(=v z_I?NrP(2rQ-|fwx2*~Jc#&*61%;hfY)QBXii-pu^+x?oBO_vE)3X^VKnjXhLs~jAK z)@Y)*Z!i?oz;)Fz6~h}jk4HAYhN5k?7F5ON~p`ORTqvIAUc_BNe=^Y3Oe`ZcddmQ{qU zQG7@A_V3DC1gUbRga6t8DEkDXj3i(1l|Ye3UIf1^XS@;e&#(t+`3QjYH@bXn{N zq6=Z67DZ>z9M++)!eT5p5rmxoS_Vi$!S%nl06M0poa9-KXHxmMBha*BU zPR*0mp8$sEQFeUpACu*wy#eowRNnS_xe zsg1x>pbQAI!M2+DIDDl<2PHCCOP$3_^%inv|(1;Xv@S(nI?EF>ikkP)? zl5%*rI!nOO(c(1cu$6?Vd!eKrT<$A=y><`mC?D2Bf}f#ekqiUbrgG2!TrPRdTtW#J zhHlMJTT`$YDSPn%UChhB$k>u zX$>(r61X^dBO7igM%dgtEUf%EfhA;2#Pv9r5UUAwzIWj7FbE2+YcIyKLTk*SC8SaI|2Fb#BlQxne^zaI4fF}3wR zN%~tmlD~+m%Fcn+!{i|JYJxQ_-7s-02`nVX3GHzMuNN{)eO=t|dRdPs=IygYxxHX4Cf{gOJ(erPP^4a*0TrI88OZ>JbfgUHVqO&>3c z1}Em-3BLT#SKS!EtYVMst3}bD!6qlI{ls^uaI|6YbOBXy`Wb+}3s5?-&+VyRRc<`X z@%ZSnna$%?^Bm{$MaOIT=BC&o`o9C;m&o1R-Eh)ZpA_;A;-_2EJ(V0G%KN3^2Qb|h z!M)P4Iai4~ZI1YNBDa3L+r&hFRGNEE%l!N_w!cOW!2_Zi{amzEEkus z$4KkBMdn1Ax#t`?%J&5wTB0*aBYV&NuDGt#l0rI7`*$FZKGJo=rV>$Xdl53=jkN{P zrnLPX_kG&wXx12)u3?p4%rm}SVH}k9w;Byn!HID==|3D>NL-gNim4jK&=XAjSD-@~ zd_1sm!l2m{GixDpb2DX5_wC*1l0vA!FQDSK!Fq@jV$Ov7mGp8hAPJH6N{t(~fk9py zCu;AJ2FK6W!zin!96*kJ0xUgNZqI42sgOP=4S*wyW~?0R(V|!V4^+kgrO_aSP^MGd zCYmnz$xG{gX8hj(t7j%M8b}3Lmxdv0H%orV#}yC-z2i(^L-6%wbl8o zSh5Z8v(#`?_nB-Iw6_4E3{O~Vw9QywaaH$uGbM83ra9v!&pUW{Y9XW3pMhuSV&UGo zs?c`>=lXqY{b=l~?aYR;aE9XT2>e{En%|hB{v6_SYl>Ar4kq>;-3XoVva=8I4q@vn zWV#MZ)M%yBxL9v&prKT)aGFg~5&J&i>QkQ^_gS@jYY|=}5DP^p@Ck2#LpmH_M_@`m zFB0Xxe9Bhj!qKV07XaJ%iDu6rVVFccG5f?EX5^$*n2CfSp~ z4>|hzQ)^1_&iaU6x7zZSkniLfDH5(A$6QPn;q7fSE!Byl;FWZ zd>^j&^#s+G-LX7cB2Wy~idOPf+we(nG>i+kve={C61 z0`)=WUvI(&$(~_<%h94BPDQ}ke61REU&-`EOO;_^IMKD~M*fWdVI)C|csiX>aZ~V?sNEKwfky0JK+#4zO>M!avd4l6LY8SNH1;n;7Lgna?fou1bvAq3peL z(D~d&F_{!H?P-j+?T^{w6g_|-bW(w

wcVd9TWT)t}<5p825`(7@t>InCF68vSmf zMEG70(&Lpc`X#zXZ?sD{Ul80D*2audKcX9>e1Y&C^xK3xtluT40TPDXL}cwb<7)y> zh3u7)VEO_xXbm41YB;Hkk;E;%nWf;$D+2p*)Pt_hiH&-`n}L8WmOYn^q1OeHPnf`MWsSk2 zxgtofMQjsTX$&4l3EHylWDfa<^MpB8Zl-zW)b_I=mYlBxMh8%oC!u+PX=tm7`!qxo zn56D=lGAlFxxgz5CzH@o&3E|3${Xf$Xfva+fDN(-D?K;Wt|zu6(|D4sBym@xDjKpp zgJ8ExRThMKynoC4o7hLln=kHa&0zIN@ELta%Iae3#l6a}XJ6U(gB&)>qEL2y)Ky2j zXHnr>>=d*;L*;$!zRH6n!C;gMf3@Z4Z7_R%b5a_9mG>N%@En7!#abB^AY9SO-|NlC zX=&2Jrn$Hto1~N9XUpYb&nqNPW;@_pCn*eh=1zgPO<0a8c6QH+JgNpGxej+tBC<6= zj0f}G&2c0FY;c)rDi=5k5KU)VP0H!yo?;aqKm3)M%j-UnhHVcL+#$1XITdWwH79ug zV@t}vyj0|z;nlM>QjW28#GQWp?@)s$v zrH27Om6IdD2yTo;%b<=Qp3KyI_iS-BFg1QoyAzUue>&I;FaL)VvR0qbbGES7tDjGd z@+>hqFb*o8NP8+>9O!Z0ZnKk=uhz<|$u^l-Rm1_Sm<$iA19O5c;V4$OkgjJ$0!$ct z!w$_&34pfkQS)g@^8MbYr^CX$G5%@-?q5~z3>E;?N;6E31_?xzA|JhKZW8b*b|`~1 z5DJ~~XUc3bAAhXt+kMmR*%(C&`v$jPTfA9^9Z#;;Z?47M%^2&|+wiTww{B|-o?J31 zHCMXoUc^&nA;hY|T@q0|i34kk9-CW0k<&Mj)7$m5jf0MFJ(KTbGkxezZ9c6T2pvzA3iov0oD^A0}3z9Udg{C8mHq_*nc&?&_;OmwobtepT@qtG!fKTGH2L zaXZFOzoil!eSlCd`LwfqPu=I}yu-#UC&8C+)ekp71wH8XjZ(tE|v#~(cQ@sV?y>ENN{4%0`i+CQOBeoN0j zt{LC>A4zB7*W~-YeG~;j5J9>nrKM{yQE6$(0aF_39yz5u1*8R}W7O#GMnXDANy$bI z82dc?{+@qeuY0fiy07cJ&ht3l$1iSY2F^F!NkfmmVWGs9Tvo=f2}FYfW#}ihk7-(F z)T`Fbzzj21zQr2J6M%4g#?+PO`tvTPk*UkOr(2r&Zl&fE<*TB;nOCVlj~sInR}ZF_ zn{RK@G?M06ssar8YYQryq(3n0E2&=ZC+vei1-xvnlaUPb%4e7En!Y5Jr2X6a3E^d6?7v{c&Lrcg%%FR+VT@~!3`Qu*dmd1C_X~zLy|s7|$^iOV9sk7h zE{?OrAy4p(hw43vRI+G~m-B$uI+O5BbA>BNgrehPFu z0{`{@bz%1(9>5-KuPxdJ{KkYz(YyTo*y+!kKq9-t~U&9c?Gb-4{ID;6{a)Kwz~$v-ISsXmd5K-( z7Xxk6es`5L4p*|@A(y{C4Gilvr~HOqR_Jm(iKa<9TESdbQPlD}IZkNMY*A;G+D##x zfIUgz#UB1d0P=v(gLsV6!k9uin(K)!4MmN`>pbAvqeyXZD7Bw{!*ymvi0&T&S1w;q z7fv=VFspCu{Fe63M(a)F{a@jG&R%-@SOVb3;F3xsOaS;}yl zzwxdYIHSvlLq!WHNFbbf*M}umEEqySm-q(#0GZM7y`=lpY7M)0Ln13HUy@X~64f3E zi+zh%M%^Dj*YG_}&(kkE98A$k$Vu2USqDw9jVpX+Y$P8JE3m@{RXjLLdz+=OyvkG( z{N#|2ZDmkiG$gY9T~*J2c&6|`CE?i_*)aSfKdsmgLz@f1a6VP4_@pvOiE<)^kMF|y z;?eWs^=xIaU;ZBE>_m<(Srv@g(1`gC03*Fdu9D!dxhjvDlOW!Jj|~?5?e>ULb<39k z#mZORx8`KMZ6`EW@%O*PeiLNj;?<)G6$HB~G=`e8JTK3S5awf+;+>gT{jmYIUd;1k zk;IZm9^pMWyUkhxId+1jt{4ZWz4C8HUWfjS=mS4@_S`Nj%3fY{7;`dwp9y)oIub)W zwW4DzKLyOz{NqKeA{u%1I_OrSS7z4y2nej~a_m~6m|q$@ zPC;(iwXP#4)Eygkva4RH7U_IGAmg+V)sRPXGy`!k zS*b8v%mC@PG-auF#H*+Z!*f#FMo2zWZ||McQW12CciAxs^|9bwn&_jxAo7|0-wg?* zMSrEhqlGjtMt$*mLIMaB10{y?`Y$A#cB8rH@!v-;@em9-YMRB zW0yr_`12@SxLNnFm$kxo2=fpdg}FfbV-U_%aQBhlJaDh<=ND%JyiVkN42L6h*@%CW z8_Cw0jaK|WeZO=k7urxA9he{BISWsSY;EUd59HHtNTNO{T>l*0t2jO!!Tu~@p{K_j z@4{t;@aoO)cKtixIxg7AdP_sJZykts-4miZ@S`e8dRFLI%QxcYLSigpBaog}Lb4|$ zvepH8wr-~sn#ZpR=bcUvwe6k4b&v7TcVZQ8KKU4#mITHjkIgY}y|Zg_vv)mm@|{an zOzKys=)BJ!#9hHb)RRpHuBN!_`IDS+$hlr3Ajg6~pnH2Z8vPa&DiIt|X3%3m$g8Za zz@)$!%nPMbIT8p<)F$oBvS+X?wTT^8xOI8hYWX#^p|(d)q`WoH-#)J?FY+^72!@eH z%Js*}8}mC|#^4{F-T1ZVjn7;Ehj&uzR(T9YE1Vp!aQv%5DMwZ=;>I9ef*jkOqEl*^ zTTs+X9U4w$g;^NjBXWI=ar&53<#X(d+|G^pvfUc3kMHYw6UuH?x4Kq<{kb%AY|S{$ zp8J1zhV@L1nl60As@Zs)-Wwx6*QsjDCz`r5+yZp#zWeSH4)@t(5*E^+h(Uut)=nL#J|+r z>^dbwc{eL>x$<6a$672H4iY+AMj!8>B*SpSHu3}Plh*cNQbX-5IpozZ$jHwXECRs z*^A`GBy$u3NPCb=ioD;5ZZ|9|&M96Tv}tZ_PW+{#^yG0)R5v~#b9x3~Y*}TxVRS$r zyrxk!DF9=Semn7KQ2e_xE0$WgRf@2@M(#;vQR2In8FaWoYe~uF4y2!@0UVRxn@1Y} z$Y{l01!`{mho@v>7+8YwLWvcT;DUf{CEW6a`UU9rXD6E8-5UB^B3_B}oC-LWh>Vm> zT{(p9@1&2ZN@T~fB+a3cCT0HM3Ww$uwwF`aNm}PrDO;B_;0O#wVZYn5HU_#$b@3lw z{B7zF!jmAfX5hsELfse0Y{?SlCm)A8?N28uSfA4P4#OZ9920iaO}6rzW|@?8Y#o;_ z!$mRkaHk6x$z{riI@V`ZE;Q$S#!w4)$A5P9nN9&LxOygB{@d?PAmnhB06W(F2D%?# z;=qa#7WbP>y;BKzC%ngJreRp};{Eo;_k|%hJ9A=poC|_NS?3CZ#^52U@<`DFQQ?Z&whw-bDEVH$ym9m&Qi?QMNy zdsUUGEUTnRshsj{WV7LIWoUB$- zQojJGvky%3R2DCsCtTII)4=H28fB7#_jU+pxqx~997J4jU#e-zn11UznX2dB|DT>B zp5{G4TSFqvd;74y8)iz%vyO#SlUkc=z;PS++~tbp{a>#RCDGGsTgrEw)mZ8-3hRu| zGDAq^zpVGaf?jn~6^BzWCIW^n-X%>*dkxh}O??^bysr~qBZhJtC-+hxoO3BNAH2j> zrexe9g1cXTqHJTeEm^pu#N`QHIz)r0%mCa7vF?zeb%_r+AK=To5*$nG+6L|`!PhF9 zW*u^ZX8Ae4(Ik+UdKD9OF~#qev#X?~bz?uD!GfV==~m@Vbt|{aZ}sq|-0w1g1%vvS zrBapCQGA1LRGTa(Pi(eluji+XnsANuPA>DI9ic%M`rgoeJUjPGTGi-j0`*tM_)a##m7wwDZ^{O-2)#xuLP}h61P~e=k_H0oYT&m!v`l0 zZ8r5Qk;I@92CQBWa?D_O1m(_dU%D(%<@w*qn+@Qb9d3}j>}pctvcxd_+x)JgaU3{e zqHf(P$lDDQ1!;eFfQ9Y^t=Qwn?TulOa^yS8uh;Qs68FMp{45RYF~Z$-DLxI;O=!c- zV{n~whWp@0lw4{LX_^QG)K^Clh-ouI-5Qh!jlJkQTZW{30KX03D~3r8LcOQYPeHhb zkZp8Ze!S#{eo1Tj#fN~$buHYBF!aKvY|zBIoZ`oscH5$r z8tluh;koTu;q@ZRLtwq+3IDtZETWUguJOFC>AMZ`WW9Uw#>bbwRYG*Gnyg6q%FB5W zYSLrAx{+~D0m1~%JUvMpJ$o&aOoMUwfbl{eL;d*j|HH#$!6_hKbdh(;VOR1V1o0f^ z;-eNKSKiOJ10NSxkj|&bZ7ezSryWwJ+0C5o?a}4uuB(RdC;p2|{8sy0-2YKtlgkOz z9G}31k}^7@>AghT|;_+0K=Kp6ZhCcz-eMTIzT?-^ z3!Omk5In=WruLJP?_CW{>iU0?cAK3W&)adxsV!g+ENX*%KYaHhP7@b@p;~n^FHO&+ zu0_+!zMJra?l|%4V4=$2RDILgA!Vzp!Hps$*1WxH10;}bXwaIgOahtuZu2ATYadUE z(4M*{Y`SwC>nyhBJ=NR!LOn-R&0)<__ZeFY-Fk$)N$F~37@5Ej_Z&HBe)&44s;aQ5 zsWbQ0wsv{&X24I92iJrO10b}U1Jh`aB|=(|ga)D_=)}yKyjuTSf#03%(yB0|Rvk|P z$TNHu1DYaLaAJHe|BIZG_qb;E;U)~)yLY|?*o_qH)T z2J6DJz%5)<92e4Pxe6DsVL-p5LDbPoZW{SMdK}%H;(Pav2A_9yua%7YU6nmE#~ciI z?vg>mZMs_O=l61Z(W*~ZjZAnzM%z}OYSk>s0d^89H#Wnby%hjTXAB zGb^I9v4~H%f14vGr1!v}{TrA6Y*yHMJ$4Y9kCL1aF;2nLk))H?-bbakA#@ z%C#|aT2+B@Cwp*pABsH3&u0E3VCts1oR@%=or7JtaTQSz0pbMhEjepMnFhA~n15_9 z%o(Bjh+h2sOnY54iU^oN!W(wRd7b=3k^i4nG3 z#ZPxCgbb}DZkEDhPw%iB)e#aRyc42lC5q^GDBkV9bSQbHCqrbrpz~Hu-p8l(D??=r zI>5w1RSJy>=U|OqQD(|ASG38Uat|&6#U)y@3-`|ZU|;hnT?8&v+kK0cj(EhZ*Qx}B z*fO=bRnwLL=&-gZc4M`(HY>-?Ah=yuz3FWi=Ee5bMqi-8qVeJFb%JLgNGD#m_J@Fw z3s?Q8GrG4wseccPY$*(l8X-EKUPYpqzu_#ej}7m$0fAfMQJx)dm;H_C*$PA-$Nky; z%KIDBrji~`6w9peh#cvG@Ph{)$~xf0>HlOtKc<{(T0nxb6GDKS$>3VUUJaA$!3}zV zwHLzWp=Rb=d!Mvf1`qL*`?um#O2=ZE7qCxVtf@|WVPuB_tQr5r zCoDDHT8vU#pPRmtGFr{^VIsn>f}l;8+fh1tW&RiEN7wXf>Fo`tH(Xu7?mJuTlk3kV z3U_Q8V)Q1)_iY0SZnPg&#zqAU8N6wWZItjiUc`FDr=H%joTzVgMPkKVA~JW^q66N- zYg?nLbXWqs!~0Ep?_d;G*5`9U$CE*h6kjiKQr@M!fo1OF`a3eT1vG{&P=X32)gG-B zwiE6E%7`l%jc`obc4J7*P%E{ye$rNvkK{hiLB{k#;f~OTg}-ja3lH`}9ESb9lZTO;#AU*M7KA1xAsMbRYwq(;;ZtFm zNv&-t=WkmAribYdXJq}gg!F2OMzx4U3TKO`}&|2AbcZqvq>LHj@GZ1GK>|bep1n1 zngG+`W_Dl|l58RASx5ss!S76G+qd_z9>-wUs1Be{gyvKbZ>h0KZc($=*!Q+lCJAr)@zb1;=pb7S3Jn3a#x}h6r8V zP4jOqku9}@Ilo#{IFk|-aRAqn4D)q*7|DZ09B$Llws)@sRFHWXO7$UV1JkWh0tq_d zg9LTHZIgczdlBwV_$S|!0paTBEvx0SD>x&H+{Y5LSA1xDaSVR$VoF6GjEfyeVI@o` zm4ggfPC)f)0zRWtafR^Xl!(pT3IwS}fUw8bU)blF-CQ@T}_WIdr7}r=DVowd?~~PV^D#Ejk133+XXWpR1+CA6mK6?oc#q zzd1>nDM>BLz-f-~)>A<*sWnzMJ0 z@+?q-#a71pR*~C!-w|&ll1mB+1`XL`GtZz==Cps>g$wQGTTem`QW}=^r<-2W5+=Qj z*SGt95j)sjqEpY93%_UYp40zOUYU2slIteOIXgHLjkUdDv4Qr;1`uw$p!Ln!@l}4L zz|_XRc2pLa6=qY2F6M@Jh;DI*1ekYGI&;ML+B%LC$681I*zXh_62Ha2!#lV#U?J{5 z%r*7FzMNy=3_7cdj(3+K!S$?<^3e=i*C_~rxs%|UsHR*I5B{&rxdmM$L)$PDI3{7* zRUGPpe-?9;JX5 zxj#5NadIhqpkE_U^UoDBHM}%c4m0__vx!KAqLWG4XHWYo!O7H5b|`PdHs(T@~! zQ_OO(XZtkv`jup(YsPYyy!Snq8U2q~wKyst*qX;cS#)&RFUa*V?2bG@9bLP%WsiEy zh3w1|KVMb+v2gbmvMkM(lphtt|LXW-_EmE%l- zR=$B3qJYYZuav{9^_Fxt-0Q6_E9}^(hRB#JL*q^sh>gS_FVKg#^M0;Rnc7~28t>oK z9Iw2I!yik#YjgOwR~Q}=E6?RRpjO}Yu5k|j1H(shb`y%W_=TFWbTzebcoqbYfT) zG%qjG(tckapy;wiMmoXTrbTFfwi;=D{y-MaHxS4CSCsnmb0XT!*A}8U)7&DxPY+J; zwylp=hk;g?77r?{nX`&h)$MYDQl}2y#2>?6wZthsdusKGOa^kbKO&x=0@D=yZM#3z zPZpg=-et!UX^%*V&RiRWl^yC1kSwHspzD<_CC-0apcJ}msrZq}J6KiT8X+KlA!mO6 z-e=W`JB?(UAy!w!mif*@-?O4PL_7Q>;KZ+*i3~k^|-(8uAVfL_vRC zip&=KuROZa$gNr$!Gh8cbnyuHJ`u7SW?)3nF8*a2E{Ggm{^fi!iq}(_qY#cym=?x| z@{WF9x9Qq5GXW zC-CuLh1`X=cN)RwO`OwbJUINVzv2^H@)Fm!1b5X554PI_k*x1Z1pj1i4VLs@7QSU5 zB>uRcxtB?sk-+|F9|NWT7b?RX=qsD`+FaAhs#zx~)-sj_iG>%WTNUFiuW5Zg7=%`N zL_ODj_}pEJ6^YY75_i};aSoahY{lkce5Pn`XTLkzo>;H0gXZ;DiRgxVv2andm2V>) zk~arNhjn~EX21G9xR=$-dOYr|OvsW@M6O=eP~2$y&Nje;LgcyOSC829zrY+#7Yfg9 zZ!J?lXqd;{_@@fXm3T;1>Q~m~fg~iYtMX_dG{5MT7n4M*uvN3wgQvNRbW^j@w=0`z zg^w5dm}DHn2%KaG{Ni&S;O`m1Ntsu)KWUkYr8)aQd~@iqli}#X&g`l7PbThXBcVt4 z@SyY+tPn>ziP_2P$>mcoIVZ@ffavLb^;*7el5BlHg&uN^_C2_S)g#a8E(1=9O#Z`j z<~#Mi*UH!qs6R565Dfk^nq75d>(YNmEa6nh#Jl6(>;^^!6C1}&uDD1`*jQx;) z7_@K^IOP9~vxwj#+nNQF+BLy`J}5nJ6dLJX(!7^!Xss7t+-Fn0to#X_Sf&eXhxATrd(vmjcV*2#qm4#W zBZXoC>bS~ffZiv*oz9;ttHR!g$j91;=%EqUTh~J)@^1bj72e_TSmIj3Qy@;-(W4~i(EeSahgA${mdZ@<@X@1phg$GVfrDpF$C{wj>#%t>L)5k+ePKZ^k&v~cw03z=`rQG-^&jCbe1qi;{QpvS7X znkjnYiKkFhCw)tr0(7&=>l9xji6;8uF1@LaSDI~pXXurKt1CT$c6aSme`^r@3QFvN zWvFY?s)vLNh)5eH)4wP(uR10GxW#Q1vZ87ypL`N++MG%NGIrA1CqZnxIo8Vg53rVbSx6ISv$48y)NoisnQe z6*GVpooR<6NDWp_Zb=7D)+)UlsngO5PIaO-JKJ!QKg)Q8e`$8*%s68c@?>8Ed5Hrq z58n z7dg4Zie-zPLI|GHTHqlt;wT0cwAP&x`o&f0pJyf89WTr}{&+>5z?+X3OmYv$r-}~K z(!-wACL^9^6eRNuF77Is7Sz$WLNDE;;AkFGx~A{16Tqta1Oio00X-Gyk~0 zV`j9NACB=^@9XQll$y}MK|?lT<@pY&>0!~9Gbhmcg68mi9s^>q;qvHpQrtk zWQ3!tYQFKAkureppo{ z00{8lI{sBFv-VKQkXPvmd%T0*l?!TNAR(5eDLn?ud4wx90S^k6(6XAdr(#v!95jY_ zYnoRtCbiq8bzM}*+aAXRt6ID`qN<*+lPAOtW9Z58CeH8ggA2zY%yr66M&ScDn62l=%0#wVcszm4Y(7pfyaAM zHy)8TLuG{OjN_8tC*IiTnl>7*w{V81_yuWAFqp zewfR$+s>9wp{Bn=vYQr`aIlBb!a^4W)Jm7C%X zim`(G5zW*Cm6!Dm!M{QV(Y(e>0m`2pKaP29Rhd~B?0fzmOm|DDcA9WE=SR+U%Flk9 zT}Pcb&MU~4lQe_{opOBo#4K|@^`$pke?el%vdr4!YU4`bzDjIkMbo}vY`Aa+0J4c* z@CKjtoD4lg{KjQ1T4ua_H1Z<(nRa2*p;BIf@d35lLF3a%1k-J;oN+?g5qK0BR#}Om zBF%)nmkTkf8zZS19ufIK0NCP9M11%Uk5+o0iSJT&6NaJm7&H;jmH`e50I?KCdb@XC zW$_!@#LalaagFO93j#lkZeN%QF)%xW>^wg9dhqpN5;}ukEvc6T*r4?hpNNO<%S!D< zY7QE22i6|kjy%EhTVO(+56a6Tm0Bub5tp)UcAt>Wh1Di~N&&xH;w1C zYedT*t?*VCJhnEL{jUtruwv!n4TpJc-(;L7+H!&2L6V}sr_lQR>)XGZA~HCGH|xUh zItbs^M*OyVOXbCh6KUm@+UAfuDSRG)_TiJT%)%enbwxAOf73IG$qcA$rOmTJ>@)nz z#DJ8i@)$3Lm=+x}b3;urF?DB*0p|o8m zH@fBB;n_w^me7d{A-Hc(;Q>zxBv>pM!6|RV%D=nB81R?KYybkQ7nEoEYq+fiI^pD` zH#n^&z&*P18u_>OcE)N8OJ-=Z%3p^k0hDM5ri82aLClTutDvQ&$W*hBpo2-P`bfOK zohEmaml(n#s@G>J9b@tGIp3|Bx~4D<@*o##j9QET6p5Lc0_6HRy0_1I;+GI@+w`6) ze+<{}N_yPSFuX*w@|2tS^asA{!*X|_yt8#GFUK|r!0SrvN)&yt{#HvF;k$4i9Y)Jw zf|)1mET8hi<%wm|ug2`71+jlT{P!ZhOD+qsoe6T@q1X7q?B48ft7QxcTHo5m!QP>8 zh|$2+e!iB~S0Yj@DihJm_LOPNNt7m*IigybD*Nx`2)Owjozg7~+LhhiSUCYhZD&QaU$=)t_S3iwH4yZ<>QWl&Q%6LGN-+WIV~#GYQP zX<;EPlVhf%(ojU1)rhmnU+;^`n-6R}v?Mt2JVf)45Zs9MY@7F4!B5k-;8RM03`%5AZT7bi z=D~t`m)lt^k#4_UeBJ^WO~4Y`$#LMeWY?$K)1x_#<)8<6_%24-A zl7W5D{mVuemd_5$KU*#S6eRl{(%>KHEk(ez*U<99ktmi-3_uc_^8>~~ZwB0gUJHorO=?bJ=biBwo3lp$FK|-?okoo5npzQU(-D zZHlV;dVDI{v7iR5g2X<9_PA7 z;pd-wE8Z{NWU)KXgnevyKU(y3`4-v(jbB$4F=xqhTs3KzcPrXScG;HiGRK{JG zK&C4zW`L|;2IB;Kl3v=H69evsRQo}X&}}r~09k?rl#>5lhXO{u(r&}?PF9tk!{qEB zEhwNgp9bgv*C*T)-qXs^S=2}OAw`W*`w;=+NaX73deLb(yo=zlb!?ym(KL5~wvbFN zPdXQy|NAy1?pWk)utPBlh8u&hP|3GYCn0{-HT;3gc4*2P3osjfPJ=i;5=ng_cyscP zKVUBXE48n-;ISOf-TUQe`YZAhh7NNacomKgM1gwq$S;W+ItUnsZNglh2^JXj<;ZGh zskiC()s9PQi)4SLGiFjXnDA?chUea%kpmi z_qp6j$L!x`eJ~uEag8O|nijgITpm|%48rnS4YPi?`&{ZMWEQriuQ9rg-RCsc^MK36 zGxTNpPFr->X(_WG742S9nAwt7tIKNg4`|a z!S*~{F$<;>a%%2(D_G&=@GQczxJNvvy{pZ(Gfm>7*#Ob*(i9+A`R`*N6YGh_Z(CzI z%Ke>6DJ2ITV2CVhmvPnWB2GbUwN;x>>5Q_of$m<5r~Qmo

v4oLeD|uyN41evsX-b8Td)J^9v_hclwJ# zzkzVCy#6@SC4kV+Dcj<|N3WNstf@xf10%YU>}H~}Cx9CV>NlT80ZrW!Cj|$|5P|c0 zg^rGy#Ebj2xbZX;3X9;uINXx`J^2BvvT(hpkGZSv`WVD=5jh52bcNpzWdL(B?<#$? zr9oe5(9Hw>Kn&%qmJ1c`1}oVxMY(buzpXacE+_H zYL|{|Qfa#8!r5bQ7J4gtbP$Yk|KVl+Uq@pvcWeun!Ub~Zif}umk53Vp>2r`M7dUa| zCNCEim>i;xwEomXfE=SU`wveZWo}s=;E0A5IT&V|-hTWo*KqcJx_mA5ZS2QV9&%lR z*TkrUfj|$M<28hFHN1nn|CXcmd!m@Tdhr*@G6xPN#IkHD=bel?{V4sq(V7UYV@k0* zVd~J&)pez&q^D5^+Ke)4-;5+#zpAh@|73Q+)Q;y!YhEIX4xxS)JbK?(YXAFl&rc6b z`ttP)RxGYC{X=>nEW+g)+TSP=&bZXk%P$%%8OBh5ZdJPmHc&;4b%*a5SDLdniljd6 z>T5-?u8Yt>7rh^fh9tlCTbL}t-Q}mR_!i+OLft|YcFR&7b0nS24XQ6S43uRw=1`Fo zD%&fR(FYBLURXwtqQU@j+T$RfD08EQ4`X(V1=LI~VqMk7_C7jYuG@ zu_`PK^=Fs5J~CV>^55FgyU6c6x##__mqXu}t|@pqek403I+?|@2l9YNi1J~v+zs_kG%ia1V`Mp4!QA<#Bw55!+vN3_~ac@`ARi+qlbmMdrrLS zz{_K+ix3iGKwqacKl5bglh-T7N-f^VxyipoBTu}YVIyl~SmKv=N&)Zhi!L+>x2_7y z0|39_S%J3db#Ux@=fbbxD>IR^ea$)UJpJ?3qUgPKN+Sx+rt!*HuE!c)Gd-F*dn!?X zVtwS(@9SA~kR^RBo>S!P_6xU-ImMJUEqqM#E|+A0!D(B!GqHRTosjnSde5et@<5D*BsB0^v1}-2%#An1OB;z_bFl^ufZoi z;9*ijRaF2fs^Q(We`uU*t!e{Fg_Uo>HV27YG!k9-_{r8TU0W4!d|*L0x#c$O?jv^E zoanv=`vT*$5I{z=j1etdMw2L#T-6PJOEGMw`PBn3I3~>DtKN%BdsCp8^yhm*DwYqO zd$+aG4PF8vgud6RPHSA?ED;(bZMxsQWBXhDln&zG*Hyc4e6k&AafJ^I2ROk- zI25+db#d4Xmu(j=o4CQPXXF`;&V%!KRzX(C+)2&XaevmZM=Pf*!;u~H?e`th#;!by1DUTv%_8+~~&-#UeRc|C8saH*XXNNk7I zx)|;%XV@=&d7?Z(sNqnhe!Q(4O_lZ}d1k?*qQPG)#ob)AWjyrLLM9{yOMHJvwtlco zkb&zL&+4HDBg&7c-}NP$85=7zWrCJ_oER5yd1s<^YUPUpj3O$}zrL{;Gq<6692e)k zrO~jl>wkf*eC5u^xrS9go4Za}p6%5`4_(Qi4UzGO2mEAF1r|*Y=vB7j?{u|zuy56$ zGgy^VlDr5XRJ{T(8FsI8pbZf&(dMKmMt6V#kfMRBzRlk|o*}Yx!pW&~RaWW4=ej3R ztN4TNZr~hX)BXw3HZBP5^c$8Ft*3|`4N;j@97xRx#Z84E8ta9WX{Do$HNV_>2D`K7 zQb$kl@XyoxyU@I+7)rivvcXzp#GBa$K3OuHf@NdaUltU_-O?QSDqPyCb&e}qd$v|t z`KyvC;d1^1nT3&Zgp+l_0&;F%ifJTvz^Zk?7N z4Y+=xQA_MASyrY@>94q=Y4rsRzKLQem(BM{*p2!Fr32cNt;Ar7S#WrwEsjn~bSeV9 zpxDwm_c?b+3tkgEdbq9GeioeZvNqn3I)#-ORYx&)PA&cPrXC@uca`4KK3iQK04iHq zJxJ5GpVwPi*CLrulNi9n7;#0-7=6+>t&>_J#}_Gm5Gg;8AI2-=H#%|YkEM6|j1ZO5 zE6q_3YIWM}vh}#zMhSJ3+V+B*B*Sr2Jkxlcc|Ne}CU4I0 zN$FQ>lPMFG^8dQFld@ZPS`?R zu2%1}6ouT(8Fg9&8whbUa%s_w5l4?cPtBW{ephX9{=in(nGUd;;6k zT-l`R7)$%(QvEA|`zwv~6q7>F^(i?2KfKz!o;IOp*xx}^sGR6A*ED9@h0vOY*qO)^ z$L%*g1b=j&?mQ$)r1%*3L)LKXR``CW$h&JZMOiZRelZ9rB*eZ6MDwm;OWaU}W`FWD znx{`L*KDp$;7x7GFN7P@F2yZ$rsdrx$Ka@vp1( z#_X%iP)5J1Bln?Rj>$_^!aIH($m4YJd24B@)0-XN z!@8MH{K^1+CkMe-tZQnS5%UN67kkC3Qz}@|=|8fSzCEAsve{bo0&p3dN1K%X3JS~_ zJEFZ6F=LRgxeo=1Iu!;ynsIWpcz$1@Grvd6@pX)m+*Ng|NiML}umC&Z!V)RYaQXuD z^>f#3VVkCC!|<6j^Ww{hEfBbNAspruL}VA2>J<(quYbNMB{^M?>z3`fkKw~D#b7Vf z#xi7yuHP1~_#=j}pS`1_)Nb_syfq$~V7h*9$1en)Xz+HD7f)18N*Okny@oAIN&CMl zGnrVu8TZuBY>c?lx#PyJ;e#U|X`nA(%! zZ66-!h;{LBT0Y|l01o3K3S5@;`Znre{VQf%M}5=Aihl98@h}p$y{q@%)KfYYT0Sy9 znGvD1r8@=JOMsEaT@mz>Q9HBF^3=-J5D(pWb&a=R%8hh;fg)zpZzqn0$s%D-1FGDH z2%DJKMTzHDPjNO|mCqQ1R>lpA`lsk=}-3M&TSCH%0hxpU;+Ve&sMh*MXLk7qRS5 z!1pPou!2JBvWBLn%bVQFjrIh^S&T!82@BVPaVtJ8ZB!Ep`E2YJQy1^w0(1sHXAgu6BnW7(I{W`dN*2z{GoNMN!wi z7C!8es3xMW!y}u0wCR6UJ~VXo>6pIV>3(&&?tC%fd%&-t4u{`Up*Zc}`m)Ih5%>Bc zk!jBA72sZpCqfNH7pvgpOxcs2L0o$=MlfdCjNhf8!>A~mUEg2S4qkUH9BTX(Iay`X z=1zS7qjV^rJ84QY_Y#H!k`VNqus1n$h%+{#+;&AWXcz)Y9^C}6WVteY>z*X_k=(#} zYCI6M4O$c$E`?*FtSd6U(cvnXev9AXW_1srZRO2ZDUf73q!Py(P zYI|u$4|u`_1_c&C$M>7?zfk1uyhpuWJKqy|t#*l5fkzfVzPM0~wp2m(XL_gHJ4!o%cznN=v=IZMGe-<*q1)nb8^h zz($4tFyfzw+!(U3eQz-d!ZGE7UHO+&{!X4PAjH?xy_oY09a=dnrx^#4Q@eGyN$_vF z%5)dixZ-3Z76UmcUxDafRH@B47*$^;S#?@fSQo9o3tD)bnl<)ZhCeKFn-dz{sR|8G zLZ{*`x1ruF7lSrdQ4o!gvPa0Do+F#C(&ASo$I6=^KB4Hu0Eq<08OWF->RFhN5Ifts zQhU-{AN+9~A4J4=>dS?_emuQS6-~X?KU*_N^%okmhVy^#IecG0-n{<#iR4+1tWFs4 zHB~=E(%6L@VyV6ml0k7by1r#}$!3%~7&!UQp*ao|&v&m@s4q3BUJe{RHmBFFkW&4U z(}{S`3RSB@i9G%Tv>}UbH7(_kSf-?J_#E!KY>m)p_OCvkyQ3vOY7pSYQn)-tPxUOY zc$nZku2ycV?GJEftv9e_6g%5X@;@a2J7P>fgST;Hp)_bZT#BP_JYcQU3NYJY$PORenp*FiVFkb2Y0^%yfg=<}Gx zt!3~z>^!=NZ%y=`Cv@$f*WI_#JGY6CnaYox@czA&{gft!&++xa@ZlRK)rUnC&c{kY z2bF8@Yoj?i60JX=V&kFJd@)HEGS;%re(!T>kl($X*-_>2(=}X<0}j-p=(QEy_A^4wFE8Y~Ty}@+gh?lR3Blm8^*YRO z;hV1>i+;I(4s-_vLCx7-_F^X<0|EeRgt{_|wP-*4W^1U66@D&9E_JtQ&pznG{V&So3X+EKEuFyxc=s)D@OB6O`S=u zfXlX-vn%M1>n>RSvW+BI2Dw-rJT%~!rCG4AJon^X$nN0JeU_7E7FZDc(7=1DdAs^e z&q1E459Z6Q!rj!%=fn4@lqtDrovDRrHEoA2vCy3A$+tq!!BbAo0u3wk?nWaF@5^?- zZU~KA_5IUni**9>TvlH#_d;n7uWw_UGBv}KHD_4o;j#{-qsJ=v;}&FuOn#SDIilU` zynp7@txJaW>+oL_KLduom`hMdF_B6C3J-jt?tgEw$7d^?uA(0!%iO)W`QjisutE^R zUgjb7*Hs1^Q;B}3>%`J;P`vELY(dW9p~$as%^7ZIqEZxh2XO4G?R!{2Dx}TjM(KM~ zL$=2w?*GtJgzYh#D#>&yt@c^c-g8-+L$^QST7Wro`q|w~J)X%ImlJ$oUCK)TlAbom zsXRYAXmIm5_-YR%9jLYr(sGKu{rBB43?e(wDX8_#Nba0>H474YI4qT7ak?G*xqha@ zr>FQx1(g<7spOqyX3)FUk<3?`b94Ph$swy1>2;D93%_SR0B(;N&z%e^P4D&vin&2| z#KAC=qrT0S*}b!SW!>t?_?5vRo)pdu;T6wMSY=?b-$Fjxh9%mhI^U_pZr;fk>W7b; z`z^O$Q}8@!dQ$l8&Q%#eX6ov|eJ9vE@lg8jgQFXLzEg5DuB!XE60;9U3{t7Vwc4Y` zKgzUy@xFa^mtwS~R3|Q+)zvqa2vnyD#>yTTSe{@BP!#FIPRq;%%FOwn*zp^+a5zw5B5s<3oVbdkPwE6!ah;7bdi7UPLZ)Djd(C4`GCMdiu@ezj7|rfbrjAD}1)gI{ zZolOQT=GtB`+erZ;6v}`Gr@;)?J2T@+l3E{YKZWX5-jK2mzzTylI1=C%&z4`Ki*ft z1BnV-kV7&1O--rzPbLffUkt82u{NG%ypW=+hN<4Sw=2y{%u4nfZU*>jzaXuQ_FA<{ZPuZeM z+?C4ql5h7?_Ph6$^q~7=h!DfmIQ6e$a2s0ta`k=+Kj#l_p)(ZcA4Ohg*Ns@wRuxIRJo3jDUIc{HtR~ z{{VuFe$vy(7{{TLugIjig@KPU-(?u))0K#I+dvIQ^wRXoP|-zBbijP^EB}3fr(w3o9dY z+d0Mn=z8J!qvFrSKZ!Bg=zkDAMR}{*q@c!^A_nrAjyIuLk%N6(2cgK~p}LFhKfHE_ z@~^ocbuauBi}r^W+ z5M{)?e8hG*ugxj-CY581!xB#lc~>p>RpnPJy8{7R*mtIBo-a)gQnb~x&4t7_CGrQ$ z1cBS1P!}HJvi7xa%go4hVD)Kz$LJb=!BIXu+(W#klN)yu>4SrwN6B;4au2;aPxvZF z$J=67D}4^(RDrl@t~2eAwW-a1dVE6DPJ!_s#QGa3@~ri}MF3I;)O&`OJdBgfZo5yR zu0RC3k|M+fED#O==fBdrooOzw?-#R=W%nPYZT|oTZur$|!W)}C0c33LU$D3*Z(IX@ zGux5HXs`Sg*W+v|vVUY~WFF07Hh&ynaq0dw`N#10#`n;8I`6~UXV@T{QfVWRFKle) zhIRAHtg_t88_pUyRH8<#s5dvtSgEf!)jTudWW1Jpt34|GOv5Q_2}o3bbY)bJz#Mwk zZy8oCrzR&#FLNu}_{ ziRC*ghc?$6M(^(*wVk8f@$|3E{{V+R0Qd>0Sv8-Cbo~m+jlr4jH3oHdw~=^QQ|9uQZIc!R*+1btTO+efmuic49(Q)EnVO%!f$%%Olgmj`JdF!UpgKec4vDw7Xm zPx(>%9WVSAgW}D*7~8{o6Eczr`!&uw{{VPh*T1-MB|`qyGR( zzpZ{k=pGIDN2F`%ta!J>{{U}mxDxW&16?J%rU{lGE^VS*#AmpNvjE^gQFTvV} zh?i5l(CuTjEXA!Y-dOFnvXB5paCqSI+@964#!24NdJ*iY+RiBb?X~{^g3EkRTiC3x zd@-O}qv!YXt=Ez;G6{9LI0vBj>slKB0Q?p=;|75Qa^CnuLv@&Zj-uypK-tK-IRqSn zGBIC+UlV*K;e8jyvuWCPoVG~O!#TG}?j%U0njrEfKd2~Il_S?Q%S~3?42~KbsNC}0;1Ca9yw-0MQKdCEVtth>+m#=+7k}_ue~Y39 z8h?ZIIOQXUbh~h*^!W|K=jpn>8~FbK{{RI9_|yAG#TJk8Ust+qA_rLpk*8ZFmYiJ) zD=R>fyUxQS9(=L@Lb9r`&3x~5@hef8WQya#UkjdPa0jN@9eZ$n>qcm`W*X#6h>H%9(DuLWouZkZl7 zo?9_W=G2e$kU2$-5M&kI4BniU z>Hb>&hw_V}{8_j0h0c|({6f~^y}Y>6ts3>O-8aEBP(ZK%Ta}5R!mfB56fQZh8`ZQu zQcY6kOATvK5Hzl1@~pK6k6~lCVMW?k9SZPAPAm4a;P?CzaAaQ*ZZ-mdhk<^P|S`x9Gf|@NNV)@-@Dl@wZ+} zbR}Y6?0r1p2MRJUI39oxYX1Nyv3Z>&{hZa>KhtRN>#O_4eNX?^`wzw+75@Ne?*x1s z*ZwYems!4+-%Fa>1-3VKer?2Z&lEeuPo&%es?U%ga5&=?_>22r{{X>j{yAuxt9&Q; z1Fv0KrNx_R+QzeUZKb{1$omhIqutrs*?B%ntdTs=6qzJ>A~r(x_1Ep;@pAtF;RozJ z;@fK*=_Y>>-f6lr#pf|AE!Es*T~^ub zAW}@17a`OnG4u?~4%3l^4Y|swBz?4>@JkPjf3*k3y(DNa7lm|fZg}l%ZMAr0ONi8| zV1m_ME@Y9gPCTg3Kzba1K|ix+>_y-Y+55uQ(|iH&qFfuBba&N5K|ZH(8=z%*A9-|K zF+NNxsDzLfyYX3;wt_jKLd)Ji7`_C4!9G3xk-Sx)YQ8gNJ<+U(4d!Hzt$%9y^^Gs$$AWa7M#ka=8V8B4r1RmDBCZ=whBYEZTy6~b z&j$w`4SdVtX%Tg+n@F$~&Yq~ljBK1a{P_CUhh84hd)ppU8)(cR7P(BJ7dF@>B92E@gK^#<3b7DBZ=m9 zrtY1Om_9Z97t_8D{06twbd~XI0Vv0B4Vez8C(`xq+@U%?LXoQRIm(VpQ4(C4**BxD$jpJP>>n;_rhz6ZY>L2aW(kRTpxlJ4^3j2z&3hF>i}9G7F-zDNH6g1UTMwzl|rci}w_Upq_q zWY(5y@!%!apl{=wQ=dHK2|$i8c^*avas_3P(u~|=Ew6S?VO_<>D|5r+wby0%CE%?; z!~$f7!%5j_Cu1YD4>QQp?q*`fX$vu9^EXq?dDfTU-78nWmsQr8VX?DAad#^pl_7|3 zZQZaG9R5A)-aJjHCa0zNXTUMX<=pst!oDB3yI<9sHPI)A;$WCXC4tBA@#qKB{{Yusg8DNwfitHsAEb<^DVX|k^vBWMF2pGwAaw&xR`w>)8dH1eFPeEqTxYwf?;W5UyEelN7wORzS+ z74b#LP^KWe4}>K6PveHwHtz7%`L~XBjaZF7=zu)YjN;aB zlrR_&$@3hmlAod4-h~O0*3qL@^R7@1hX-;K`Hm}))pRH?^}%m(3eOz=R5HfO8aIu? ztE&USQVs|;+iKQL@Y}~)4~2C`d9+PZ_$HTgBW^Iuy+6>VC%73fG5hP4KQiR~R$pEH zPO6uY@yLbdSbh;*;>!+-v_6+5ry4tB!CWiy1u$){79#K7S-m1 z!~Q(HlMKtPUPl@2h?!6^_5kG9o@3 zwuTfwo@23YzM%z4uUUk$c*ZgJao3D~6>H(V$p?q8%)c`{(DnZSWg8!sY5Mc;wvZ4{ z$bTLy=rErOJiJ^|n@H$BDO&Ei)DVIsxbVF0bDWimXh_G?J*(hfiTahsh2yK&vVuAM z`x{w`$O{mBgkFG*=N$VCSLk=e8*likZDYpoKf$`z2T_GjMRi^wCg;Q2bm|Er+9b&Jjpaf= zTIRg4C))Tq+IaUF$F+SO;=K<~vG{M`U1rK9j!T<8O($SK);+$dBB4H|W06PD0bY_( z=H-1)Mx%C8N09tBdx(5rV<%~?}a*in#YG5Ow(-( zUq_aWWq8%5=)r*m{MgPw!RT>bY4C!^U3cP-hxC2N{U-6D+@yy2t>#n6yNhiH$d~RVGBsO~$RU{Gh>BV@>>8Z(5cUm2dw7wRx@r9M`&YHFskupe@ z@=RL?%E~*-B9am2D>Q>?0U41&7_9w4v_-VJw36*^red*xh{}wKy_!U3360MDCNs5i za4-&)qIfev)?vD{vyV}=iGO$}i|pQDDYftzz8^e^0M7560D59-`o*-?%^sO=3$Y&x-45#@;sw1#3x z1b_!M_b0)tsINQ^ZzC{R<8Lewa9e5T0QJZD#e4YdY@-B{JL_SSrlT#70)N@l#W-7u zCAG3xvPdW#w@`DEI`P#0b#F}g8zSArFGB5*5Z+PyjkN8;fN(N0aB>D~?5QsFX19`5 zibV^YU<02_`uo&2*Gn{E-2#Ki+y+j2{*~l^XjOV#;&=*a-5dvn`~jkPXG4=ugaahl zGEB+=W6*kGM{msYTtDoqs4s;-YA=Sm%-fl5{8gsfI)Dck(|=|*LFxdyVtt3_?H95U za1;0fs~*GD*UJ9@+RxxGi5C7Tn21+E*R1t7C!)t;e-+s$8U6^fT45?me97*5*lKZ$ zRzB(YW2BD__&&qJ7ZQ1r=!n`=x#Kcv(lwjs-*hKFhP^^K0UT8?6=+g;bKWO z`zMRF_CPuWxzwk-h=1R?l-7l;i-FG-{-yAaBiiM7da3USzxgz(0nfXO^kXy%abLDHj)rF zlJGH!xaSs?$nqSql$EZIMR1Nql)@heJ?!(SLYYvJIbO|OY{g;gB_8~NQw;H(GX zT-WZ|enH4Tt$&x5B`CWy{Ya8c-E}M@$f@%Z2N%wuR*-JmbF^{D@0|3hBSKe! zgYEu*r9I=ykU+@?f%?`;N<9b4+1Xo7Y?i3;qhY|t>33xR0N*36Z9%E%@QBHlhot2U{Ty#YQ7u2ji9)@k~5<;}PAA;)Y9k+cOm^FkrygcGKDe$%M`_r}I~I0r?8Nd(&QGp; z*4~wW9CC;+A`?aq(qh>JqSHPH2E*7oTq1F(xTNq8|pUir|J{kLmQ4HlEOJ0 z5<3{=Zim(R$fWqoVXbE$jy0~wt`7sXwb&4#g$L+ zf^q=%9ji{l(lLP>90QJhd-ScT;rW%W(lMNs&s5O#AA`C(TuF1P-D)n%petB9gd-%8 zxQ?fT?N&7_e}nffvMkrg6cMu8(P9MSzV{3Ky*=y9WrY>@7CaiOGblSj{HvxodurP= z=JGu*(^U9dr(1b3=$997lHP61sF^?RsbB}-O=wB+UKW8vMb1<_QQJt2<6b`tvw>tBfEJ7fNkIwRcz#SbXBYYzi_Hq$&0aSY8g*E(rZ^Cy(aO%31ou?g0#z()cMc@w;NozE=agZaH z77x1)tT+q^KBKRG{cF^v2MAc=oEy;^dM=%(*he(i4Y(vyG@Rq*3Vkz*%J|c-#Sg}B z6-8-8-QTJdI?WvyxUR(fm^+aldu z1{Uy_2jzDR&6!wr^4BbRBCz~FrT9ll@o=@(B|44Q*~7&mAy`8rx|rmOZe2-b!vGj{ z+C^<$`0r5g?Yy23@jjMb&80WGbc*6PQM4xY+!#o>z&@VYuMF`NaLcLb`mUB8H@wOe z;kNLyE)QPDxMz=)(~Zwsju|b_tu+seUk~K9j>}N^bE0VyCO}OKON&j4VsVKiTXsHl zyaDECIO7AF@Shp$QLfvFwCU_^-NUL$8zL@7G6@*(QEC1^@fMXCX(LOJqX4;d<0Ot# zuU>elwC^8W*|TShNY zBYaAax;I7HDsWVF+Ii{@MS8pZNQzcAuvEjK3O$5wo)G<6;>D;~Bl^uA9^vxw4W(5=fH~&_gU3@*-DTuNFe8rT9-~m9nQ1jn{9JdzLF@J5=ojw^PH(fkSm|v-~-7d z5_{J@e{l@fa={z1c@g6qa0m{iaxx!C?!@}$xJcnb!KRs22;G(B@J3GFyc+2~ zBx%TF@F$3^2wQuf25H)ICzxbA^n1(eNe1EpatD?P$E9)eD1z60+LYh4p&d!;I&=s1 zG)_wB9l9BQ8y5*}HIf4&!z|mk9#EvLqaMc~06DIf8(k12^Iq?YWg;l!XDa2+a!UDZ zHph%GE~k-<^ImpD?!Rkotx1KHV<&;*o-Wu!Z0zN@ zjy-*&nIR4(ZzL*hXvimP%F3+S9TYYVN22(mQ+9&?08Te3EhM!;!R89 z2Cn`p)MFOL=xt&$QJydrm}e)hGNXWc3|Bv<+g|9m>~x(%A{iPJHM6pV<_s`WS3UNV z!Ov`ST<`X++>*7*f~<^LQ}{vP6O4O)KJ|X;^HH~UxxSg^lZIz1smKG6r_-lT&b3sz zVxlQcJy_e_PgaqNyz(qy>;rJ>J09fvoRWR(kMSMV^Xv9kSF=b^D2?|?xP0iq93FmR zFiv>D6{o1`Uun9Q?^2me&gMrfLgcqO;~4xaj@R|A6k9EXmryOk&l19hRP!I_0PS3o ze(~UQ*NW=HQ+LrE?X#qU;l9hcTlZnHv|~BOM;XsS>Ds>3@D8(U2ZBBt>RNJ2f;c=w zr`!qKb3E4ijCVklOB-2jWd#^WNCSbK=dVohj->r->purcV15tj zQP{MSOJm~m9PJ?56kq8Zh(J|N()LO+OJwAOj0*i9!t7VF&v5PaZGG`;^*%eqOh2{B zaS>e={^ptYJ&n>ThE@enPr|;|{gf^~)#D4z8X{C`T1Dhp9YZSReR%{NSLf!Da$ek= zBwrjnKWKPMsan*I-@J_T{{W8kukB6c+aPH$@d5M~EirBt;BzSPJ|gi~b8G@Rrgwwby@T ztx_xKfJ=mn##c=m%Dgy6@=jDXKt08Ok=$99(8K1KiVfLDa!YQKeb4ouhqEVFFQHZM zC`JA0AOF|f)7ol)`$GHz_VbJp*mG+i7AA4yB zf)C>7#b1irpTUn9*lJ%3bvdoIXuV^aD1s-pwg0!@0@Dl&E5`WEVP;i8(<_AM9o$o;0gJ>q*`6!>$;8lIPbWis4b z+*rC>%M_`|bY)V(f$yGc3rGE}ziO=#?I7^)#IF|Gpu$HKI@S8G-T7BO2<=(fEqq?pPG9+7*S|4XRZ$y-!T3TpvJAde^JQ`4&-3GdEHKv|KwsuU% z129DX6&S7m01o_C_+M?}&1TwNe$>6Iz|0Z+#zpz|{Hx9fARhk!TKHS`Oo9AA@Ymu^ z#lFEB=r+ufLUM=glU+%)9mhOi9&w(S0=(0}Wk2B}@eY1f3vUvE#}WC!ab7IxMe`-H zb~r1+PIhPN`}}_RL2VzKF0u&j-b}xqLjM3Nttagj@NVQ!r^l~EWtJCs5XCaH1M-jt z%#a;`&rWe)nyuuB%&3_NqnzV8;C(4HEfV8j@xO+l(+1nijeAJ4P=4!uvi2BuvBAJ4 zz$5}jO9gCT#Aroo#SAoCvHLmue|XQp`p3hs5qKj@)!{mt_=Y<{Zkw{_PqI~(FiMUP z43pCrMdhX4 z!Los_u39**BbcH>hjWt4w+I>WovO`%es?rZF;_dG6&po9r@8+C!AO29cvHZiwKs~s zF~m|<_-?u`k!d9EtqS;yRaLjsfk$wI>@Z6$?BrpB08asH@RRnw@eYOJ&)M(9o)ywH zYb|Q`Q)@_JSZ*VlN0w{oqDcJ~`AL6IRAB=t=hHJ|!eSXVD z(QJq+w>LU{g|(Cr!ginCw8X2p$|lxk!B@vu;NJ&o8V-y`zjC*_we*d7sNAmLCcB%N zSyhV_j1jgZ&Js0L906Y(_?!0G_*vo)iI3v{02OIEgIHPL>KB?OkzgXqUFsJOX1ivZ zWLFItCKlkvCXsMU5Hht-_$hyZE$-gOL+~!CJgUSWw%^*qxgNexo<}`#oK`e&wVYIK z&{(I>IWzQb%SE?WSpXS0&N}hwU4^xj7%WFl0sjC#_4y~M{{X>BynL3hOQZN7N3$0( zmsgth>hEq2ItZW#^8&iB2Y%7t@K7I$nk3ry!;J$>coF{qbS(6RV;7bMkn4d zFsE=e)_gtqx$$D_;3w@G{w(okwSRGAW#OGEeB0^6MJ?v7G)`4iqejf3GB&9yMtH?@ z*1~(KsV!~!o{k?cs-)9Re2?2Jtu5_5S)0PJD7cApDOkY9(0O7y8l(F~d`$RJWgH$E z*L+W-$1U#tkUj0Uobp8~Iy6AOe8}BHh8(Up0tm0=uH#Cd#NQqu{?-2gO^)YW^5ZuQ zW+9D8ovZF?WRY2P>N#~s z%g(b{$Wu$X{Wkv3_BZ-R?C);!TrQ(=3V3BwTX!KRmJ2n!Vyt-!BbNH-J?gK(PabJH zr^5Y4;_pg{Enx8-@<_@@`by7yt(h~~Lv15(J9*A)xcF0Mx=+F=beI$<7rJ5&yc>Nc z5J(=uv5&^SS^b|i>lyw9Ug;Jme68ZE>uZ9%?Pt_P%2kF#28ELZcgX4uejST#!fkbz z`Jbj?*YA?@I?sxKwMB=Ae``Gxz;-?ux0RCKH#0P$cH%3kM+*_-<^J`2vHt)CXZY!< zc;nzsh2dMBF5ceXQrB%RJm}?PBK@8TrN%t5!!%n@BLMo>BjZhG%fuhIjl9sGHEm#y z(oaD!vYIyU*`{7A%Rk_yS44a@@KV1dYm-8A#tKx|YdlaY0X<12tq>swZjjb0PRLk#xq*#19_ ze!S0;J^SJD-KX6~aCS6qdEn$Q&1m?7;s&*uRB))-Zv72#9vRkQ({85o%K~}_iQn-HNw$n-f8iX~?T@zTY~%`0 z)DkQ6Bjd;?#2P&E43^OBuumMRoOADAr2Z0h0n z9Be%6&kju~`CmnY2alPflHhan{{RXwa{cnZ^||5FmcQ3?%V}|rJN)LnUstqvWPw}&3}9q)oORE=dhVSA{{X_6wV6|k ze-K{J!9f;!_5vAy}a=1N~0T#Bn-pUjDmAq zoPV_E!w7+b*TRNA{{Seq9sNaje-nNW=sG8cEzFV`lHr{Jke%ZS8w8Gruf2SPJ{|ab zPVlwe-Qn|ZL{dnAF#)po$#c8wVzIe3C`uv5ShjLO;XvpB{7E(Ubog%#{J)DnZnA=lasIy(=sk1Cx(|%?`L&M^ z*e;I*R`J5{DN-5Lv%lmDFhR#m``6VU2L3DAX#NvACH0xS+cOe(g~m^*1ob_u;13CH zYrDHEhGyRuLm&1*#(jDYde_sQ1QypYI?EQ~oGSn_ILA(y_pU5%RZUOWwVRDFUf~rQz7!|{eXFEqA^IizH zP&9y;l$>=cGmlK-yKN%g*K40CK^Q-DaqEsN&Bo<4X;W=2OsHa%npBFT{ zLGcp|fl_OGR$wv6!9)K5px29mQ0Hm&JzL5y@jpj?BWp72pB#K`uZ`YTuXUyBap!t5aaYH3h6?F%-sGG+CT4`@1(ZiU`UQ*PZl zpY&{ZCcbvI(Uw1+DP0`kec(spG5J^djl>ugOC4I1`!bUIn@9SAk>mZo28D0tgkRqj z&bBzmNXX;3>6}%URshI=ouIcI9y)M%tQc)I2-D@c$j?GC$@+f3(z+`hGg_UN5h@m7 zLV~26o}Rp)Ok%ueNwd~fdmjG);r{@HdOo$GuZZn@VQRK7e`lt7vS}L8D|>z$?AGt( z+02Uam0-AbMk4{vMtWUK;J1aeJ1e{GZ^u?PvD#|$XcIH&`ud3VOL&a(*}in{B}m5d zr=pIgzC^t6Hm#~ds0r_85$X*HnFD~092|&9Bbdn~k=&DA){EiW{WikqQn|jG@@u)x z%ae{xc_0>FkHl7L1 zSYHSDJya}SG}7hcf%Xkte!pP<0QFZ3rdiwF?vu^gx3S0L-mVQfhbloU?@d!n`u_lc zWa`EAI>|ga;mcV#@jjgDG52kCFns71IP6KsHC(HqlWU?n$ow~;OND^;wP*0> z!PiVSc<)Lbat(&HkUR9S#dkVy!tVjXpXq-Pw54Kr15MW(uhUQC>s&h7L2Jq?{c%vv z>keEVM?dHP0Iywd3m2}rJU_CR`*MU!Pi`;MzYT^7%3|gYLn{wx4 z;79{~@CZ8yIr&dU801yLgd+9*Kgfkwx<1p?{s-D=S623I;^-xo>rvEn$RlkNUX06m zq^Xkf01bf)BnX9=6M_fIO?;iJKA>c2nXz@f;0tD?5FH1L7HJ{D<&@_jd;0^^nnhGR zf-G-oB~4=b`%x|fO#XPYsG#i_@2Pr z!48e#Uko+6#;I&?tZ%M^62r_ylwTsB?`M}yN2RP%Nm^G4q#$47mR(f3d z=06$h_PRv3-|&KXO5zI{79n+QrQi85`HEc;ifTk|Dj499cJqJ%tPM-X{u;W}oBJog zeh`8;MPx~|+sRk~pW_;eZUl7sMmZeU?0@W~`yuOpwEqCY{{WAF4Xmu%<502Ed@-%) zM4u`-wOcEj`NSzAVU_M-FKsIWg07<+5nq)4EAXDX2iWynI|=nm0lB=JgnzW!5`JWI zF)I2TfcdK3u_)+738hDQ9z#bg2mQ6<-vjsol&*049r64*);jj-QA6=Y zNm1^k_~+dEnVRLush^ zkHY#UlF%!Nj)!F}jgq?MgFU-TcFiefCFJumc;RqKq47t<15DK8)in78*0UgNkl~fN zz#G&t3IWOK*l}F$w{3r>cxy_xy0MzkkuHgLQl?2KaCVKuB!yglHTEdhG(BF@-%PZ+ zv>@5Rb8!sO$m;`mG|f`t!pd~Gg-_oMB*Eh&a)1v{ z$FF+yDWxr0g-LSkbm`)F^l2uukHWqZ)`YGIlf=3*>enfS+C*~;m{K^-0SE!-ob!R2 z{ymO%+b@8=8%Klx04)3>nfB-HhxM*rS)ye|h_>=Ra85@A;~(Qu=WtYzu_NWiK<}PC z{rVZ0%3=bug2U48Ta0M~q95h3#P{{RRt!;Rba z>>dxa-M1OsS)_IL>zwn9S1PJqNKjiG4y5~Z>rweqZYWiXqVbXUK<$%RwHY^bzT*X6 zPjk7l@y+eU%g1Hl4~LR$LgyY3)0ikx^Dzf@037EPJ+H*6VvXki0E9!~^}9p_h=;+N zS$QzqSVqPcRVqGzpBY{hlTY}428A}Y{h2nE{k5p-LQ89FDdZ5LEfHY;V@VPtuF}fR z=NxTP_^xe}U-7;6blWwR?JWRU;Y5l2n0Ve$kX_GLBbFyPsi!!pS}Mi4O2=5AA3RxP zkv^5-?+IE;H*fO%FQRRHpOl3*e=Zk;lG!|hI#u0A_KEnHG-~2M4S}&7o({MY2zewT=`JA-q{V*bQnJW z0O4A+o~)NvOGbNz-^cHbx_yPM{mbe0_WF!ob)>_?o*z|_1_3d%>1+hZ48iZ?%0MhCgwi2-9o>FuN zwg*99J3}~?N}b{eQ{*;d&J)`Zgb*Qy_#!K)%{{Zl{=#ttM$+0Cfx+JuW8ovpHOnt49<5qslq<~a(g3o%s$kMxfNI%=C$9hIH$+Q<2? zn$fB4`Jk@TR!hClvA^J9aIKee>Pa; z@>KC2^hVo_%BLWHTYtf4zieOGZ&2~ChqO-!Ur*xi82CCH>npU=8*Z5l*WY9_!8PzD zWiZ@YI?659Og;wThOg%`>-LEFpFPZzMI=!xv0#d_5TKlZ36PQpPI<5MGmf%~c+A?p zdRB36>8`g)zsBeKdBD|b(aEsXaWQIYHfgIhtGD4}|Iq$>zu>O_00g`-`#XQZKzvW| z_u=$5R{kGNCquf^zq9Vr^4e&pv(K3nzGJ@EWNa*f7-N-HDo^7Vhzd5hbnE*g28yk!^gkyP#+Qat4uabljzzQcGISg<|1RZc+Orh7jOUu1317H_|4*&Wd7W{ z+B3FNMhCFx{A=hk=``D9W(ip|taEpH0hPx%CjeKv_*i&HTD7>K9jXPb4RKT^MEO&-F>_FD0_v1croHpcTw(W9NAFqtCz zFqHo7i)}k(;BF)ka%<*~4(n-Wt!j6#G-|WzFrGeVgE0HPHV6 zV<=&SDod;Q$>e56&SJ8(ipWL=IYP+J4mSgyalribpTk=8_LsJrc=Ke`H4!WrE1jm- zBhv+W`rw{8ubyor=VR|_%0K(g`B=k^DdsmKFU4L!pmb%h1I$gw)4UA9TIQkxP zex8-={{XWcjm?k8{U5|d{>&^i{Z~;BfX8Gyb;>J%0U)=U$^QTX&$zi?*}I(fw7FTr z{@0rAhm5`<{4AOqi6ok0_q)Q4qwN>TB9MBp20k)5!RStF_S3_@J@6Kt;Xf4ki{jsh zKjAU(zLRQ_K9PjqG+t5?6iDczNM=zg<-)LzFX6X_{w(;aRQ$|`0C2hFZG^?9^$iwts>ywEb9#};LX}dep|=Kd z%y52FoDSd$$IyHculTdXQt93uwl_1g%`3AUENach_5k2>&;n0jabI@-0KrCn5d1Fq zE&F8XUj}{{eWJ@s(k*mri~lS1q{YszL%Hga;oc3D3me8a^!e+rcP2 zG2qQj^s8?*=9Fpst=eXlo6c!WTU!oJalqvKpx0gwwN*uP+nL|ZbhgmHW&OQfAaxk-GC8Z|wJOH*k>7-?hr>QvYS%w${9pT7{78cO?(4_j z6E%MhY5GO=)sOaEdW84H%yJ`XlHA7;$v~@dI*thn>z5V#eftA^aQ%h9Z9jp{@K5%k z(DZAI%c&xS{a$vNSK0fNNWBf{Cne_d*VK?KBcbBB)8X+ zJ;b*XIVMvf-o?oqvRjf#JC9tP8ulLz{A2Mq!EFynx9~;eVt3g)31eA7X zdj2)%<{3kAQKanA=-}~il%?&KhxK87{{RFa_;aoNVbm{v6nt2R!nay3tlEXPiQ+&n zE+C5MbW&PLlWlnHrSlZ-&y-=m%win=a(>#su$RVf**{;K#U48rmtGsaTb&leR<)US zOQ>4e-OkYni7y#woHB+aFwQfIt?^_23hU#~+V{hr8rOVX;oUPp@i&CyTX^*iC`crj z&Z)jh4%ESr#tWca21Q_N(!MSJ(H<@RsXhuXfgxAblS`XQzPM0X-ZA#XfX^y7c*Kpl z*CZjq$i;Ke;vB+?`kddm-nZY@;>Pf%2CROy6=@`W=kZ75KBM7phj)Gt@bh_h*IM=E zshvUH1TblEqpFdRCEN<+jui3pV!S{05b(~a;r{^HyFl=EryeeBE;MU-kxv2AOM@vK zXCMx9&&oIzgYk<}w6eRiMwj>6rmy17MN|)*rJT0b%tsg@O8{~B^P2bX*?uc%be$el z8)DQoUCo~7>`dc6pEGCbYsAc;p@sKW@;n?r@wM!H9pOE0PY{0CejwLEEM+e>Q~?QL z{(~~x2FV>;W>59&NBk82);Rn#;dmm+5o*^r1E3jrBv5|^kPpVG`~$H%kL@4gCAi$P ztez#D=dM>(0hjOx6`}tC1wV57H-)5(U`S>@yoE^5^y$+z!^2L%f>(pMm68*roCpW6+&_<6>%JaceYaJ&gYJfrI31fKjyUc`eGVRn*2jtVIQI0uT++cL|$&Gqg{1G$_;<%p9Vu-fF0Ub0>7EGN$^7fkydQln_rX0%-c7_%6a^i-3D(Z~pYZ&|ohW&(}Cf-^`EZ7e956npBG%50sH2u}o&2TV5HB{v%%T(A*FPu&o(?Dlmy90j-urZTr3Q zO*`?VcQ49fTA+=PV>`#N+8q|8cAMoUfY72>DgWgyrLV5%^7j40v;}O$E!^+11b|mS z?{v|gu}9Ej#%mg7C^4W4OMx-8uw#GgqoEg5nMqzUWvSlGYc`NCF$4XHqqRftuH>G^ zryI17yG(oi?Ak%Vq}pzKWDV}e%n9=RM;8Hy=kY@Q9-f|Nw>eQ!4i6ddvaI{ zzGc`S+}km#sGzz~f4`I)Hu^~(#aGne>{g#E8!y-|n=lJ=ocL{15JzE<>V$~E{lR>b zMbkg72I^B~xvfn^Wnv1VMP?lXVaH5MwOdJ@;gtMdy-1$Yi^b0qTTbt6Fq@Y>I&V?Y2?^wLj!xAS>=r8=!5DrzRdtN^a z#H*$tPG8dPH~Yl{x`_2%s2oGG3-0JNTpM&-ZZ+4wI)6OlHG>DJQmk}B zBO1b1rFM+BevuusL)Un7pm1rf`CDQ0D)Jz-@U8_A$K~IOG6oUdUt){MUgwj6e!fiD zHBNVusY!wVMM3OkcyB4t0@RzGKdsI^JT0z-!G$=1^_Kf@)+CcSF0vW0CTFj?Z{!iK z4BOk~MS@vKgjuv&I$uan@fncezoa8V;c#;tWQfI4lx0)Tw9u=|YLR;n-V{_yo{+4D zd^{O$Zd4Nzgc{%cz8D(IX+c9fHNb-`gl@9IJ9bez<%yVwz{Dr71BxJ962%t34MM0a z@RqQtE!(8j5%~I&fuwrd{g|*YHwvdLzA=q_Acl*#@T}cYcef(D7uCm!8EmqLv|5&}Ot_n$L7o>ufu5Vwzb=|jqS!7u zg>3EPuf8n27Ha8IPIeWVt0^+)wb}B%$hY!M&!YE8puQ@1A-{8B%I$CFr)eV}pz>sw zy{n+&3Z!?8x$V=qEM-=2pxVlWNan{Jl0OH)!m>0OL53vL`+JgK&xTzY-+mtW`(aC@ ze_WT7)1$o%saYAj`I<^91Qu=-I=(DeJE$w0^!kS!BQYa;(wsWgDnPJ1^E{c1h^|0oOdqrwuA2v?qH4r8z10U54K zJF3j0+fNVMj)IGy`_p9mKO!O(-M$VVtp06mC(d?bAPR~VPUXzF_Tg?lDdU+UfXc^Nst_5y|;4*w=h!~RjOt)DQymvO#7tgDqmSH>R%ln&`$37^rLbX`f?^M;*mu+GE=DSK?rnpv{7I9Wo?N`T3{^-~;8 zX+52N*8hmuqhc2~jUrFsl9=$=9dyr4(NK<_fIF^krrQyg7y|m*n|IwXq zGE!lYQV`nHYy^MSIHB3Fr~&YmeyTKkL`E2d)JMCN zSSL-->oz_h8_=YZrlsH%-V-eA*gNbU?vwVu5bg2y$}UzxF7~Z+$aqSzPrZ!(h!zgR z(?_}%nRVrIN4+7&N<2*u0qW?n15{3KaSe59fQAnSNo@7Y>OaNM#! z%}~s04H0jq>RFLKbH;bMWJ^X?)^(U>$j`LiXjf_A-xcS2+TXlJWjb;jzo_eaTa*C?LZZDeRK_sf8mbrQG`Gy+{Mlbnhiz(^KK9DQ zX#fq==U8S!g3rw1aMy=4qHk-Udusyv=uLeL2rh0N4Cy8||E+ z@}D*^Z{@48Y#3VjLK$nZTFUq+kd0dfT>~XMc(O6Su+Ck!{%XIUd%7tM|M2^b>;irq zP{9BR60>%)pOF@&#bztceSOtDE|-c@dV0v9zwA>5J5o^ur=&BjEd>cgiB=gSFB#Jk z@+Nj!=Be+q_q;MMs>( zE^GA@6n&&d>gMTltbQRs&>9eD3dYoW(3-x|o5-&r6`8nKP;02^`$I@eo{AIJIo|Le z?d*23&T>y=r04_pkbAgy<3A%b1hmnyxL`R7ui(J^0k2G1RRhNJGD^n9_kU>Lh(F!% zIv)Xkmv1kU^8mGfIob#^Je)4Vz^f{lEvgzU<5%UP5p60{2{J=MB7C9m+O^|d5AnD=l z2Nrh)rFlndEx|{3DXH8%;v&;$W(ip3+5_+!V1c7MjesIL!0BdEKou&hTjKoBOq#6^ zBi0b_5Xc9(XAO}$9twZKKLkzW}Dv7IuqXFF#wh0xd*o7Ifg}KnYW6SyED1gXB+Aih3iJ_KlvN;#@A} zwH*z^e#JVsUWH7G5+BM#u(ku z5?0br-eT!M(B#EiLtmV#I?`KJkF7<2j0QgEDIg`f)9qE;iq<=3HCNz7uHN(;E1y0{ zT)W>oNOYe`cSx<2SASFf{@Lg2ZEzv!e00ZZ)K`p2rXAqDeynE%lZ{B_*RILxFKK>%Wzap4dPfp7( zW@Fwin%wgg1zz8eKaz2FVocRdC?>T-u;dSC%a&m|g2kfNsw*d~WnCv7teI|lK8W86 z1%Kg2SmVb4)!R{L?vI>WAGNSAwxnbYvZjS(d9mwm8_BD_9dG!uoA1WtZ1dFg&hH5C z7d)FGM5a!7N_0dD==3%rm}i-+j8ocGCHutA^mOphh8R90!Wev9`om!9GfBQtYk zaThT2K7@O1z$-2GF17r9UOb{nSYjk1BJ%jc5DxHBdQ48l8!VDzL*8OX_KzqNdn?#G zUVHxNBO*rNm8UB3tum2yjiwi@qkh-T)`&@WZuagwsYuy#Eb$PxjLt1JS``?m_iqeG zxCMlwb|sTi4x1fiE+hGyUSBSgk8wZo;~Es|6~%FlN_%8{%NxB*ZW&h(Zzm*AzOUFf z_>hxq&v}pii)A8gq_qf&6+cA*!kVNoQmZrFD`ZY$iTTme97PITnYj+re#wVxvhvW*p%`VA```xgJR{K#mG*`8?tshM!g9O04sx|1A-yb(wMpIZ0Qd9#0&4u zp%fFt)gwA_!Szh>bKrP0Py=%r#k|+65y9fZTS940@8!+N_}GjUkA3YnkR4aQu?iJe z$xm2gv{q=&F)Na_Qk5nmlGtW!BhYR&q}oHCm$!zQCUFFsIMS{;K1-0edu1Q^>fs!J zI2cq!**-Zl`RPA@t|5@;SoTG07juv-%WZ-zUDv^X`^hK!k%=cP)pe;AD|f!M!NYKX zQWKeo46bTk`kwUKd#3|Tv~;w6q_~uF_r@bYd^j#NAuOCiH>o8 zu6Ii0G2gH$Q8lxVM#^r)e2^uX~RC^B;c*jcPy1o%jUaZKwIY1u4AC!s~>O zZn$A+B>StTx(b*a2qmlcy5hcGY^62iE?FnEQ}isQ?fXI_lm+w^rJeG`oDxIKlbs~M z@A}E`1pF=avq51S{Cg=!13IqWxv*lzWqUAb6qaO2?SHQHaH05EYp&yB!#RXkmG5uz z{INz&mC=jOoN>;Bv-BP7oFx$=$W{(GrPpm2yg9C95Aw-OjYUQ^$lp0|jerr<#s)qt z8_*q{{JzzS7gWl9=jk1JR#03n<4iQ(|sntI@jV<&c9)kxJY;r*LA zWFgT(SBgjTqYCK+t$_0t?6V`Nm+OrV?tKInoC|&MJF^273d{*NC;QK*3MYm8~xm4H~kX+&1X_W(LV9N7*)7cKW!mZ1>4A|%0_*U6rOdv z6eISXDh@9+y#9-E+jK_2huU!cQDDHyvyFd59^LO6ETOuvcgqz&T~f4%kf*j@yNUZ@ zS2H;?oK{E!QhmBo@iHiSM|V_O)VT~~J|TyZL3F*B{Iwaye@->SQkfh4E3B6^=Jpx( z72z*zXiWC$&h+)>d1t5VXR#7EBAvj1xcfzTViR{O4MV|0vs1d}l_R{PHYWM`Ip0;B zwkTQmD1mqg`38w?@Ev4Z$i2xRz$%DAH}ri1fmhR7NJqVqm5!8&P*?De5 z?MxEA>^(SbO&qjQ5mH`pouhdA8`U-Fyubx`2BAEjHSNav>e-yM4vaH_Y$Khm`Lz1_ ztA%OE;7XNPC!Ap~EhQFb0-v*(br{VnC84%h7^pP{@gP0ils4;M<|IA*{Midh+%{PE z{Ykc;-m&a9po73B$vOdY;#mlj6y_GuPISWY8a?d0Nb#e?x7`Sn;TL=J@6d)yB@4JS zUmXsmM=rem;){|g1YqSD5yS69fSd}0z&e1v`VzD77{r% zP~?TZ=oH;dYUq4*)Bn2XY8Ifemo0vF4j%=F8Rr$RY0bgprj3bh3zl~0(V*X?13%z% z_X913Fy_ghm|oMrxULDYmlp&@n;=@@P1wr|)6Cv30CQQBM(&VX>9Omvo4(#ZB0lTp zMtwU1eU1Ss1AaJ6IG5jdFJ(Dl9W5*?h^)wuNrB*kIF{^xM1P~!r&BJTjM8=l-pUS> zYdbRgBvtch9u{*%`?4m0dhT$)XePbLDTj8guu0SBd=43NEg`24Ihc0&bG=SFf%uew zAon)j`Hf@RvGd#ty#mDOgW`4N1uwM#8ICp6BtHx|(!Fr|q;je~v37|YeO3^{rODY6 zNxU*%3vrR_W5fI-nubs1)}Kt9w%Fgpv?w9Qy$3`6)uTyUho8K|d31XNNkut-TTJ|2 zlrJ?>mf0IAISKQ`+;G$wz$*xv#p7?fd{C(jJ0{ zALPs6Gxm4N_pt9lN$WT3djgcpMn=v$Tvd@Y*n};MpBqrgBcr?Vy<9J-_N&AC%Cs&Z zh>nG*i-l8O^18x-$t%CHa}D$ak$hsE>dCnJlO8^TPc`c%)Z9S zec_2cwDa2GOC_;CsJnswD$7!Mw|^VqFI|~oXf(e~Z_<$+Bg=r%3b&f@mMc*3+NP?G zn!FQ7ZnMp_r-rBB0}}JQabn#u9LnFHMZ|w9fSmDuDmE6zs|!B=o6;efgzPXgFq-LY30SMA~tb-vz8C=Kj-!qX_hplq7OScS}D62)M1prc*2qL(9ef7 ze>4t1g-nxE^isE;Ld^w^q-*J7gT=a}mcWOZM2>L&8Jeqfw)QA{vjr{lUSF@#FpCf) zvyf3{zwgyxS}a_%2~k@~(@(rxE3{zRFD$R-G>E}kBT`Mscu(Epy%V$h}gMyl5zv^pe!j^fQsH2LQhJeY>Zy@m62p9Uq!C zFe2s(b|*+e^1K!^G?oZMHH_xC1Pi-S5&XXkf1ydB_9HKgK>XNfug{Ix-h^6KYn8Ne z)>7{eP+&EIYZm8Edk1^k9rVcIP$*lVwsq!g*P;vm29+WEwA8Papa}Ns+_tZ7aeYMX zQF#A`VSZw9JwLB~=Af6V>5w@83(@HNsfEF<)=ETdmwtz4wz*?D#)T7iJdH=<3siQ-3apy_i7bNy@uFH0WEghRmsxpJ&X78h=DlrkgZK2)u4%D3`naK5y`0_1ig(X#h^pf{6*P~k(u5&|t}6#PExW7^wR1vyq`R_H5GZ1hbA zYYstNn@7RMEQ23)mgYx%Kb#?lLi1>+gdK{rksIiw;|R6a5f487Oi?|I^8NwWW+4id z3X7>?joaYx?w2UjLJ344Zcgx(h;G`IPkxFfV3^1W*58lFOFyqenDN0;Dd)sdUPYY* zh|^&~xZz#?qp`x~0E7^!K0jJg~(;l9EN-;zwOom&;*8jrg)P59sNv zfyh`jQ^@qY$|u|&y)4ZErxx-dLjdB$i$v_8En0S>Tc4e7+6Vn?jC<;EGp*y|)~|Ke zKJ7l0J?BpA&grkb*qr;^AML(_kx%k3QBb*++WgV&@iLP|=j)*20<7jI=i-IF6lR~1 z&z4PJztto{!iI!%JSnMoN6+BEH{p)j>7P$Fe`^jG=Ul&SDBHGUS5JNA64Ge_ZYL$E z<~jN&{TF6{_7rQ*i0@qojnMK$4TZBldW`a+;LV+oDPR>aG7owYjpo{ptakkn{MDz9 z{m~&`UW3+4?gc)xuJYmnZboOkQ*)&wO3G5FE(`KQFa_ zHuxP3X5mD!A2DJ02Re@a;%Gv1NdbbIVgtFlVotf@{yyAm&~1F;$T{x56=66x>Q>YE zEd;ZkRJ>1xeF5ve0z10Trl9YxAMus0F)|>>}!skpb=+SeAB_X_f znXIEeWCmZB+0jdX_n5!kyHBk6xhC3`2k^7!fbq=$Y(c_}ZaD1>R}r3|u=Ds8@CU2G zsq@V<-1~~#973)jY>v3c{<0ivK^TY7HdS^Wna(<`f8aYm=~%s{u_y-;{F(Fj7cR_h zA+pbK){Dgja=GgH4w$!KaIsU^fj%_^9N;KtJ@(Nknuu1WOa>LCHvE~1g0a z!>b0o{_yFNwkZGp2J&M7=&+ zC6E{fH^sA`-LOfnD37DYO4S(;v*MP8R!&Da4QFJ8d<*(ZrF0ONwG!vjA@ukby)h#` zDoFZoJJFF0xN?JHnuz)1umn$K?o3_@@0)efB#)<7_X84V;4c~i>&fF$4fZnPC4s4_ zRu?k8yT%=2fb=_lN{y4A9E?iC9C5)MQG~e0?U<;YBs6t%Ex`eQR(vCVVd*<^)BO&XR2~RII za`}_kYcIa}>Qdo8h|CTUEnN1HF*3n{^f4D+V{EfQ96tp(u#)J=b}9fUUk(!)b!NCt z+|KNiT?s39q@j;>X)z9< z#V;oO6^plJt?X6JOOWCS%gY$UsRUVJrPc+$U0>u`Xf=b$hSyvJ6~=s~5%rN;lX2ez zMJSUm83cQ`h95qqE$mD-V!Z!wNLsG!@MpY~!1C;WKk|V~gl1;GH5ZKKNr8ks#x7hC z^bHEHk{otUY~cEmL8mrIe;%|tsiz-iGHpF60Dr1I{u($(*Y4-Ydl1`=G)Qphz%#X@ z7re*KjrUWSZlsK%2t~TuA1eX019eYXPYl3WBfM2q(VAlM+iP!dl?}2^LN66$hI*~g zo+?B)U>yyIaqYLKvw+gTUL0?#h=i}>rR0aLCv>l?OxzkEaZ22YoAYFxywA4NV^9dH zNPBoLT(Hs=(}_yY?N3e+K*v_1QeL@sS~r5!}z=Ds_z${Ie!n|=lhcY_SN&O z5x?n;&y>`~3U@SYe@6_OQFUDSF@GDE3^^jd`(HWf7U#=8aq?c`r)b`a!1ACT)tW=P zSrKIDWj0DaNS}E_&~%V^Svjw#@rb=mtz6xs-sav%_01XgM?HS528kM;4Z36^(W+9O z!?ku|Z#X|y?anN>)HGm)Fl@?wh<;u(L-ZB@r}TAP91KJ ze#@KxZr4DtdHePj8Z@#ySV|M?k&_!9ejHiNvQx@u-K8PA*~mnBhbvQGdqgV1q)>DN zg<&1{aM^*&^}Nb%JEP$FM`SX!Y64>Z#8F_WdKE3vqq+%2$KjLqX8_dv%H3KdMFUWK z)M21}#PHKECW@?2t}}}3AB7%_3stF|2urZUQWX#zU8HpJNHzz>qqX8rZvRt zW;Q?kn-+c68ZEkS_Uar-ngdDbU!&0RJw*Mfm@Knub5jZ$ZUXzUBTieRy;Dn5HS(27 zsJ(i(dpVX{@(F=O8pV4?D^5;3AI_0zp}rPuJTa6NRdCESk+;4(&}x#I65eXyZ)W`8$8@Jj-ScgN!kKpN zs2TL5kguyGz9*}qIUQFd#_H3*O&qGl8VS3DU5a)wf}Dndo&WCx8j~f*0>rw8)m6~; zU%d+vt&c&;DDTp@RU5Z*Vf**!+Z4$V+5Fg#V_0-vefuPrH}B`(=Z%8 z;fjZiyx^|EMkBVsRM>3IG}UB@Hx(N8rBS<7Hm4AZXRgDDEKow2l$Iw&D0N~czyL+O*&OSoh%&TQsK98Kv{NL=YO z^%3oC`W6S@5eFD!o{2XHGF!eBd(re=mU}AsTB7vZ^EU$${GN|21>1;xWtaB%k7qkg z!t|!L@S_wpQn621LW@+vyxNR5Bct7u5U(#?Gd%&J!c1wzHCEA8+$7W`o~mq=6oFa? zE56de^+H5r#4ke=EteN*kFDN?C$CFyCnY++WW8j{)iV^&jQcdO0`mz`Vf9 z>KR>7nC^@yt@O>{g31w_Cw**pmrx6psZ@f$%&qhF_uJn-jM9RrV4!|$s#R#ugyCdf ze4Tbh`*XMLJd*V7S%X0j$(fYC69>_eRO}~AWS0`2ifvz5@Fo6jjdpL@V65(4Wi3VC zg|E}i_V@>>>nIfXLC6%C-I~079&4hVKtJ-?%6=g(V@}z8MGgCXy{w)K+D=8gILy)O z5x7>YTMUJHgVJk+$l8e~nBaN;h(49lp6U=zKoIT|dIa_?X1lv)WfmhDJN&y@dC0rGy1PU=K@@?eZlXY(sNGSTYaza{BROFlWag#ng3p3_pOVVfg5_j6= zw$b`iZC>=o%RA*%>lW}@+Mr&~)~qLoWS0yFGR;NnwiglLmUg4}Q}{rg$I8g2F~ zA7I=uHP)mn#r(~USwt#V_J^CAu7fW7mHTuV9=Eu7eRw48_PHy8@iXc6b~nN8D~4)GT)PQ`QyNN| zIne=l<{#v#gt1n~S?LlH*_8L`h9C%@jyCsshPGCc^FSkMrlQ528Clv02_V9KqqYVS zs$vlM{*$A=Z?rG2tGnXsm1?PA>Nkw`H&btu`AZzqT43-M>T1%~p7$sz@gI@E5T>&_RSW~&q@)R_K~ zb>a*e1-mQlkC{ILP=nyznLvzh*qWJQ^-CWP&H#gmwNy8ox7R*ho12;ADx=`PMCHl| zrnlvTV0FU#aBq1`n4Ho10Hc=gj*655y5IC$N!$8SG&)WU`iQ*x%t!9TXfEMPlJGhbqVJGUr$E?4p)ki zYg7yN@6d-^>uiXTFfwg_{pixA9x3V+l8y7>kd56Z0ZD|?T<k(NfSM^Tg(h+D;=LV9zELTPRDh^Y-;C#1bD#hhi z%nf(*W@claI}6ZZ1IFcjIKOVZ4&Tn07?=1Kg6>|zk^bWQEI_Ig|UJShuS$tBkb}{CvvE#qeBL@h_ zKYhQcE3qA_#ASjv%FRo!;8y{7eEa4Nc*}dhD{u$o*Cn|Xy>=Y+CYn#=YIwN0Dce(O zh+}rVaPM*1N0t3k_R!|mHDEi{-C3N8nU`KU;%6@bB?h5g9ge=61Aj%z2lz>(&A>SF z=AP&Bb7sw|wZLF<i1fr;xCz*p2gX*^(x|=*PwGqEjpaR#DSZNH6eB45zXXPn%UlJ z;;}ho{6a&$5cB}yp4#6@QJt0=^A@BvMs^qz=l>Q_t0c3r`Ss(LUVVKOu-|B;9L_9H1d>u5*1(rvODD=T3UeQRxRMgYdH7Gw{U2D|Ui728<%V1UtP$EB1v}`|(`wYJMdCDNZ7Jje>vQV4|nh-JcDyJd??-ZuF_nA!Lft zq~`t0DzZh1#nN_ocWVrcta`zM&L4*Uv2f4tiq;m?W4ir!gQAmo3jh_pnxR1J8$o|o zu#f9%v?Sw1amWQPbKc?dpX@+ZxTP|wJK&>@l+)*R>`__Kh zF3J1q_>nJn3+j+*eO+M)O-djXIcJT3FqMlRgOMOwqiRo^m`iI)rj7447lfA{aUo*O zI_UtvhCO6mXGmP1Qj0NPD_nToM(d$c4jHm>!lO<9i2_;77QjyX4VkGFOX-;uT)wkUKi(Cudoj`*H1N z_KEDP{}-GfC$!d~GN!Yum8D5abU3#wZ5zgyoNIp}m8_n{s?x^9vNgxt6m{nn>{sDh zi^=F!OD)o@;)brxLKA!IT!~#k^;=_oNMjMR#u^4)#^3h=&W+KrbCs#{EDkx1NpnZ^ zUuY}Z?zLH392;<>3fj|?;}pE$}bCCO))SSegN-h*toRL(dgNFkU- zgFsaOgL=UiWIS4$LV~+y^*`MsQ?gFHE%+3z=bHdRk0Rdri%VQHAB9NI-BzvQ#+8Up zxFa;Byb>o`l($V&!rxK;2bUnp#LeX3K2(_y-c3N4%_*xK8r<|?;ri4fQ;sVt>{K>J zs7WhHmZnG*1qJNts8;${X8=AjFIG~2VN_NO3k!`DHMmB%%xJb zxKZ0NcjGhJ>8lz)uoF0;iZ5E_N~}~@RjrPabE=|s?GhD|Sss~z?h!($DKQFy;CMQ>4u3oPSSMnHd{cCyy82MS?CmW2t%pDXs z*k9Y1|D9SU@&GOI)MRfo+5L~Q1UyWHLg-1Y(XUTYBZ7q4lU9p>VXr_t^vN@(o=h&{ zWvO3`U>@AuM6O}hBZv=Rn;sXtp37MUnyf)XSLbi34lM2#2^=V^2B!{Ubijqh%wI=@ zM+m;T=Li@1_f)^+BaSqI&rHu!1{vvSl19Y$z^v06Tb18aIW51$`LTcT>EDgPnGI%L z<5G8-Bj7Hqio|Ec1b=ZFe|vz4+3l#2v^kE$@tNk@@B=S%`hH!Z6mKZ z!T7iwN<^PYHN$1?C}Q!+$|s|^UBd$3sKb|$tW~habQ=Tdy#?I@U9I=AH>E1pPUYGj znSZ89@XwM;r(Y6=mO_tDk##bdu|E(S`$&QU3Bc}~bdVqm+1#hZc<6HOFELzFRcVruV zHrC7`dGRkK6DN{)fi3@HEhLq9Xo|+N?0=e+nJh3S ztzZ7UQivCEU4GlgWZf=J%VV3ZUZ2ma|6yo+3vqK-7OyFXqZG&SbS}xU(B`mb-18zt zo>V8QF0eqdlVs%BdgMxt-s|>^L~@Ya^%OowIt@3Q4J|U|q*zS{25yXSq>1%bmdnWS zm=wsHAsLCj9La`x;EK3umgZ_neMgtCE^{X<_!bQo+MR#t%D-EP*rcjmR?uAkCIO+) z3fN=mK$*#{TH1I_%oEzHZ_JYcq&*f7Ef#NkX1x0-gZay{Yuwvj_sULn?}-Ur?WAz~ z4Tj6y7tUPN4>E2WJ#XZ5{iRHq;pZ4MGTZi~s#jJM1lU8$G)#HsXTObzWeqZChAN4f zv)7Wt<_yG`0}}lI5y_{ieUU z4=udz)UWp;NjCJW3UOVkJ$%{QVM3Xv)MPl>Bx$h+Jg(4V@(osYK|HAR1{P^SUD{8o z<_BC3Uv~)(`2_7v>xC*}=u1};`7=Hzz{&HtIiHZM6R~G=23!=R2)xtNvg$6dZYu0Y z=GUH=QpqG;z7Md(n2uTSTZOAt3J#pSQ90xE(o%R%yU<5kum|K%*Y}Ja{#jaoTMur3 z*aRl+LSnfQt(_(d9VxO8rhPQ-EVGi<)YN$tINsbRy3gWV(D7MGk4buF2i^vlr&Gc< z7h5Wv^wct}#mbV&Z*>@wXu0ygZ7rr~nJyT$`a2ZnxyHX^SWcrZl zYx_3Xe8kNvIPNH6rjtMswydqL&6MYeVG9_Fw{AM42!0H8u8(3_50y1)gD=CdFX0Ea znD1+VsA}MX)CFI2q8t{r%3@!wRVDIy5bkg#B%8{iev`_n)E~k7!!m`BjxoK{gY?Q| zojY>Y1;}iLX1oe~X=PEJtApdwohk%-45kT`1kX9a!3`Z&^@U3I6Rouj9q=f^`Uh$M zI^YYEt+g8v$)M2P@+ebrG5^2v7{!Ef#lEHcS&p33Ui?ZMDWQyaf%M?j#MT;kht2I@ z+2=TBG>2>|${pgH0;Xva%rz6Fb69?wHoQ3_Lb|-z=;GWc$9xFVL4Fj~Lb>Buy_{G5 zl?mblI70Fp`p;)}OVZTUg?DV@%xPVQxqj ziaRv5Y7i3y>|zWtROQzE>ygk9G-w~_WvnyiE}Adc9W#{8^hyo~tpTrol@*^VYjs)) z2T<_m$mQ>b+{c*A3Dec77L}z2d(_XaMkGjZvb}i};1Nsk<`WEPq2Y!hvtUM7+^baz zc`mHks&nmk#q@@ffR^m%j4!4X1Ur=mknUFZg&+7UVI*B}Ygi48Ul-tBlYtl^A?(C# zFne!a_Uy$P*CkLN*(q9;w~EO$UD?;Z@JL274}~?C z_x?qmFXg;7PrXu#HMllIu03_S1J#1QZ-p>n2^}4w-ad%L(S{NFTish=x1MbF>Y>xg z6i%cD{oQUtWU%hwXwx}14~$5*OMA+wuz4@`hlEgE>&4}56wnSl3trXg9A@ra>tYwj z&aFv(n>=p@XuSPLWLY!n=8`e)@JHVt^}e3^uZDEl<-HXbM@Dvob07bk`$3dgzZ+is zD4?RY(@xUaym){~ApcN#DpM}tCAI_R2Wb+4aPS2c_t9c2Ad_H$?NOQZktytq7m8pUq7X_gal zGxbaTw?S4;`DbS&p79@%WGp?bD{U(IYt|s~%3d0TS*A#&O=&)MvXETAN0FTVvM=u) zv4~^JtFOo%i^Is9``F`)ySLegvUh_VC#w&A7k50cKS7}#BS%uCRTP{C?4FH-8;_j7 z(WspOe@geUc%~)`-g<@1Ss+q6;8?Deb}(fS3mVww>es!@fJTNdm`(7D#0qku@zC3E zm|q){WeE~?2p>ysu}6>N9xThTMBNfzVtAN)3C?p&_AxSl0qIliO%LjvZkN-Jtk0o#dq|2HFW_y!SX~2Hb|1z`5T534y0%x1qB4f^d2x>wsv) z1U#C2fq8FU`zZ3BVfnEzv{=&j-%JU-rkiE(|7y*&?)6fyXY%7CUqhtbu8dzmh!>3Y2(&#ykL+iJ5%VghT6W~ z1i}PVW*6vSo4Kx7UCZt=rp)`d%9whjqOTL+y-vR_`*5at}?F4w~eBtf`A~M z0)kS~4HJ=;?woXtMmi?lEg%ij(%p>iP8r?O8#!R?|L*;?kNe%*b3fO0opYUYocn;J zsQ3QB=Gr<)H$1hw!lm7EQZm23JwdtAfJ9OA17)O}rH!ol_7smS7}K?d;^4Z$e;Qw& zR0?cI_@(4DUBFNyK{cxJ%#o}7UtcV>N?x8DY7|iKe0ubVgJeTop)KG6^EZyl+m77{ z)rjT;XfHRnGdi!ASvUDu2bT`ce?MPR}n zl~aZ`sdwW)K3j(#={|hdrAX?@{IT${I+^FT`F5#e|Cgg@lFXOa>%qcH=wEFrCEwo* zmM{eQdwwZ{%U52+d7`eH*T$YJY3ADbf=I5K>@$U4ndsK7wa~F4; z;;#G-T~CG0p$L6?nZ{r8*FcCZXG zmv%(^#$S!Z__2Cb)-j8iHRdB~wz{kKyuCE%$kv~m6h&Zdg~z@}qa9w5(rw`1VI-}` z_Q@s@8mbg^6o^wg?0>MT@J8>P=k|>b9IAJaD7su`$)&0M)+Uiy7AN^4H2xl2#tFE4 zcx#`&wZD6u$u^h;AT}Cg9%MQRq-zh}0XZNmXxn8^&GE~gy<&-bwb-@7&=H#w1pF6I zK{_qnb}gmb^o}|vz5LkFIjZSQN`o~IEP<^WXpoJWr%p(;!@g8(Q;3*5=8_q_{rzC; znBSl%O!-&ZzW8mdLr>p_o)Pq_-UZU;Q5=B=!xKtRsJTmjd-wNmysLMV`LcL@RN5G# zULS8AR2Ju5)RKHDkm%#(Jmaaiy(%OgsL%#P5l%?d4y*@AI2EHrJt zMlft+)oiuBRU^)p$CpqP?+3mA7^B0H!Zh;CwW-Y^Q4ZVom}ylDX0_B+vOef)MxtB@roC8T3|Z6UfG3}lSq>dJ-cBB8{aBYlaPsB|Q?X<{>$D4Wv)~_BJg}rbIZbO-`Ns@<1%!mXK`)&g=-Oi zSXyfTKrh~>GLCO>#;&f6S~sqN%j;oSQAz!8PI=+3L9iUoCuU~miU1(2B`{U0L&IpFk>XW4YV3Xaz~A{-b>ZDKzQ_!kbTn7VZ$ZtD?IC4HHmg*yQfDK!vc;n6=2$nE&W_ri$Rl{aHWwMxDhWw5TlK|IemhCdvS8cSdo`JJbE% z7|949%Gn`$wP9C7+W*Hk79jmb{JZQ&)D2meFdAMrwOYDrT<44adeP%1}H`a9zsi)r*)e^SrAYX%r_{Uy*2R?a2< zB5-}Uwy>?laO*tVTn5he<<8OS!9}Wr-(mFQ>MapW7S;A>j9}B~P$*5C)@*Ih-Pcz* z;3XUXT*0^(a+&?@Sz~T%%}9P9+pCs@{L%j~PL@xnkt7U#u>kux(_{W$*Q<7PK)ZgP zcI-7Y;a&5+?yrRBIjbui`{r|0blN@EjlL4a1Ip;`P}TpuUl zaOms8K61)$!Zxnc<4wMr>Y7m})=C`ffpUkTjegG0y(3hGxL!9eA!UkXvA%6Zb}ssV zWELlM!G#4425$MAz)28lw`vQO$H~jc2wD1C9?46*dsLg z#%>i}_aqWH1jgv3D)i@=MTsJYRSb|CI|v#*A7`_-jsMPBlD4Apim|>NCu3a3Qi;Jj zw0(rnVs=clZheSBRRwEk3hZfs7bsXgFQKV~xU!7@!uxBUF&&GzBi_6B27+K?7&tT9 zft)I$J%jbE$Qn`fVMvKNlDXmU?OFjjPM?zASl#X=WXQNs-qg4u_P6w<7!h{lm;AzM z^(=K;X;gA}t7L6Z)w4EHI?yHOIq0=cA4D#2y-y@8eNY+yH|PR@(+P&ZmXL*)%w4EW zu!3kb|0s|=kL@Wmizl3t);+9Jdld8!nDu{|Io!WV)?fs+^DW*Am$zWzbW=WfzacU_ zAuY+DiAcDuos$4Re(+a|zI*`1B7MOA7TZW=ZjmHp!VGGNL1hZ*CHk?@(Mq9m`W`~&2JhDLM5%5)?}le?PMu`~2; z2h|1V*!?a=0Xsm;Fz_xS9M8@iOCZ`rEJNVDybw?vDL_unD0 z)byogrsIwt*lQI0_J#qg;h|MS1nh@wK}l&)U;n?%3t)?BI)BmS^BcTQ8T7Uvsm%Wg zxK$3!XKbUPhc64AFo0QX5W&;H=TAidTeYCPrkwQtniai1D0u-O=A5Z98n3ZJd4+BD zgY`xVM?Phnfg`G*Qv{g?F2k~bYdNd0{3$N zhZfQVZ%C#Ow4*{in2tC9mf!v*Gv`^w7cKrqQ1|`{UtBTx6quKV1Id=8zSBiKJ~|`S z7gE%-CNeyL?Ds5j8>|{;lpmO0#no%B>Jti`iN1|OMT6~$BIl7p^MVYKhRY!iNUq&Y zQd^i^%goBEfx)8i^s#b;InOod;p0<_8^z=u}4x zLz+I`hIgRR0<4ZD108PWWfD;fOl%u;W1M0(#aFe@|6xkp1@yoPryJ1y52JQ7jtv&b zxD4!VZa%PG$XJ+mpvG2zCc)2ZQdpzzU31WX8wt1inb)*A0H8*{Ms?sgB^v6M_i43) zW$J!(vn?|_klzfExkA8D^jG@Z)C@BXK?BoELaq1lWoh-}A_4|`=n*x)cQRGAYi9#U zo&*5=qs8NzY7+=?)Qj9Zv9ubmfZH(KCnnAF!IjSOX2*Vs?ANh2v9}l3E%DcoroaVL z#;*8ndO=vxs*dys$tZ1q+vx;4HV$18v2N#oFI=0bX@*NS3}P+wd~nXOdyWs3^QvK8v#_+lrVyW;petk&g0_a}vA4s;w#s)ck zWo;7$ww{ha_uqGC*gFt>f8}|z1&92Z@BCP!9INX_#$$6u(*%EBeEwYGp6Ytp+H(lU zRiEt^K=LnqL@Q8N%p8D7)py8CX?-=u7ub9T_FN}yi(?K8|#As>lcFH4J* z$$(8P?VoK&E%ouaW%sJgU4s zi6m4R(T=p(yTunP7q@Y9R*0*@Rqx+@AYQNPPj!k;SjP#R2gIe=K!T~RH(M!F9Hn!e zR;AvM9jR=nzm!#2Q-{1G?Ma^C^!eQzQE^3oWRDAuYvy4$HG-knl5NuBE^9=@mIl zP&)E|w&RSUnwb~gO?<>N=uy?#UR_+iamF`QuVj72$IBe*(?H9}Uah#xvRsDbZka`@ z`Q{R}FU*NtRR3=DOW?AyPBS**!f7nfNyrp zCaRo?mN~;OhSsIjXc3@+%(Ev*V!Lx_lL47m%Hv>LJ#Tgx<}=|v`tF=zw0(brLk!U{ zgz$Irmu0tsfEOU;P{0F4ao~L5%qHfU6sPGg` z+dqRiIW1jQjBFM5JT4dZzCH0?0GB9`B&5T+w2Kg%8lr#njEl=^i%4Np|sD2orS z#aWjv#F2Ye4l;~$H6wyp`4u6QVbGuOiS;u!%?fzN8_KGi{cZw4Rl|Fd_ zDyrOjrCuY_x5TF-y7}v4v}c+?&JAi#_b)gJo@r7ku*(xv7LhR$OV^Cnpr=fjGSN!K z*M-Up0m~(;G&?uhjVgBmB?<=bmJ7Tiq-E{_C0xeb7=Fa2Jvu3}12%gE|j z8JQ(?U;mtaH))TW(NxCnYV;IyT&nYk9k#RnmpnTC)=*AI(|a!yr4-IglWOs2t9;RZ z&TeHzT-+G7(x=;REh45rDlO$*T9erZp9+^E^k(R}-Nfl6^7%M{2+q-Vc0OTHRU<_&y(0~OY06Q~~+5sA!`*Zj;An>X@=oXyoCX_n)SQtb- z&fHsmFKbPTW#f?;CCq(b8VhEY;sl9h%zaxt_h-tKtB!gb4c({f;}F8RkBKZ|V-UgX zmw5S$zJX$T|3bI_8Y*MM0*^1mUFz<7YLt?y8NP^? z`iPh=SEOo31XZq|)D8Wp;QD!)!8!0FQD?9=nDy0^_SQGO5o+roib#PDk#n3m$8p?a zQT#>*bN68l;Ie_xGqZZAz@IAf- z177+|le$4Hu5bd4iML#@*E3-Nf-joeZ|KEBmXRp3>!nDz3VQu;Nv>DS|1fEuQsVLM zwr27~R?aT(OR3d&f4fJ3-DEk^a0bc6huZ=tYpyGrBKn-}`wlTAxu=cOJ%(`gZm$Xc zZlaDH3Xobd*}UkBRXYJPp{5Mu($e9pJuO#cz>lW$XT8owZL+z5tWA@d`V7J# z*oo{+(0>?-j=%LUeQCb-NIy9A-jO}|U3*rH1IbiwVu5S>C|i`Yz(iN%Q?ZR~lr&s@<^Da4^Ts;6Z{$OZ&XlzLO*OHoGs~?OYiyoC!jR6s;Kb zLECJO45le04VQBkbdZKjf_c1eAa;D>M3;M&|M z1U&I+2rwmDowwaM2^f&%7LQBb-+Z?0vco13A5r208kH-2>lT-$+D|F9`Q&>I2=j_X z0h%lw_=@85Ey~n80|~%lDBpqpJBKH9H0>mSBFbnGx*7GoS>Zp7dO`evKV?GCr!;pCfsFc!-s9BA?Z)Xe`=7=sGe1yKMZ9hHG5E)i&}%U9=IHhG)oqe0 zVNq6*PV+0_Nnc_F~+I5a(z`yotL>v1FKdK&&{r%e~JWAFYHB+)8 zB3@hBLst}=O?!Qr3bt({5p((bYv|S}*Y3BSW4U+9s@$yLd%+hhQTs-%1VO?q+^2jx zMt~1fIHHtN$^WG2W<3lf*uZ}jL*cJZXN_s;W{dcW9JbrUvqeRR>5iT9^{KYEr55 zn*4^1o;&=ME#8HhcAdNo5zNgH=FtRG$IHD4kd=bW?`Hm-j**eaVaD&36%{6w8=FFr z=igUY6QMrwMs#=t>#Nq!Bz>*hh2Rmnk8fM%tL_x8FcJ%%cAxHO?))c2*kILq1DRb< zQ#n{h>0=R6nw7TE(SO&&)C+5;!M=lB4(}4;51wjr7=jBFSualymhv;-GGkPUL=Lm^ zFQh_ShbQ{!+||8@rRTpVkROwax%mje$8-g>dPb%wBym)ESM#eQ`ntse=<&ZSIGK^x zh2FEQH1mlE`||Bwa*QK^l}7eb9~YXqq_kUGGSz209aF!(Az1o(q;yIy#C5DJSro&o z;WA7wCn3L!FPp3|6 z%Dud~Ib>YAi%p1F?eUdsv1f_~1$@3zYPH?vI1s2C_H%B%wtjI?Mw?hjwnuGU$ygg8 z#G+}bS=^lcC&7Q0?XSy+*34bCg~mPXUjr?XxNr%qa~9uwhaLts!>Knwq2C1}>vD3j zS5T8iCLsbBX$N|76yser^N>)oi8n~AAr?8eVmu`Ft>Mcy_*rh@e;8GI4mEEL=qDSx zd??vyAABE|)k^<0XAV@onusm?i%aA4DiMsne)H++?hC1uvc|VW@w)S-fpdW$7r88T zC^F;1pl))qQyFFeT4G@ewqwx0&6oCb>qqT(Vfp-Z3)i5uNRA)m2jh-=1gpPX2)>H! z*#tqm%@L{9hVpQ6#YJaFw5pQ`58-8vumHI1>Op~FmGEg(GilfsJ z%db%j-MV}DCOJN=8=htxb2wK)rDHx?+?pA6dct?OTUOemX_vXgVbXuQq5mgL=%VtXeFB)KnfE)C}I|X6N z0}M?bNRVpkdEz*eV4#qs_@FcGx2K z5z3GH{^E5AE$q3o$2W$Y?#8BlzusvL=rQe{C6b1tS3|s&U@XJ$!Hl^}ZeD*Hc5^Y% z?ki8%FbMMX3F>9xqw=~P47DI7CdB}IdV+46)I|F$mm$&~rc|oGMit61Lpe8jKh>EI z{GkzL?;tuMihyk8LSUcKkdA4bMOoldE<0L|C?U;wm3+8kTIwtie>4Jee94nmn@amU zt$1YU&|DEPw?Pe8m8a!&>#VN&pqhDCsHCS{ytwLlf!K-=b9OqNBEob7uX&`+dmcJ|< z?N3EwDpl<89fD`yFj>03##9BB?3yfcSu}rRg@=ESq=_Yy!YgU<3=GPnkd8 zUljiC$^aTfYtc_nybzY5E^=8MJx7^?YhEeqc8DT0ME8=8ihRRQ1-o-Edcg>}n&#bQ zng+B5Z(v;Xz}WAGD1BoAJ4q_o^bi=qC3tL?N74BXO-&XK=f0lz*V$HR1Q3~++NK=w z%a--it+#b8ERX0H23h2iI&o$C{b1_P$8xyx`ws&lI{Mu?*m)~2a+BaXW;!}VDL^Vf z=|Zo*VYbwcnYU|TsR3uf{7O5+a8hTNk1jW`&yt}klMP8LjUL}emJBRf;px$p!o)@E z2tDJ!{7(n*8|t{_lGwgKfiQhF1dCTJ@YHv6NDZfOOOi zeLx#T7TCFBatj#RUc4r)iyvb%uWN}nm_F`ZqH9vANTbkukX+f}jvG#K~-X~ho;KMJ@ za?sFA#uFrv6=ay+GrZiGKwyLb48ZJpj{|AvGv_th+v+_a^~pX?+}vibhc1Vi2j8%# zh++JP(PoWCZOqX&7rjD|tt=_V=BMe<#wK}9>4s&e-zS@QUO7*oeL&0=m=S&mZg)#C2Ub>MO1|P1-UW#g*A&=qEA)S(k=^P*(zyR)b zL?<@`$*%7mJ!+_r#3X~#m0NlKwF?M+`ykh5qHb`iF2^NTF$jO_Q z5Pl3qE&qo>)uhJo&y*j&zb6DC@t`3h^-kq{8$4*UAkf?#yi#Qu_GPJ#MYp-`S78dM z2(Kn@g9=avszC_vqF#h=#W)HN>Wt`;^eZfms7}$FL8%?0Z&5tB_p`q^dLZF>C8%bwk7UmjnUJArw+d7l0Riu-FV+<;#nONlOU zmn@m5$Ye{N)+$OR;8X52w|e}w3fet+6Fa8p@I#R;2{xU88?<0&RSOM?%a*LL;P0$-FFlUB{n>K-(U`YM`^dOxI)8+xtR{cwdHt9qz zbB7`&0qt2_-J0$K?3yvXz9b#Q;aHN#^~z2Tnis^-I8_Bp(d(S)WQO&t!`Y$)uiW4= zHWemD$0@_sWOa0*LXNI31^YS^3gkRVc`59?ci>aB5)~`$sXUO)pA8Ne+>ELP_hrGl zvV&!I7#Ix3i|tPA-h3H#+1Ua4WpkuG2QWJ8tkq+iC3wP@k+gAeqds3{Z;M+t1jm+I z$SrYa;;{H9LGi{YVRoX1RoPM_S+>uCKC>>`xKqc-E(BTvg-wYz{#P9(ta=1`Lk0lE zf__H39?#bVn1--1d_dUQpYd%CD4h(jq`t8|nPI%V+a^$}zeRJkY@DX#IyE2kpM(&p z+u$(rC3L(k;86lms0P;y4`i@=XcTI1I$JOdEdFzx%ea73T(zF{0tsm_$oSn7KSX!I z)E8E4?KQeX(}RX%UU&o)tyL0uHb#c)kL5x-BirdEPp+0_p9%wMl*Po2?1f2K3Yrz2 zUp$vA))l5=KWR)o|D9KQ;_K$6OjL7Ss0y=62y<}D2z_7wMYthMmzHd%wNw8|h5zx#t6mop>-nWd?*f@pyt`s(6ymyP6@+jofUq9rsL`0!ts(qi*a(Sf)+_wzWNIpEl9q_Ic zGP9sr9b4Vm)wfz7M@z}jmdFM7spH)=gDQb<= zN1EO%k+Zg3torf9)wzKgAFH-z`C7_iL4s~4PHeE|hED2$l6-?Egt?cgG{qfNqB8jT zGzDeQ?FoUA*Q}!#ubw)WP=fH5=o@ka`i-^wFbexCx{hx|CDJ#8fl?<7O>DzYell=C zM5p{t)|u+^;AkmpS$c9n;gkxiZswr0%U6)Y)Q0AO2AO{K8ub$zQ3+O3dT{Q|i`rnz zptuv-P!e))S7iR_A(2x+@z34l*PyWKm5D}Zrn1Z)5tJG~;k9@Bu40QjZT-5;OOaki zJd1H+->qO3*JU`TZtz+hO8v!r^&%$oNi0h6ABuIMU{UI=$%v!*tJ^j!qgYZVAK7{P zY#60Qtoxvg;e7qt8|DKmuH06Z+mj<>eLAvu-`eKGiFBysVo6Kcm;6Cz^aXpA_{OXK z)hA1O_g(Z4A}S~EgD&|#!`qZUNYQ1d$}_mDS;1FYB}Ms1hqR@C&(n4Zldb$XToiU% zFQUXe8rzLbx18nT1-D2ahkEzm*?~F-fzK8^R6aC0V+!PI`|iCiObO{2ap$zhJH(E# zSjFl3)0k51o-AlR-vLk$q%s`#fRpw{`SBHxByZU6m}4aiXq5i?oEs>Mm2I;Dk%u@d z#`u$l3w<-#pEr(ojiTLc@puAVp7hT31J1%=?NAabMo_sh&WNR-0Ze|I0XPOx20LZw zs)f1^et?1-1v)%%?kv^m^;Ojz_W!fu6R5=Y-$ltUOm>Az>;2ZFpw+ypj6{Zyd-c?j`t@?f=n#h`Nak; zQijk<(I023wh^8I)Z@&Pc1`L-xv9eSUB(mAao@{=-$KX$)(JHJtnk#~zQvzagRz$; zcJ)eM%nzVijj5KD<1UoE_eSxd)Tk}%&Tu2YDh`8qmbxUu{hxcmy95)@j`D^D@ADqD z11CUmrMV-A0`0Drl*xu%Yy&E5D{J-h&q??$lppqXMV@!3uur@j*5|W=jG(Av(|iWSA&SOAVp5orfony^){(&~;h_o!fSgp z`g@Gxj+ut(c9qeI(SdwNZOV>hcV$~Mm|rrrHjJ%})3E++U>e#tK;yGAK$r>ystb69 z@Kc@LwcskgMw407tqw|;Ydo0-5?hu3>WR`pyisp3S~2u4fM;(XIgwcC_gqBaGz9=p zN=Vwdyk}%y{}`V{a4~G4E>$_pa~Bk9p8V`CvHlRJqt!W*Orcc~wCIJ-`8`71`+L#3 zT=p)1KL+789CxCeJysnxPzvp+F0&m#0{!~{DXaFDWq%b9SCBio3AK2Y&hmuZKb^uP zOD&}G%<{F#qc+s! zPeB@MkfiAYfV$zcm|}lCcwV>wMC62=R39CyDTAeVtL8<&+N$S2$6l+8Z7|MJen`o7 zbg??$PYps%`Ie-*T9XL^->h_Ssx;c%0UIRfGq~7DNPmWu8ijk?98^xJFRTknS_c#F_sKi-z6J!7jOf z44n39o{t>@=Djf%t^MhkUEmLWP}$CCl<2MMQj*eeDodlngcl(aep7P{-bS@Ack?0E z1AE>#FYZH!AxR|zs}VB|L_Nz_AI;1_lAVM&Lqaaxs_0tEdwn)~SR9c4Tqdna_fZX0%EL_m0ZLa-YtH5NW(lb zt$Ad!g)71-)EBE+G@cT+&yh{}7inNL_{)b`snB6<5k<(stB?Wph{$WU0 zVhlj#-O+nhl}*5uJS7;F>;|-p7yL?0NdOP;9dbAA-6Dck>=1>YGA&dKj2yHjyVHf% zk~is3*cT(Saj;XoCbZAh``=1q-Rniqf1UT>(lS{f_BnGE?`xah@O<`*UUa7+eW!Fl zjTm9Nx_u1|o3}mlp_{3*!C5b9{+u+yJcRWp!Ys*Us#TAh?2}J8PCPfSd$7$d-*6fJdA6b@?Co6bJ(wEek**H{RM>pr#UfGgD&(>M_n(%bne+u|8FG z9zee)Yx|dCkJSPvAz|r63wn?7y> z3soAbg&l2B5P7<8Ajcw?xyEDG+jGZXMk6L7YjFY!<^4?Bx`X%XiQWk095iZ zLcg{td~d^tYr`wfoEvbK`9s{}rp7wt=T1>hf|ThvRAfq6577J(!L~U-^vvJzDQ^ev z)nCOPK%v~)`HNFn`1A3$&CB5ZVvl!ZvH8fTFbqN*&`|=3qK8!()%o6~wOT66zoL*t z+=7Qpls07RCqqkW-f2Ia6&L{|IS-Y>gc~1QXP^!;{5^l0kOc0L5GC)H8MEZQC!t0) z-*l95-kGco)%P-;x%@S2YUhy%CMw&3B>Kya!bQGr6rI?YoaccNlGA6-gjfDjA3qdR zQ@e2puA-j7yoNzWtM@pKE#Jn7^GTB7FdsIy_|EL_e}K*1VY>_~YG_kHhPdw;2Vx1s zT!x!{4~xj8aZ&>W`1Hzn+>sFeQ~5BA6DkutE_^;>9^73HOU;YW+EI(}sM4HUFfu$| zsux1^%1QGb`sSt9O3DVHa{#ct7CK4c{X2TWV2h8G$6Pi={tB?tXcjwWop-Dx`Z6}0 zmLX9Ej&p;TXa&wK*=DcKLS5sx8{=`zurM^kRUYf9m^mM)buxcLIbw;z>sdC1ti0c8ukWo z((J94pLJ~vN7+?sTeDAbtOOr79wuk#Pk$c4U9@kYDa^rf`+{T@UwqH}(#WUo>r6O+ zz2oy`zB%DpK@>*p796JnLi?BA zO>piQUP%n%z!-x+8CVNSJP#J- zR?<8clRe~x1iWQI2kd;%iH80L`dgvoW!x)xUn&QZeFEAP9JFS|DcRPJ$+;u6-Y#mR z*SSuXAT`~8z%-R^U{XopemzF-bhR}E&i0$UyDTTkhaCu!Ba@g!vK`B_Tir|Ftssi= zSdXn>j$5Gg(pwX!K53+|?p~f5`|>Vs99#78j$RZeT?1x$@9j^&EL@;KOW?%^#p67U*vAc$4 zzz$?i-6HKdETGj(eCm>ya>2wPR@m|+Tktm0XlYr#zUy?Ws|=Ztl=@6lPPz3TMz-D+ zIV;xyDtrIVAt4`W11B0ny`6R9;%?M3vyUw}z%HViUI)C#EZxpC1r2;;)+)-dq@yGj z^BMxpBc0(g%8QNV=)lyLcP=q38GY7 zy*P}0*t)w=X>q;|lOQIdzrTC%n)FpBDqfjBOZKcgQQoB-nvu-^q+DhR|#b{Pn5D)tk{zKPC}s2tv;GL-lSZ-y?FVn zO22UqM=_KUNYCaA#}yA$RokYN)`pHlZZX>MvyF`wi>m^L$cR}k?=(FswP(Y=<<>W} z(1(ThaoAgA*!X65quoyPXKGTmeaTc7SBsOpYtY5hD|EW)OFd%A3j&XYpR@B3-I%o0?) zI67e-=UI+Bexe>>u|4%~62HEXQ+!jPDKaJF3v)GfxDfd9Ufs(g4^?|}WO7ixweIA| zYpI)XH=D-hy&G%8cr*P;bc)>yKOO5%;Wl0l4fecB=8dcYj2+By07<~W{BCtUn7OkBw=*oNw5VWCIwA#RiZ zL~`7Kg%4K9un1c$SgDhXWqt5$PUO;F)ble;C$)`Z*CB5}s=Xn|We7gXibc6L zIP#dk7m~EvRxN0vpo8m)JK=gV<42wRsI(xVSge?DJEQ1YW^cPhUduSH=EE`%8~a^1KoXJ3IoxT$yQq4iyO;%R>*>*1inQ@b8bG% zoo{>r12lOJmtA5tCSwWkD=2L*C|R87uh%WtEWpYHS053!YaL>W=!fHYE&99ed^6N* zgS1<4HhXF+NsKhE>ROO>$~bAr@Qm)B05B1%(>KJzV1q+_HmS`2=nl(;R@RZd6K(2D=ePejcGle9;VJ(Q9cAkf(l?q-#&2DMrzebH|eBdjUu$=I>ne$R0P%Z*PCOGC#DA`gzsfF9cp$! zZqfB)mdQPN;^R$vrtLk!9sRnh!nFVry97xO$0-&?jP9%}_j@i)gyCf5D_5CK{rNR zk#5E=)l1q0)nC)wOKtOYS(t~IB5WBzHF%4)!K&{VFLR(h%$43)9+)P8G= ziu7lDe2yBOK{Kk>wvZP2o->t_^a&ZaIm3|^O6ZsNpAr8g`FN+$v=e%13(FEVoFMTg z5ka#n+yi-iTzhZCmS+^d^g+XNtvAMKEftIO$1}`QUzD3e&7R-Ck2O@)BrlTXqvbd& z%X)tis}g?d9ob`WH%)oUXItK%<@()}*00fO8&)~Pa*HyaZe%a~NU7U*#!NZrlj=4) zYzk^v>v!es$}G8EMbSQ~ccX`MEYuuF`fL+JPGkC!PHtGos@lby3cg$axQ*0P3~Ya35G&%aoyQdj~W7fyJfXv zrnObUuu2EE=<|R4^`Vbbx%zApKSV_e@k6QtsOcOWEU;Oxr52X}Te%F45xrf8n}h#h zkW0N@vlFBL<^R|*oFdyAjvY%*I5c;=9Uab?{oycAOT*i_4}vG?Of2ub(?y{Ia;N=M z<6-A?L|YWj{X!;$QKcxi$9H)MYTC5bC3G(c#zQD8&9j9Q`^dH>D*QN4W#Q?YB^SAW zL)?Y=+I030Vqo<|7b+CV+}3Itr5t*Nxh_`06Q>T#%Ca4u#$~M^Gy^F<9$FJg_RFQj zW_6t0)4Rf*dp3z2I$gapS;I=Ct&;ZT>9SuID~d4R1!mt=ZoV4a&JkM?=win^wF(%L z!BWJ-a!=O&n2)*9)xZ0OS}p9X@k|fRSZ3+8HG{=pi>kn%#g$nO@%=ql#F|HOdE4}u zzJQ~{z+g}HV`moF3*&s1l>}kEjNgb`)p03lFV$2hBuJul7p_1DGpZx-E}$v-En4?45-7NQBhQj$$UIn+ zZM)#bb6!;#kK5kgrh|o193?3#RZ@QB>2*gTPO@^@@2R7IyPL>ApJepV5;^%(k)yLX zIQDmk6Q`=UC*8Et{&&c0`;IFJu`c`9De;l_E6O^azT9E>TUY7U^%t|Ql(!`W~hrd?X5KP`VP~#@Erh1l_k}TST zD_Uc#79z!JXA^hhNM|T8;ozjqSq+YOE1xEIZca!9%WVohB;(e-krWel{d*r{QvWC% z)j2XeF7eHy*OZ6Pb%e!LuMi#!uri`^yYc_m?BGLa@FlQynvzoVFeqAzO}YYMqtz1r zugn`c_|AZeMPY=c91AABD{_*pH{M>|;a6kdBG!=~%l#5Fmi?wzl88>l_m@9)1A@k=LXT@+ZX&?sFI6dD$ntESsXSfoXOLGK^H)?^I*j?>=4Pp@M^-s z?*Q~9n}yhpHkpwLO1f(0%gwl4t`of4>`9;<+_tS|g|GL{pB5w)hqfnAT;hdrYLs#B zMfMXs7k`-DU6U*F6= z28=15{E1tG&=en32fBxELlM>QbNLz_m0qhz#d|9Hu9!R&`TJQsL0E!8-}Z#5hR5Ur z0=rW?1@p)UB#KAyhY5IxcTMNLtX}mAHV|OALogW=dV={^qj#ab1~CfEYigpZ?iZ7R ze$=o1MxIm~8XA)U|1CZ`-})>$djLsZAQgVwaS6nk%M(UD{)Wo#n2Fp=V`6dhPl83f z@?grqki15n@N{_L0P5|MLlc*?@_eSBvPYA`?Ahtgk1+mzj)WT@@xPBQvVmK6^zWCE zZ0ML~BY0qf&%2y!L;Bz0RxeleS0<|Z!?9AWY2!nmKs}xK-2TkJn<67ui(|i*I_q5a zNGEiP47n$Z`7Rg4#r+u^T>aD|KI3`?U^_>9Yyul@IuBMlwRT>IVFr$qYb6qDN4_}> z>1Eh8je_NE)r3Np#CejQ?fvsB3p~RsD5Ad>;XHhFupVMNu;&Q zz-*laZ<9(FAarpX0G>8Pt|L+^8oN0WNgf^Fy?i4Z&F1&RmA!o*1-@e3ofzM+Pp$~b zeR!v`h7ebJ%0)V&*SJ7!PL5nq4VDgroWHo9jL5=l9E>YNip_fYSm$VprOIz2j5{y# z*h*dyv@)RnJoNXr2agA^7QeUBj8BP}{8-yO-4Nk=ERjb4bc zPI%?5bR=N~f5vr_W4_Yb z@THp5y*``Bp5AB2eTMBA>%S>SMLxg)E^ESY;X>mO*Ez5IV>1!7H7<$y_4#d(} zFg}qR!F!a|uRfjcM+FZ<@`-L*afXoBqu!7~8>J}jbF_J!M%yUkpXci%#|pH5%Tr=U zt#A^FS{0_TXC5q=EmTwPy*)!N0K=OEdZ>ZDo|@1l)2HL=AU-`$jM33&?w8kxg4>yC z34TR=%gy~~#1WsFMoBQlcuWeY+=pFMeRH#O7bk)4M+Iu>+Gch@V#|gFOEiANEVQf< zJV)%_&4;rj-4^5WOlk$gCc10)jBwWg?Zh{Cz%vzZW%JS;z$1dKCUWmqwB7JmPxz0a zZ62|}-TVdB;CdPBizS|t7m)-5wA@Ct7D=nI(Nllg)d6mEoVV+`y&O+S`#Nb4&xwuaCJQ7+t3%dJg;w{Pz8!{0FM| zd&C-l#V-TNEw$i%{(MYZX2Z0QLhZt~0Xf^W;EZ!$Rs0nA<0p=^TWcQ}!}bkYILvI= zA7{GR_n8RK&D^g&_LJ7Wc8)LD%axv_==(_CJx|gtclI>C@xQ^v_?_@KRP%q|YuUIpc&#uyG55zq7O`!D|0{v7zN`*M6x_&?#+xzs#2eJQfhuC62x zEUBi(EiP`FN05z^$_qynY80VT0Kq5veEo*LE?)QpURX7&`-aj8TX7VrBgG3yK?=O8 z0(!13%XM z8~Ef>kX>zw*-m$aewbXA1Evl;X1|&5fIkL2H+B0Nl?Q;kyBh zkSHn{VkG_M9Fj0ie@ovA{ujfhX||ewg}h&;c#bA4rGQE^tU391s){-`2k_Kj5lgu-ApZY>(O3SMg4a z(CPj)@U^^}mAtuv7&QYNXLY^Ij#CF=AtYxT4l|X=KcVFrk0vTIZ$$q9ue+z>F;IW?PA(j408!mdIJ(lIrQyb75h4~YrZem{6~EV2Ip1N ziHke7H{WIa2t42%SJA(*-j8!>;QdP51)diX$Xg(1%VK);`_0=fP0Q>@GIg!hrh7(ht~ACe3)l)CXHJw8-_V5I)MFg$Ope7 zz1|z9={nPc>X+P(9P1tMLHVC7>;4<@K8DjvIEq;2BpHDva7S_2pMF0o_OID%Oj}JO z;!lb%QAM_!;k34W*=7F#O~1DT-~Lz;$EmIp;{=~385%3gf4arZ*=l7A2Wy?STt z{jF)bf9(GN@osC^%gbe>#-MUp)_ds&G0xq{xjly`fnQyToTY|N+qAzE%ECoPtL&3L zI@T{W8>?2i)PpaWs>I}yNj&E~0&~!Q6}RCFRDDL(VquiEc?c|WcfQeqjtIx(b66fE zh2d4VhiOg2jDyJ;>JB>oRo!T4E%jMsMLD;Q8)e2v2$^o4rN27*Oh+n(;wmz=IX>s> z&+Mt<3qK2Z7fkSep9I#g<1Z6x5xW*vlIFuo@TQ<6SwQZWdW?<~5ECOR4=3@D!&-zo zUEYzXN=kWiSkC8+6-J2n_FVoI`dRP`TCvlx?_$pk3m^41Il(5*jS8XJQ4U;y@#jCq0NY^89St7Tj39s7~lcB z2eCNhbjNHSyo&wv{e=Gj;G}wniSY8@_I&uxZ#5Wn==NzBv8S1EmWhG(c|l;(d3_@% z=G%qjjGx0D7eUwWwFR`&V@V_AlP}2{1sk`!^cepD^$Pli`%w5@@Z-aN62I^ZT}k5f z)NUr!uVk8EB1=%1g=4pGmK7{;#DgPkCnOwK%JHuc(#)`~l-1tp_Gjw85phg%?AnXO zKixKzmzDKDs9FC2X+MJcpNH+V`8Azd%(AN%T@vnAMj+sD${6RN2N}mT?%%ba#GPlx zx{tved8~Np8!AL zqJIYLyd&{bOt-tguz4QZUoB+0vbsP~C%N9ybcFmS2+AYhYC50LlhA%AU2>&!Xu*F4V5P1thmINvX#3s#spj zrrpRPWhJcdVP!GR9syCEpTi(Av*Y?bgW*TQzuJpa@UM@)7qph%1=FmPG0Zb?mzKb! z_IEJOU6NFAM1b+#Nm48NW;m%+()u&}bj9G~Jz77EU)m@3B>4C6OH0!)?({o-UqiC7 zNUn9S4-1>MIA(EkePYG#tz=-~AIpwNLZEhaC9Ct&ZC-201optq6Aj2y=Te7|*&Jbp zFni=z_6hq=TzJFwZ1|sZy z13%!Z9|E+Gg8u*&tv(|BKGNF9!yZ4k^B>BLEvy>Vxj)(wDPNW|aD{FXA%M;2FzDFy zuvorjsq=XJv+cnA(;Ah-WtJvFGxxfI@9kBg@dl$6t=*eTh;PIKWh#xIv&rU0Rhdgh zLovxlUFu4@)*00mD2zDV{J%Zsg9^yxK1%GTLbXD*q` zat6)fPC(#=+B=-rvkAAcJfkr>sx>@FmhsH)8D0j()gQIIz8PxLg*o0l#_g2;Ov^7It1EH-%LB*dUz495pc-$- zzZYr5soSjIzX0@>>S+My{gwp%YxO!ODx|6B2l-d#AMI$#XZuii#UR3@*QqEZWZ~^? z+o-H+mAGBpDjV{&mHNJ+$+qC zrI2pjfyY6dkEtK2u5?J)%MMRm9x70im-&09nI!`rj7d;b8ne9fG8 z+IIBn0sjCVD*STq@U+s%HjvpI^VDNC$^tiUL%4oDy7BK;y!i)&+FOsAfd?CMKK2g- zo;!Qo4;wwER;15R@o<2fXbao^L{x#hVB0y1zg zc941#*FFCLr6XJ%U@;AY-y`}Qf1i5n_I=J5d!1Z&t1ZclU;=(ojAxE|^*`iSX4=KQ z&w%apshlmvz3ib-wpgnnJHYF=Jbi1&{MC_w)?o&oEQ zxhK>0tm;us7Nuj?$B49%%x3_4^#1@o zs-?Z!%Nj^<0l+yNan$ksD=kdc`o882Vvpnl*?1j$cfiGYQHMr*4fQ=0E*EGUv7Y>P z(ndv{Z^Xh>UT$0meG|bKf=MvEKy;VakFt-~9go_3O~QHLUKHw9ykt@@1D| zZfWOhwpP__G%HgmiEPtQP+Kg-PjMWfKmg~TE$ps*iwG31UH)U@p%PnOmZd8xp~?G4=5Oayk;(g8UUq>?<# z76Ph1?})AS>~DnR9OEi@_2-T$^XoS*-#o9m#xeBZW9eKuwo^|Vhn^O-`DU7SZ96^n zTCb7OLmSC1WRL5Q!{6|7{k84X>*9Zcaz+H7C)m73lFI6O78e#UBDp=GuHQhIMdRC|6U1P+$<@bvGO{m;zTCT4<-(KHcv6kZM=5xAPrjYW| zIqG--VN@!GQ3$>_5iD*&Ja)(B&3W~%wF}uoWN7yl0Dyh5$n9F%J)XDXtJbja%nf+s z<{=7&MH%ItimLQD$pyP+zk66X#x~g_#hxLzc1;q9?=BgN{{X6xka3PjO#c9p>s^fETzuH+qJ3}`B`wVX5xW`jpJn7p10F|hOb#Ex10mmfc7!~z*!dY}r z8EHNt@C2ny+SQuNY8aK>EE~Q>1mkxhvD_*2uOk;h!&x18d8s`QNBI4w+cm5{CV?Gp z5mq^skE(P(q{^F{+-xPG18b@0$CYz&i9^H~4tE zezr{S9DiZINeTI+PyzP^h8wTSbI*GDhvWBy^uG`4+BLS6UMqQ$;y|oRY%3xH19id3 z;O4Nw$|>?a&fHDq9SiwSH@UIXOwt{u zMTzHRP@zFXy+9zGjAZd%ZSZfxW5mA^G|MpwxU;sAZ!XB;L{Ng`aq06(u|WPT@H*qu zej4h~>t7OlU*cCgjIzfNZoGMt1GhfWjE?x?vc*(#a!AgcU9Qf1;U{10&A zQF8Xtks6i4OCf0wR{sE&YDhW99@X}-zrVH&ZEois0c^w@131YMWZ-)7UmzrdQvI#8 zWU*vp02{#Ju^p7eft+OJyJwEo_3VudM|J}f$j5wlt~^?_VHeowryJd$Jbu)^E`v+> zQzoBtcQ4vCE6Zr2jf%gRk+_;+_Y^J>NzO;0JlE$P-l-;@!gyrcEAAj3pmUb`*W16g zO~`E{!#Z-NUQ4T~TL&3ciC6i(dqsgGXlDXt&Q)+gJUh`OF zw%EhWY!W~}d4c|Bx;;AcZl2c@9jZt?V;?93Iriu2kH)zK(^_Vgyst6$gago?{B!MI z!{HwTUHG?M(!3oSbALL;F$7^_AH0~5f9aaGJ@H8F ze+u@MSQcH1yJL@qQgRdy03A9Gp4Gwl{{X^Po+r`nA)jIbj@e@iuhR9 z=B@22cRrp}Wg2#ljUSc%GrQFFeHQasyFY1@Yc!6>E}_2jdY(>u;=CzuCu?}FtQ^NA z8#jOkHz~@o>z;$C_OGfuN8#9J(rzrFV5U)ynH+PI`2M}C=8Z2xy}s41bn9>2x&h|N z)e%AYnb>i+(0>u&SLj$ev6LG}i$b2Zd!B{hn^`rvX7M(Rcge#$V;mgx9RRGzVz;o~eV*A{%Q7w>FY`#pJqsTA z>Bj(7!M;QNq`(GTD`(JW_56C&%59##I%+D&gI<+1y&mUKjYOt3+{#J$jxaOy>&1B5 zc%4}TJ;?-&oWiG{uX^@fbS|Oc>xklPg-|eg=O>H;I_IT)&?nSTYyjYK(0_$?VD94X zis!Na*6Cz}SGD*X;yGlQ9d7h(LRnKIb04uv0pNE7u4~JDEvSgRBc^HafG3vP+$-SY zX^AkqNI2WhNdOVgabBei_S1YW@ivzMd18-5)W%5KNLv}sJM`^eHuw)wl0OaT0_EOY zn|nxu2)}h4iqL_`AAIA0eR(4|uIzm6QD3onTyII+x$Ab&A|yucNV2nth`}I+#&QV8 z)6@Ca=9i4TH9fb6wR`z)?d1Dq?0Ty-dz<%nx6?}24BHj!IuE?ee1NMrK%>0g@{_ZNC+?ET>OzQRkZYf{}$oE^y&P_{iV3%NMy&MSs< z(8Dx^lDplO?1V7ra$MVuk+i!IjxrFFjF0!Lcfq&v-slkB5?xEMxt0R;XM^&&CMA9HM%%}q8+U6c~3j+*o_g2;iQ^Hjv!8c=w`_SoOG{Y{{VtlTzKQeKer4J_?udi z`VBk7+LofxFj%7A?-UV)h+zDooQ!f5^VC;edJ_7TTvB#XPv>MTzN}@ArH8fon*7h{ zBK}CPu3F+kQq_Lyu5$kX-2)iF#d-dzs_FLkEF+0`%^NE)4I}N(M*bWF+r4#OCTZ0+ zzc}J9#d0~@?ibV6ydPfDmP@}j`+wgOG;#5}D#}4Q$pGVMBms<%f9JVo(oWGor)l8R zTAA9d)}N^Pjc}>k#{7=)aCvEU*_AUvIWJ1m|H=?-!fOhR{ z?#@9N$TjRvyd=3_Q;l6p=)v%ov!Q%R@XA_hw)=eG)>rYB^2mL-bz$ai9$bK~K>z{H zax?QDk7*~v{{R);co$!dB3p7}jj&tHg^1X4Bp6UW_TEV+HS2yA)GvG^@g7KS8_9=W z^I(-Hj3A2s4XjnN0Ul&PZpazh2sP6F&|VOIOF~Z+SzFw|bgj}bkqa>lM&%N>R6S!+ z^8x@P74bQq6M}ja>tDM+8$1#4r~V2V@M3=n++D4u)r`?zXt%17$u^&Oh*gFZ=V?+_L?jZ8o}gFq zqx&j&u{Di;TPG@BdwDWYf^!iEoZyUuf;!jfcfs$BULp8T@oP`kZD$7FKbA{xG4RPC zHbDqoK?Zy6=;(ap5PTc$M6LN+QxB&h1P{oT5#1KaW9+knBW$J>B9qo1U zySL$E^iDhC@9lr~`&rI0ZFEwyQvU$Su71J%H2A0DX?!#X_M+=c3_Bh${46>7l zjn{x!1}by7aadoq=Y+g>@#FS9@xQ}Qg1T*`{j(0@m7bOYD z%m+2(ntjZgHh}&lj@sJRXnB$C=2umXmw7v4!zzM0Ge%b<8$cNFUxMGX?z!Q=6zW=5 zsj1C1wDQRc-6xqYLbmM2yJlcUcD7ishWtsUUMYo+rK3T>^}LNA2fhoO{VVbF!q+j|_^!g{ zLD7}__49rG>-3Hpr)brE9KS?}=zKrp4}iZKG|LYX_~%r!`(~NptMR7Z-9|>( z*_X_T_VfFDo9aH76RZwmOI!C$k+tKhE^ z+1%;Z?XSYK+TKAmwZf>fu#z&-PO-?LmQc!BKyX1^f(?CbJXL8WPPMJft!?bP{LcJW z)a61IaJN$B_wUkI>EzM!>P=yFBzAXEZ)xRFt};w`>V3{SbjMz6*gPxYokLrw{lse#zgkcgKH^TD1Dd!v6qZ6wCXvh7nUM$iV9y&Ps|2Z&T6pRShl?ig&QFE3b^QFAwC9eJAXwUYH|zKHl= z#a=P~+=j!$_SUGDI!>8yrKF4l$vklGCESsLxtx%SF^;5@#|dNMO=rYfwYG)f=)Ti; ze>^s@Mlczsxs@S}7bGrPMgSHdWQ>qGuCQp2sp^TQ>8m7{w;wDvkVf*&C$E&r!kRBiD(%{&Pjx;lnTpS%Tk)WyP_ zIbhO#E<7#*RT@%@MMA> zOQf<7-Z{0jMT+*+47TVO&OkU>_X58Sm3V_spX#{y@>E;)erw44eD}i1!arNf<;_X$ z-FZLBdFX#TJ{0iEc$>o7rl$m?-RLokwSbM@b-Or8Spdl|4&dE6LV6G@l)07C=&qu1 zG;C3W=X&iJ!TJy3?c7&)@bATvd?@&hr1;Q{(Avwk9 z_{L3dz*gyQ_>vR@ukj}tAoS!{?VrTo7e%ao#@3!B3+0Hln-|Rq z_S)Wf_eu8nQsNaWz)50>kSe~=P=Xi^pU*~(@hV>wc;*ibX;L%Y>JzP`Y=uS@_?3vr zAS!Nbj+n1(_~oH#{ww%3@v8p-O~j)~@ivL6S#yuvK?F84$fdD`5$?`=SMHL-EB^6l zf0LDKJG76`U)u}*3Qh2n_Sf<6g#0t`eCiEp;XQ5yv$wg8trlxipKE4Vn1)gIkf_2< zrz#i%xF@;&s4e_a@zeGg@m`3d`)gFt-rHWVw6kxP(mOeo?xc-E=2rVZow;~vhV6h3 zYx$A*8{!WcS@_Xx{0}3+dpfac;kAtuj zSJmTdsbP)7sM-d?8-_d;Va5nNdvj1}w`(+;LXRp{&Zj#e zP%|qWbU;ArNEP(7r58n1*)(uTS+vX69vj!S`Tqd1Z6J{z=PFxiD;l>AAt2xc8Eg_V za0e#8Q~m<#HrE<{z2Z0#p|RDqMr$N&77$t6!(^Gq9bq{3uO`qw5Zrh!^}IQ8ByHjZ zA@4u}RMV1AmLdnF%UL8$mxWZkene=%rCRqY@e2u#fVUIqGn6?b5d3w7Hha?~HPJ5tUb%DKV*I zg$%%85_)GD?V6qs4?_wo>ND^N&z8M9ahL1)RrIm2vA1~RxtT5J+Fl_b3m{%pvW3CH z4p8Ha^~vMY>syxA{v3_`(BdW? z3(KAkau4Ox9qEp`d-=CE_J~PRo^fU;j;95-=ieEvBU1eXO9jrisJ@)KOwyFLjrXir z95*K*ft+NLIsTQ5*D}bhB)zf8>OubiKmB^sOIW^Gm7{JrC9(kTp2D#b_R>csfMy4k zU{6k`9mlchS}5OQ70V`8X3BMJbmxz!u&3V^Xqr~&yq4*VezmI(p`=?zV_SbQm=Th_ z{YOky;o>-on36rguyKL3@q_EeI6T&T@>se0rOm+vR#GFQh1Z(atSyx#sTZw*ENN2Y?Ep4rV<`V8Keg| zCy($a9<|VTWoK&*TS7o_b7a%Tq+m3*-#lX&{vb&DX0>$GS27!I=+1s!MoVu6&3k7i z$ByPT1zdsyE^)^>!0%e03tszG@s*AI^1H`sUglEenBpWZG7q>J&p}!8S}eZ~{6V+o zXfJFe-RQHd=n9@Yra1PkPla=vW2?^DSN_SlIb8g_s5$SQr1bZ#p`|%8WbaWkoR?6$ zPc_Ek(~YcfKvrxP&usd0TH2n0Wv6&*(^;1P0L#0(8T%xPtmQL};Ks)aMzmWjZzYtitcag~B~J&Oo(Eo;J!z$VP)-S}5Uf%} zj#r6eAo0hb!2JIJ&TBxz@UO`1K~Pu|(T7v?#aI^VV2P$tlFUiYcpL%8T#x5k$!l!4 zEYN^mx(suSboC%qN~%u57_E+LPBNyw4607-5rfx(!5xQgJJ+a8vt3%svXQ<>Tye(Y zM;r>^_eT}n7bfcW*vLBIju5NCX*Zy zK+{hU!(n&_uMNOI;nux-;V+B_g7r(E?MOmf>J0)Gk8V(_5xK)05EZg|o;Vfdnmyxb zR)1leb|NgLfXacM=i5DSE1}b5ZGQGGCq`wG;&N9GLZ~VW9FlM_qd)z6%N1stdks>% zRyzCWbp2oBzPY67vWX$_4ZfqNhF%6P?6SmQ9!!z&J^8PiwEqAq?%8io$R$-}QNdh{ zf%WI}&3$37co_UV_=(|dS4Z}`l=NTTkJ?q5pJR7dsUtL;S z>GSzeT+IY-A#l4AM^!4!yGS4mV4Q=&#chSeLfSry^%f$8tdlqgAH2$fL|ObZ>Gh}* z4=H~BLE62~<}20g{tkRu)ty&I@V2LRBsf-r+DU=*KPGSwbI^CidPT47kNZX3L?&+z znB#YpeN}F@wMk zt!|O8J&dEjTC<`26*dka)cjATLnMDJkOZ1mQ;vUm@<#o>>ktnauP+mfr!&x@gPa}O zJtM=KotB;PZpQmd#9D8SY}(n55-OSA+Zbf+J-L(}{%`1)FdG0aK9Y4hzFNgmC z5IjAjXjMd zWPLlq9xT>v@eRT(ds{!j95TqCFa2}}nydRg+oO0(RlIEZ*B46`@8a7OYJXD4C%tt( z7|}dA;SDoVveLEbwJ3EMqJl`@N)Qi_vEHEHIZTHjf*9k{xvvgs_8tV&#*LugUNWqv zZSNhsHs)0g^AP-ey*hRJ1u0*f&co#DdGCN9`&0I{)HR0+QX6aJjh82BiYQ^p^e5&& z)K}YSs7SJfl$jTh00TVx;=X9mJU8Ia4(rY0-7Rlo)Fl%6vx%V$!gdWNKXgFJ9P!6& zRjqf$2|h`eP%51BeCjdZ?gjq<_0z=5+R12C6{BN={jK~&1OJed5c%Jk1rP4JO^5Y|kw6 z@O!(-yYS8{nwR1?iL9gzrZg7L4nwW95)py?B4m(p=t1?bR|QeR3Nc5Mi^E28a%}Rg zGfKXMh$2HJ#F+qOZYMneBp=LXzJUFnJQd>^ekR&$`i_mLM`vRc@+`60&u}D>)5=92 zN^7I$g!1fM11ylmC|KegZkk0{S0KBn=BpJ_T$1N$F8=`4io(VUYB9O` zV{4#zv44LqfvLwC0k_F%EK9WaK61uJKR;^Yyj`MryTn>e)|ug}FEU4uWN#$)w#W;N zHs-jvix>w#PMJ0Nzj*^{akN_S`&;f|AeGkR=Undvvm&+*c?E$3Adz3Ezp$_T5+ma$ z?VoKQ#f@s(&*9&~%S3CfL&iGv3lcT7Nxn&RxMflHuoshZ%$A9r-elNlkBL(Y8g_De z`fhtMsOIeUJ~y`gnSLi~TC83fi&6088lB*8v)f6o*co5S8{9`7ytBI$OqbfHXyo8m z#J85P-D;9p*hdsLJHagv3o{`utuFP zh5RKYqUpC@DW41CuNF-s2!7LjeibFPvTen!>@9>5gyu!s-a`Hp-9(V8!)~EaF9kw? z5q|Fg@CQtGuh4OM4lgg1BaBH(OPiOfPiy(aDf1UFncpO52QTK4-yk>JA5PN8mfwD$Xao+ghUi{oHWcNit3(icmpR zV8DUKc<;@8{{S`UX`Q2T6tTee&(ghK^_tSu$6^5kAcLG8-8Xg}I@gXptlN(VuYSLc zb>Ur;)P|Ivr~lWpqv}&f@ax33g}_;~jZp2|gT;jZ05^W!vtKayTKj&1qkm_}rgLj) zAV*`LEs%0fILhZ8IIp7Y>@_t8MWE7VPohIH7j zyucx|wGrI721bm+xEn$Mtfz*-ZaN;rt25SdensUg_fkv0=z0f=^=WjU2x}MW3N@{T z%!I%{DdE^!phaJErUn{AKv(;{?>T4F};*+AX4hmU*ph{P|=%0Kom*?Trh9NEt20MhHJS zx9th>+QqG=y4Lk%k~w9SW4MeWVHoTU5RC(d0B%;t1mgz1N?1oIu2y`@!Oq;w{Yv=F z;mvEro)n)-(Bb<`u*q+G38A zipwy@NoBQg+`6>?02Qy~j=YyHojiRi^6tOk&*~>qk{FseFhefpC$MAGl6v;%*1Ye; z7R?o`vjtM|p_FbTX^yBlFgBDxO@ z*lGSQoBk1t5g(H<3b8bbA=w|_StpYV7C<2cWA9e>yQ9H#adgvL7{1dCP_Sjd*&91~ znN=VLB>nE4NG7rL4}t#x5qwpsNiV~_KIZmILmDz8pz@kbaviX8!>2_!0AoD+cxlSl zhNcxFo{vNB>C*k%fKRdNSW{%xKMCYgSq}=h64vR`bqIm;}#sZKyAm=sc9v^!@8+Zy0TTC;&X$(^Cfl=lW zvJgve&Hx{HZOaVfoK?LR%fvn}@Q$_czs6$gO`mtz%MU0xiPLYK%IPF!823^;fs9v> ze#w?Hp8?%`Kh-2%N%TAWjVwrnkbcD>Ovc|MC@YKsxbe##`PanVb3U7!(xa`r9w+eo zSiSf=@uuHSy>$@VUHN<56h?x60}+Fcqd6JifyuAcy=!c7CaBUfn|2c}#6ZJtHo4vN z5Ho_KoF0|=qx({LM*7O?A00&Ol6gz^f)tk>Hp)j;A-aqK_}A!H!mkpaiC+vfe;7v- zIlO^olH3IW^1B@woMR)Ck&GJgF$&L_9?VjbO#LAJqco^KDt^cR01dt++N6(l;SGDp zmPU(t7Wa_6OrlUnml0xwh71cRUJea;55p}&;=@~uPejUJx~!3IbdAOXA!dk&03E;{ zp^v3|-|X@IpYb#HT=B=l85&t;j>Kvfs?1eY>>JEQECzWC^8zs6sqhbnwz@pND1=`_ zFO~pMJk(vFC@NY_jqY>;;3HbY~f{Id9Y;v4&~6HP6vBvQ2Q@=2CeMsVS< zQIud1N`)iS1RPiEM~XCitKWoo_qNu;YuHtctgRt;2o*$tv2fV$l?*!(k)MfwW$%ZY zbNJ6y@Q$|-y1z@OLh;Dp$FOH|1}dL<$j=Nvhu`X|r7I)nFw;?tq;7w~K>Ta3_#gHI zm|EPJE<9Nsy@I+-+k^|Z?Fdze-EcT46_Y)BSK$8u!)vWR+u?qpq*9#tQts3!# zW#1Viwm~3$EBi?JmGB$kH^#4lAMm8UC0N0CV|8!kqoS(GZzvd9plp1SDdZl6@Nx}* zExOO_WvBcS)qHW_FC6M~>oMBgNn#@~i&mQ32&iQ!cV&>|h{y*gp2ObCsVO%~FIQ*z zY5AWGpH=r&S!k8l=c)No@gKuJJ@}XVFlk;F_-~@!>(|<&TTS+RBo5+BbTg9WnPX4h zAW~#>P~G!hd+;m%2^sN!S=24{uZ|ksuZQfCCT6kH&)ctCBjw#Ys@udz@hC~T=-hMt z0Qg74{u=P7!dp!j!xk`UI&GkxlG@TWjxB((Br4-MCnTJl_2#@2S^c7X0paa8SH03N zxz6m3&dxBEt`CidJ_F5hvC5;L|mRp4<{f8eoS2Ywv>+JCUFuj5}0*;)Sp z!gb)MPZH_gU`4-{Jvm6ev65aH)@wr|#vGIq2;5+(Bj)c4c*Dm(5jB|gj}hHLrP#tV z6hU_)gT_H(1Hb!681%2F^?!@s2tF8iQ%kV;g{*AT^{dgQ&355SsJd?^N8PoNb}1*E zj@7@msOFjFM-Liv=Ubj9{{RNcd{R#me$0M7_+=H-8+`}FlSvVZW(38m+QEJ#j|3DZ zSti=Uk&NfC!v6q*=xN#qz(3g9KZSGa7cl966TTs7OL$f|tJIn*xiS_sve=H+0u z&f|w>058l%d^7(51X%c!@K5$D`1RwThdQ5+p}4TrG}vrWu0mTw=V&)`+N{edl*cF9 zVw@FVqbPy791n5-0D_8mzu|ZM6?^u)@CWS2@SDc=zwnIs6G7E{MX%}>4-D>Zt|hm& zir&qfm|EmQWZcDFVn-#%%%H5sqNP%GXLTL_01N(Sy~Fr9;p;%mAYMOzJeA3)qNfK^7099~_ovNy<$JgHie{1i6{{Xc< zbZr~MekagwXMv}l4SPwvw}LRF21mJ)CbN)&*a))Z^Ry5N7+(u~VgCRG;Qf^+kKniL z=K^aUGP-!}@9exY1lsnyJd4rwE2xCiBxqaj+&pnc5Ad!l>8}X%BUkXqebxX)kS1jl?D{7W~+Q*Kpb8Qts6DQf;-!5#%c5FCs0j%fo*UJSNMry2Hy%I(D~kQ7 z{hT#V*&kcIvefO(ESnsQtu@Wx$D~Wd11pU!m$E;jW`Aa@#>z?XgWPVH-Cv?u|htsbL@)P)F)K9pVj4JmZIJ znKqTXU$>F@7ZUIkt4E%3_m|sGMwZ;md-rR8&E3EA^Zl9ip%;kt9R~5X+v+y42hU7H zFh5-FV~l>a_LuGb@s2$^M)23d4JuKy!g_m~l~*kr>l);VeRv211ln6IylPKC?4XWC zeuH?Y{t0F9GsJq_zZCxfWSE<;w z(OcJV$k!9%bm{OMIA)W)rBS(gMP5mK85H@FWD@nc;%hd zoNt-YBLrt|6mG(e$2sI5z5f7aUm9BY&*3k`-vww5d8k}!9vjts>&xau+}~z>k;iKQ za2IgMX(VO@3@GjS80jQxC6!4|c9F`GI5<3m_*dQk0I|=Fue>ASFNjyx5-q!0cxwJq z=OmVBjB2FkN-lTM{(h9D8(84J33xuw#Tsx;2f}_Q@ty6+KWewJ zoN4-VDH$mD2^^9+9m6t$bAjH!Q-5T?3HWE>_rxt1SkZnB>3S}t*GLxjc{MF@>|7D^ z7MB+BO1R*zaCiqN2jh>~cS^p~t>BN~tomN7t2oBZYgVjzWghe$_NKB~_tf191P z`&IC(Mfm>!ZFv>X#4jJ~8hk_rmMbf9Z*92r3i2$0l%6mc{Y`oGs;N!3wL0m-+cWaV z;z#V`;r&bFe~Nr*@w-~G(>0A}R@5~$(d-}^wz~dw?D6c`X>*X|R#%cv@dY$9%^-%Ui?Kr1-uUt@fF@NJck z#SKeZy16%FQWp0fe1I&|m4zb@F;F+K5$)V^NCU7H^X=vQF|$RcU<~ygKMz{|)bPu| zEUn=v=9!gGY~+-bc7nQ1Cc4|tPb2tS#ebw$YO&SwoUWg;P;M?;$=S7a(RptC%-ugk z&^$R~9hR9CR?w#HvBUsSSe9%k?!=Nwd}AS z8lG|g0Fi6)4<92Jk;YQ#IQ}WG)cx=1^Hr?!P+eg!!p!{gK!b-RbL-!q(ypQ_D9H5$ zA8LvvKtXldGBNmftC6W?AvwU#JMekOzp(9J#*#<&bBlWqF0!CSGCnd- zZaZh3cA_*d)KR3-cg6aH>UVHj+K^*Nl|qbyTPG*33GG~bQ^^g`l2*>v!N*aKIIfOA z4(m}&v0Pfnh4K(f@;5$#%Ji-4KZSlPzaC|^tedh$1>fx3#udujAdc5=rVe4De4Kfj+!fsrYZ;m&B)>!=&kj z{N#A}Sy}SHWQSP#U}1Bf->oWl(dcBNRaSta?i=M6H^Ldo`}Wjo6i;C#Q2Ir{ThmR74B?(|&z+1;L> zhtj^)v;B}hC0;^iySvjY@}+{!eQb}wAHn9c9Gw3EPPMIN`xWZZg_7&UdR!xrz+F2? z0qRPntAc)vYu2H{S+xEPH2(nb8Gn)F)#9wOo%!)J*ZsdA^FA)Tw2*@&XXIdW>x0&m zTQbMyVBj2MA9ve|{R_4Ijr<=Qqv{?kjzt(Oscm-L4%t}X{PHWWv;CMp5b5!VZZ4&{ zL!8MNyb^!FM3=iU+nV+1cx#_jvW^yCJ1@KKkW!86Vl?C;a56%XkzcC1CxX5OXl=Jk z_<5$vyzLr>p{m{m&r^oFk0D1@mw6Oy&Ooh}V(1HAkLb_Ud1pkHnkZ1H=$Gyt%cK5%*hq82()6tA!ZP zJRf?qq$SRs;!QhHVH2v(g+QL(LnLJ3_8^yW!WQ5RJub98q%qop@u3^w+>`@R1F+}-<) zbIzyAD>FI`1C&y^)4EI?nhb1h639WSZpBg^`#{bRQ712{p0OB_wE2%MpgZ?o{*$;P zmW(y34Z8;JFzTggpfm}qkS%Z~bdvp9S3S$bkBk9G=2akbhg*y(G^$01E52aW-0{+g z&AleGyih^eLo!^QiDk=SW+t4+<*5~QZdwFVG;)oe{X_Z+UaI6R=SX2s%b@ispT1C# ztxh$FVo`g2O2H340jI|f_2%-QKAEd6e|@pp5ay-0Jjt1Bzj6gT-py$hmq-XR|1jAn z0V&`f$ko2MKZd84I(&bv^uaUjMZ77*14ASWDGVai76l%ZT)ZeJtknHSQm>O$coR#I z1AiW6ZSh<0uc@&YncTO0G$CxQr@eW4|4k(j#nvp^5nuR4(-Ez_zLUu^L|KIPsq*gF zt6OpX3lX@K@y_9%-kBX-qzCS1(>?x=q|d-8xCo2u4Uqadb>4jTkTeXCMhc&zSJ~v6 z>509TvpI4yo*%gCfDgVq>b&9kyI@dj&igH>0Jd&-1(WVT0wS7n5(|J2+PJ|I1`Rb0 zqFx?(IS~D(+-EG$lk`8?(>uSsTguHFYQJl9X7~+41+?vVvvUmL+qb6ox+9%SJG&*e zwFQ;KY%HfK#y?Pz719R%yTS)ucjbfox^fxO58k;D=1@DSUj3U-QWp95`qnDqHK!r= zrQ^0}`wl3Tl9|vd*AEj^O=76(@U;&`ehyXM9_-pu%a&7 zf_sPP8)d5~H$;Vi@bJ7-!$Q;?Mm$PYY#FaMU6&kPU5mOY5D8Al5H@tjbiUb)&(@gg z9Im*0W>-ed#lg46hNX@)H4-=g$9uCv^7Q)bG09 zN!^gQ69RiGymAD2za;DrosXIB3j?ib7cP6^PJ^D`6x3)k0G>H+5`7s$*u9cN0s2v9 z&9~L}_n5|kUyOE5`l4Hu!<*qzIn8kYU^HTNZ2YuBuD>ci<9=+!X)A2grca600a$Gb zXHzY44Bv3y`~m-B*_ zX2~J>km9lCP=V=h)=ke!9_WukAFI1S7dN8t2^rbKY(*LJ^k8ivNF%*I!^X7gn+prO z51n_XIEC7ic*R$Db`WX=;x*I}2z212k@1E?xyUDo(!0dJ%m2=X3eBJIMpifC?8*|r zxT!bAxy&4RE_sPf4&Lm%$o$wig8JaM@|{0kDuk35H#6nFnb;FQoNAB-dONc*qc4Dv zPrli*0IIK2MdbBY=SuD(0FfRqEcs?Gp1lkBa`}Qcc+N-53w@Xr!FbjH131{7q$;)>b@?}VDHAtE z{M-!j5U?LK?Nx^jbz`pW))#zSI6}Q?ekQf51h64o$5s56M8*~6ZXH>*SSk@0$1S{a zWlLHcOOVOmh8V!g5;j^TojqdgafMJZUEpouP);?eQ=K>jG7*M)52X}3n|EMij{Wtr zS1kYFYt7`*zdJ_b3uL?j105F!R_foto1NeT0U0d+qE^>`B$~R<)9?QI7vi^IQChj$ z*ufhT1W7f(tK&9pDJX`6u7kVQ))+663W-*ND{ZVk!64*Z+&Xo-ZydX357!T^wCO|e z!70q?RFsi03aM$!^mhRTqlS9(@;pbGyG&b32Op6CE9cbN#TqTIH^7A8c#d~Z_GAzP z)i0t-T%dg6C}mPsfa}EcC2)7K@ye%AImTn@1R{4S48&O7{Kmfv8OaY4MSp|$GU63W zI>y|x0v+aH<$ji2#P{L?>3D_L;|>IVVY|s1#%b=;@71aOymV8#Fn@Z2`}^`xhmoa1 zaPT+Z!A42%$}Zzo<#ny*QW{J^v#kA*c(skgQJ-RTt5;U*+7IgL+aHX4Jr1w|a|iWd z2xRnsBz3S@5#5%+-&oa-S)AtxU!KRG(PAFbL(D9ysoAT=;N9SlPc8@qww<)-}0|mS;(RnsXb0hWEii;aPJ`;AtQtFIc!tJN&$*%`r3W`xMj)bAvW2>W`Gej|xYxtImqhcSez_EcP0AbIZ#Y+Z4x%56&BjlO7k# zF2Ubm^KqYo)UXG|eK;UtG=#DS3ql5K@5tn#bi!Q<{iWn%8HPSgSA2RESZ=PC&t)FJ z{h;yG%>KI~DMp955<)_zVf8w(cF{YTw762i8!80%1vCg{MINxT7FAisP|$+|t6;+931fX|O@Rs(*ZE2tG98ro@QTQH)O(%sEOX;2MR4wq> z_C*aGV9%zeF5qL<<$2fB=}CREUV%&>c!bn=v$>utdMxjTL&`s=%Oc@7r;ttEhOP+s+9Z<3z%X<${g zr>9LEeHXna#f5X4s@{8ctBCu)&^e34@n|%HTd3h3UV;zVEPu)DBmQ}Dz_+uoq_3uZ zSnAu^`6}N^pvu}%&>v&kO)|73YiSkdw@$#hdk*QPGU zFh8ds|5#7@QSi4~)u(*%$UO8`s5+2`bw4NAoeTm5GWv+tfeAYX^Q*>h`&sXNTlS=N zH-`Sc$QOxoI{erBBwy(ZYOdFf91GEqgT>*W6q~nB)7_SAbo1taKa8(7(Xjc3yiNM% z%t6Z4ekJpf0eILbrTo(*_JR~KdMQ- zFD@QDy3GL{0t@GNU((+?B_@T~()Xo^%b#Q5jyK;W9WJaAEEy+v-hhcow7<+gO+Qdx z>-zfY@YU{YgV((m4NLZ}T+LqQC4fEPhRm#%p2@9#YaPz#0T8*#;H?r z=LqIO;fo1gaJz3}=t+I3*c68mk3VSV^MgIYMM?k8@-Rrcb} z%qF&79plMmc~h6OQ**WarK{MT#I*R+QXf5tiFcM)yW90!fM(+*@R{t0kfLJ`u56-H z+%N^`k0-@_4iZKOF&kc@7k#fDsSe`X#vJn~`h6^KBAuycuCl*Knd!^dx4R?f0}|-0=J<(g z7%YlO#Xc(wVw|p^Uj_ZcFM*2UkASv~VG(cTN*<*ov>>ClkXv&b83_>NcR&j z%~1O2)FnE3oW_47kgvqYIhPJ*;5V$>X=3q$BZI?+f=p)$sBxu(c|Ge|pUZaZpr0g_Z&O;%vXQ&xLFfCj`o0T*H@{74##Vvzn%+tZGeN@Dmwp%{$_aT!L&M0} zJ0i*@KtBrpD?A)z@5b1zKdgT#Q1?&Q<13GWqQR}Qc!9cN93W(ZZBnNk3+X})a8938AI-UUi6zKO+U2`0#oSpmD zQ6tT~sv)uNb6tWA>F?RSO}H)=(B}i}jnk?V2m2n^EXtEnfc)(Xe8MrWR z?m|K4X{+}z?9?V91^5RMZWgNfjn!n~42>^VM85^GxF1tD8-FDzYW01~!#I z4ZzW0^)Y)lnW}M2Gj6p|>gXouCoa2@B$_j>si@~lOWzx0Z8yt|Y4_?%F zs$Uq{cz8*6H?2M>yll3M8ZJ>l8jZ|-d@4kts0`S%%-dX$yaBG$Mz|pFO|B^ig0B=r zGlISDewD=yoGdBKO^hbW%oJ^MiY^W0`cTPsTH0p>R|pSqjCW7`9D4Ra@TB)U!Vdh96aR_k7D&ijF+2<}k8*9w(EgHhyK1WF+>Flhh%wvo~&H9}=XlgHD z9l;R7U17A5;Z%z~I$9HT(y2qEHYAxv-6v8CH}7N-GwkYO%uJSu{?;7oX|4Ahk`0+B zI!!jh3Ee;{DK}!CDvB?;ROipW7UZU4&TD?P(jkqBHJ9bdSsv8wc+S&>!j*Pa$q}bU zd@KtfXl#-8Z3F7tVk;s_GIL;!Jd$B?h}-oJZ$`qA{JnchW4u6F7I_fr&?s6a(t>NtoUFStx-rHk z#(TkA$(y?G3CXcnSf7UpQ{etkoK|EY&*BI`q$oC_+6@bU;z@6qBq51b)RH+8hXJGc zAWU(ph&T6-uld9P`=F2S14IVGk_eJPHt3?{pG#n(<9xm8N>R=&!f4#i<`MTD-qPC= zjaH zE^d(U@vH{|@C^||(_{pQI=kj@YJ>mzkBaSCcAy}bG;A^|qBp}*bo|7Y@5Wa1_rPHP z6B04OadhuguTt9e`RUN%SpU+A$zm2?4i$@P}B-KyB5y2;+y zdHFJ3ZA0k|mPDTS&*1Vo{EPShqN?KTA#6b$L~q^JCq98bb#jq)B_O}z*z$acoOJkJ z$PDS*=K||zNpb7`eIr&iz@8m+bZ5@(m6=cJ2f8VHHWmn8U-WnEVgo=T!Bd>#ozX0$ z3eiJL9i*nk-DNoas|p)MF7GMO zj)9YkeKgjLKy^Um05vJ@#wZD6*?V4?y5v3Xi(b=cdLy9nF3FXjb%5n&9vP7f+Ji?! zXk6Mj-`Z`^h9THm)RDgRjZ84xSst^T=N7jq>oUtlxa;fuI__qC<zF*$ki{%aXYQ!%SAE?q6<3g*k{NZM>@dt0_CPB+x-RX`mq@8eCjeuM=7!2d zn+VwEZL}78>oep&7OM4*+Qk|(ps8u&u8p*9s3i|%&9tRNOTJxtmJOOtP5ritn0tM0 z2vmhQoNzk99w@P5b-FDd)zdl6{?)PhUHVdZQ8xLOMLwxF!^a`I0o{H>DfE2vH+%uK z^DluC{}FY8srpSAiyHYkj!(|8>EZh7wl63z8i5NYoK3g>WB zDAl$%5!+(EX+rd)s#-wVMt-GnKZwXdA;A{&c}?q%BL-!*3tM^1_8-X;_;286w@?TZ zkoXKSo9p#mxSA-3@`Bc;rnamPb#--#Y1D1)hr`3`Hg8!!k3G7rI=Cw<18;zQ1T4C) zVJvTc3bmMQ+@^^3kr~Y>=eTWCoio!mN%haAkXkUHo0o;(RR-K7lP&$hvgm_Wpm9$_ zsSlyG10{uXBCkIT+3lzTVDvqYu)YI_lVttjO?^?F*8e z(}nLeVBpqMKxE7s`DUzYtb#v$1t`mu9^lgiz+F8&n!EXPQxuk5iT>MPaJ!;_(T?X) z3FoUj4+7)r6Xvci;?Yxa0<9j>=LtstAir32?lf4(wYZmDHv&gsgS1DrL?+29+U+ne z9~v7%H@()Kdp`wyRM3#Z8+F=g5@Vdc$8)pPCzz8w!lK*qNA9*VV7N{px7DBVW*fZL zp~Q$yMV6bbpCk&u-GNq^YT5--5oA8In{g#5o#y6+@)SZCh!HjR9sfU&WVslNSQT24 z>P`0eE13Ps^~jd}sKNkwn=dAOo`s%5_Y!nn+-}8KYX8kXeE3NKYW7;h;c`l|IRC=Z z{uDAv+v#MYsJ0{P?V=JOwP>~VABhUMfr7&;$%!7xVIQIAZ$>auus{BfB(}KBNv6GT z&uT*T^enPokm-&)q)`3V>GDVPp1!}049{6feofZ98nJ5VM=cr)R#8YApRXg5!8OR~PTv$X~$+xyznlxFFGT9-y4+6{zZy zm$vD1@6+pGFkXHxN4-xs>lm*uG&RJa!<>KS@UPuU__4b2%?yL^wmtVI7Y8%^#fx18af-G8A!n8&YV9vyiP*!3mmObh~XSdtN z{LvRg1mx{z;uAGmyNyycct?UqmAww%r)0dIWZ#?=Dj<3MZa`(P#AKVG`~d3vg|nT^ z*bE;*$LAFIa^Nh;lGN!RFtE|}$H7lbiSeeTLByIboP@BndZ*bQb%6EilSYdi zEn1vsS!^bc3n%4qb>#j>vKL+4M$t}GPCUfsuq2dEooaGeP7mM@;-GB7nq?Pb- zXcQ1-0%hGpQ!0sondH0Kl`X)osHOhCM%pCy!jN#ivjvAEJ5LBIp7oCvY5 zM6?P^oVgxRe8d2lg}r|>PB~A2yI#MIr}+Is=Uye?cLw9@DFRg$@x;m{T0js9474q` z)`8W_TPLVQoFg(Fna1Wb4CdzyRR}afn(ME-TIkkiZC4@u_0{a|WFc8j5Y#I-EdTWAsj;J__*8wkk`9BDm#9+s z(GEa)w&fUteI$DLmUk4rPmoS<=C#qU|5%V_skz871?|hiuhxetFv$wITI;FE@Ug3_U2e2_pjGkA)tiNf_uA;NAJ%iE z&5zDRYF%FA;cgAWF3L7j?i59DLk1z;Kv?CQG2%80M6g-X+@dglZqEg?j}iyoRy-J& z)WJRm;u&73q{vpZkp4M1j{bfDc>rRrY=aNcB;PI z)SCc?RlBXQf|o}J@Of4rQT9Dc%Gw~Z)VGr^&MdxyyH2!0?tvep&ZW==7(@-?sIXM1 z(hNJa`DFU9oe9sh)Q~d&%9WbYRQdeJaMf?j{AqXQE{Bkic~J}~FOVbtbocd%Ni(}+Bh1Dl*n(;kQj&D=mJK2bjq|5CATlYQ^jzdPQP%lsBaqBws z7my@R3%n&!qR3;muT~d8KafzjBC-O7aUC02j(F6Bs`5x(T!sN1;fti0uu7=qO$LE> zUl;4^^&M}7rrA>v&Tq1{J37jAm6Fe;(Px_IMFLi7BrrPBK;ow72$_J<+7m;;*U=%Y zOz+>#a?yt_?s+7fCfP#dHpj2vHUE*g8?9_$ZJ0Ig`hU6N36u{U3HZG_HmEJM(Y5d)OVA@5==jek6DJr7&4#q-S>Fr zXeRWqH^Bxlq1sJMR0D>E%#H_1Vq<^yqbvH$r0zd&Z=p?7UR;)5&LyZ{8|}dRL>H(x z>ko$>W)&wXklksrRlW=;`qp`-^EDgUlJ}vUaBMun0qVaMyLa<=y_nXWaQ_(+rWkTI zNuc5=FjTtN030y&z8GKSr-L_~!5_9j2ee*vTpt(R39$H9o&Aw=Jf&|u zVDr49f;KEyyvf?fPD@VsM{L8Rz|C7wnM=m=aPZ}Z&#=mr@_70{?EYgUIp-^9uVt>> z-YPz<+a{1fyIx5YaT*b{M78RFXzxVAhs#(#Pt@Z?S%w8vUS40SK^47%~WPg2SsDLxtO)umgVD%B{u2OLCn~JyrVQ`pb ztPwI^G$JbsaeKh$H`Z;iCCc@Oa{t#+4jO4^m(}4oRc-`zEq@jhor+%?u~+SzL}bgH z()-tH{Po+ZwoI@82hqcRG6c*||Lc_A@{GSA(n$HDx+QSkq_U%O3Aq1Aa-G-89{_s}rW_q8-6_;h->Y$^m z|M}8K^?NOUDk-Y2Vif{5IcA?Wzb6q8S5*Cv#InYXT6(}L+-t2!?&7ixISC2;kA(9- zk^{ru2O53ip0(FU0nm!qp$bcto#x+oSYHL8qnbs|YVZ*9sFxQzf}V zNO;TPXlt$VN~8NewiAv;TTMcHX~Rs3?S9^$wIHc_wy2Uo$1+XMY_;J_TeHD;?u>7j zao>+DVB$%%9RHDsSy8+3$FE~DKmzI@slspX`~TW;a>;zrqk8}6uSP|oh4JmOyOnq7 zGyxXcy7e}b9iKR#4&I#OGP^=1vm$p&?A8w)(pl1HuQxtIpT4JEn5WGQwJdum)=u<_ zp21d^<0Y_A=EO)$dETnOAG3S}uZrD5Ge?b*^)5$lFwtZYvzxU;dmn2VtYj2r(ApAP z5SwWQGCLMbjtDoBxquG~0%9y>#y{0-iRx#OWd|X>&XUCaqP7xjlWG?t67s3LyBOr# z^y=fwd=5ZMQPbtFE=ONR-w!eTO68p!}z`2+#FEz z)c@>^60+Wm+;404)@>40!;$<)Vn<45nG<@l5Bg=bn6c44%01)C-DK{v#H;%i!mI+E zT5Ou}KXXcYOX;JaP5sSyk(EPSt=Y)yerSRZp4Ro^TAz5Uw%_~+vPa`E^(f!8b@^wm zl8F1Dx-3kU{JM4B->DV86`}9#;u^nXrdX4!!y>I^-9mw^O5#|cVdYIZA1ZzsYt!jB z$28Z}!bB(a$K2d_#3{YHzluzke~C`}_(umoX_1fzLB_!DIV*97^fKql&a}ufNBO|% zonU(^YJByL>Aj0jlG^(}CyAq*j??+rBThuLH$c1K3>l2Ew zlSO{V)6yNCcIT}q+4jrZErcm!GmBUG)*5P_%09i|9^dDdj@O09OY$*0Pj9EirX^X) z^7?H4+_N;+91`Np+>(MWV(JFLojX^cc*2zX8$IRLdoXtgdc|)|qX*B~{Aj zq~)5~VJj^A+W&**i6_88X0Kr@T#d(_gG&zixrwOyp3s0TWQFw}9=6Nut-;Sq*?QAR z+-ht{T-x`}?tR>q<%f3F+EJDFp>Tf{eE-$9h{*|xE}U8qrHPb9IEnSLq)ykryw(57 zV(bfnADpy!EcmR}P`dlOiJj*V$K%l8FNwvzDgjmRW$!SUuhu1(G1CAX+z1={^c?dt8VpBvTwBQ*D$;-8xDw3#2yFP7ryrMX^aJ zMWNRAn4FTOmUZ9U?>aId3%%J`>fBCi=XZU*NfgRn@WgC6=`L@ShJASaRH^t}(F3Ot zGGZwnZU$ut0!}jPn()1#{3=jlXWc=t$h#zMu;pqvii-r=Ll=z9&|gfKSMN?3>DJdp z&G6xPos1v337as}{r!_dzRhct9VCLy@!A01S_NkqM3!&0Rr^^~@OClH)4yVM3{$x> zpf5;@P`FPU*7gGO;9zc!ZdMJsZKf%a~{fPExPjtMW}l;nJSqz68_-cfr+a zt{Z-=tVJrH2z>o@qru9JPHCHYoi25TvwVNxk(eEEK1%V>WO*oWIYxvxna8Ky=n9j& z_5yutOwq#<Zt?^`ZH#6T`Lb1UKAoQ==Mk+g{^mHFTPc9HSG zB==WWIu$2{u2odM2P#R-k&d-Z5dZ1@{E-9;%E)1pwc6<_iWwrW8vYoYx!y{>VVNo# z7G!a^wXLA}NQC+6#AxVxzQSQLt?aR)Km4_Y)6SAkjMC@C-Zv%nBwQp47`OzQb9b>O zc2#yWPVI@wM(Rf!qRmJLaJB2>J*00R%k;+7RenAh)obP#O7Mer(YyhZ4(mY+HghA?I8=uL@CQ{yP*#rm$dHf4FAAwYzgw@cJfipVj zgZRdMr^;&CV}%dPxu`)UJoTqD3xzUYxKp<`dBR2@FTo%5*%Ql9lHA*clW}fy z6TU!>&FzS;^|spB6sSmsi(o~@nOT4_egdY0%8&Gd^M2uQiJSba)T*wVwSd-%4Wipn zWLKT6vtsz=c<`x-_wAXO@38Vm{I9NFi9v5hw06$NYR5TWucp1yny6`?*Ol ziJUX*&e(*zm}S_npob{SM7mfJx6cZ8x_Nc)B2rWDfF9#*;rl{}jvyT4;I>XGUP~}| z1kp|i=*4z@9oLrJZzR49;79?29YEi+-UzRT@IeIy|A@F&d@*LZOJzU;?ac|KcU&}B zbpAa3kSxJ)O0tJyG-{K6y^fy{$Af$47{3{qN(qs@m+!86IZtGXP6vWZah?4bPA_N{EFQ`fSN^X_j( zjb|3L3Ssl70v-xB=`1%OCz+TbqkIXBu!Do;zFg$_bF-DWwRfcI{&R z$iMzp=r1W?44Y`z+9LmZbbZ&~rl>T`RdF1Z0k0oZy@|iDE#UdxA)YtdN_wO=0)mH% z`kmuLM9>T67UlEIwM7)2s%Bxd4iL{ccg z_Z;ql+U*q%)IyQL!ncIRejU6AWxcx+#+Zy-+fi!Y)p#xYVd<=5!@q(<+jLQb?fi3G zMHbA7FtRw>h^86pTzx#<)mDmJW3ZsIThvV%`cRc+vZ~)MjQMWnN3xaT&ky_sUSWN$O*VUt z&v9)NF%<`G6>*9`xvVI~d~1AcHln|zNBe{3`<=k*mVdWaZX@1*u>H&e{3f;>`+A7=UK!~#&byWUJo5=`gdYSp zEUtSB(g9Wg6|Anj!_1{i?D26QZU4{=7S(A9yKC<{zvFKKUN7`iIc5N=969l(}7|4M-!({t3T;_kK^0>0sf_ zT z0aC*5@Z}|QvPD#}%PyE%J>$;v*6C;}4SAz{!G|9+P9MWk=|NMdV^>TOW?!)!ajR$uN&~h8*dVJ3h+K<4!oP6E3M}4inDh_rKZ+sjZ9M7xqKMr@>os{y~P*YG7MvC(UA?8R-4llv?OrOI2O8X-ndeGUWKS{#1!*1(-92EY+P9JrCC2 zRG$&vu5GIE;6UlH_ehG)IajGMy-^CA1~|F2Q1a3yT<;cd$P=V)3J8;czeXLji?Fvi z=I$e*`+6*cCI^OLHq$O%Q_cYTXZ>sP>G_l}*3T9TtCL_(Dwg0X!vW;dvk>TURdEcHZl)?q}-9Log`HaZG2cT}Y zWnpu_Ri;a&Z=TEqU0TUILv+Cm8+W{ds`u0E0&2X2mji(5V_U1c#ti`B5N$|@F|i^0 z-bTy^hHgej3bVOA!1D3GP=7F-VX7>>7eyg73&ovtOD6|LMBlj-w!eBRM4{GA9noA;#bDCGWX$3b)Unit46}Pt%CGHO%@KO(#ndb zR-k5gc?5V(<8?wD9-CUkkgyQ%U^nqpI?4Un zlox5xt1W9;-Yjel6CYwaJ6=8?+A-C;UI^CWLsT5tFTl2q9=hK+S92PX#UvI|xM7u` z-VU!G>Qz$UL3L9rJntt^ssS5ih&eJv`8f|qCUUW?f4w7zqd^r^2aF|?W||GRmeHn| zQ|E*~xwNJ`Ja<0Q%ml0=pAJ<$@`gVvg}tR(^2$Ivim&$G-gG(*21VvP^f)aseebr-p{I<)CVo`fqBaUUZ}l zH|IXDV$Q@yC9Vx?5Kzfx(dpQpGGU-e$QR>H#gaynGpkP_gNSPBALv;vNZ%UW1sb^) z%|%%scha486>&uSAP3)UCiqkqr;-Q0GUo2GiL!I)hY~jsAcQ)kH>Q1`-teVElZn8W|!fdP}rAJgc&4>W=oA)(&ijR648|qW;YVi}4=u zfzPCVx#_f8&KX^o0Z29vK8Ca&_egpF(9qNLdw8jN?h6EZuPC{^-g7ycZmvY$x<)Uy z0yVIp@_jna^xeC1U)_TtmVy^%>!R%^VMOO$XxRCEtIe_S@;2_ESLnzua-IK73y13# z4I>K8cMq+c;vS57{a$@4e)#F{suDSSW=Jgx-UW|niav5d8Dn24<7guZyr%zF`Pew8 z;=5}!BO6mTMl%0C31wGVXqGixx_rrq@<+rykNAlHcHE{ylHiGYjbVk ztd|xiK^Vz>(u8VDL@q=Y6_z-v$Q=9RcnjP#U2o`BSaAQ4M!i%`k7pn^BzukDb`|N{ zShl$f%2KcOJsK=Ok>{IPGBjroM%HMX9xg%2>5X|0-Lu^ z9c=R!LJ7Q`m3Q{Y>USNx*=F1>q^8bI=w1S)8E8X zmV*vUGt&jKx_|uss?N5WE?2D=(ped$9944MSncnjEuPZiSWY_?%#2BE8ovvm#*P_G zbp}t^m#lK5$0@s;#g3qz^*&JSQ`$aZaxH0`4)#ZdnQ83mv*{J8pe^Y$nU$S_=-AtI zHAw~mUm;wsc-9RyM9}#0Q_tsJeskF${1p^81osYWu6XQxzaQP}N~wIO1JPLTb!Rf3 zliXwtD7e1nYydD+YZFZkPpX5VvqNjxXNW##B@=9B+0lgw4|UR&T|EmSTQ>T}uU?re|oj~=}e%l4_N zhcDhc*BfUKJvLCAwp+so%AV*&U?y%9zxOCr0&Bq012j#zkgOUzSjyV!AsrbKRQ)8BO$0y8h>8X{Wmluf;GW29?g`ElyMvJ30$zIw+k(Rn$(f^)1G= zZs3F7*phuZQpt38HP5Wzyt~F?y#8VClLHCs24vg#S0&(xKDh&MemB9uy6>{EVwqi7 zQk2}p?D3PPyY9o<-MCoT55Jrv2R(#%$XJ1|=UJesl7-<7^s_hm!>@0)X*#tcLw}F) z*WSm3)tid{NI+w8mZYeP=uLI0Qo$Vk9;R|1I~Vs6uRp5J29lmnA|BP%#MYDWHl{6e z-Qj$R7srkeQALfLV($$@t_m45tmk9%H;hGYU6_V!Nhg}lg^-syA951 z?brykTQ}QS3@_BIRZAV2oYHG2pBHGC{KbWp_thU(AVVg|`-#i>jI8#_+my_wuzpM4TS1?W3qM{|G7J7(}4-?Oh zxOGp(jLRua&s8tz1k|ivQGQ-K_6`l}aieo$OK20GhTCa2#_$CpvbLLOXfPqf!~nA3 z$2cWOhaSWYw(#d#5Ef}EJHs?UGdCsmBe`VnQa3qDy{y9UJr60?I-TjL!D!bdZL~B9 zvi@Xx!?=RgR9^6XPdZ80lN{)3G9ni7p{I#Oa!kR?Dr4)o# zl=hrpMNPo2{tL*WDkh%zh27XBFZHyu=&L}%3L1M45)uT5tFqN)Sw8bJ>NMpnwE6n~ znh^@>hY_ngNRz4zP+Oj;>@USu+}I){RyBqa9zMDmM!LQyr;*>SRY^CJz7AZei50&% zdq~3Zyt968woRWgHpd_5RB;+ueNt&GXB?mrN#;IHF6ShnBHpmj%K7m zteO9gDnIMkONyUkS0|*JypX4)9JgTGi`7HFs=ikkc%=CMMU4+rfzbS zXux0p8LrV0L>$<*nmFPgE z0Ck}y)_rDbSEJDy9X{tc?(Y+>8_yiQ**Lp9cix(?n!G=8q@k4|Y_&Z+So{n}3X@;H zo9Qi~H(mPK3K13YCGivb_($yDdn8D^?MDYyG&m3MjdqgazxcfjPW1?F!^Fr*YPJa%M6%xdp} z(0G@4mzpKMaWpMR^n9d}{#MzAJ9<9v`)})8k2vgeGiT|CYJhtU6`I0g;VLIQ&VxP4jhacomSk!KSSPsNiowR zUog7$DYGn3mi1zh1~dsj5AK$40)JV{HoY;~{DcR0zHpl6ZDoUO`NIyCH-Qme_{V5S z$J!#mFj{3?>!NTx_KDi@NyuV}oq#gJ>F`U>y*b)uv97jP;d#*)S~&6`Rbnaxga-fi zKawwQmVf7jU?!TSz)GvPo~15#tb01!DN}egV|eLglpjt|?x?yw-@NL?dZ9o_*Tzj+ z!R}o8sh!HC)twD*HB*I{e3_2KITBg@J*&?lU0Fo^;R5rfN(7o+ebVLpU+q=?L@k5! z=%2-4l#W!UNnBO<2X*Ba0(q4$c>yEf&3`9ZEl;fM^~4=5cvsfUgkSh7;lAL(Lzdjx z_Ec|QfNE;yf-%#CFU6=7sqw7&!wh#KhMYWgk9NcVz+!M&wY1DMZEDn}_NJ;RerT=O)ZViQLUh=xcI~RwTB$vo z*t??k-kXpZ5hVFPd0ysCK6mci_jR4;bsop}*nZwlLk2QTITe-b1=9{kyxL%OJ8RjJ z%QcBSb>`o7O*|JTwGEFE2Z~c~8RZ%yI!ynum_fN}Ka4=M+bX+hUG;;|Ksoxx2 zt#Mjs^M8`Ul`idgzc-dDsUT^zohaSED8OUIC|U9REzy@}ynf6&Hsk6>1{QsYjf@m2 z0{hUr#1m+LJ`-9pw$FNAWkqSoLs~@hb30+j`Hz)noXuP5H2$SbZ$4A)iK9Mc*(dh) zxPRR;2kNUCqRna%cyd$=OvV^(GjPmaxIUpPf z2T-VyfqkuyzUqeL&b_g4R9Zk5EKxS1BZWtr}^ zlkBa8UjLfjn?VV|vya=a20w09hu?jyw&luZaD1H3Y^K41X6DKs6LhGeUL6igX*PvE zcMtxFH1wm4bTy`=`Y?1;&u1g!ZC16F&S2)XDGbXuAD) zXPATj=*u#S8ib@Hnw8JKC0F%Rex6M6+~iD#(Q5b6kV@M=nT7;W8E}fN?d^>3DLw9| zU2D4L_W7 zi$F(&u#0r44HVO82ZAXiS0xXxDFPn{48HhFpG%gP*da?j!nLmqaIC}s1yex9F^Vcr z`&zH(Jl)JZ<+AcYV%x!(zQS?~pm&>b5^h6hu#lzdsG!c>wwAo(JL zS_rP8Phxy|N>@qcBSnCV9|rGuU3$DI2}EdXA|T{~FsAsFKspEW^o{F?-pNp~qxbOt znO4~KU5s_dXg5ZFcu+;58*=t}rp7SBru5!ulNHC=G^+)DR?05(p2|b(l+W-r{ zPwr%B|@qvw|3B0&kVH`YX5^Zz8w`g2Ebg*#ob(?9ZFsy#@ZmkP| zK0;M~a>W>yV2rz=jHw2(ckjw`-gUI4@4sym_U5jV`pEWpAp3%a$zjRDBu9RKx~Ws( zd7*}L=Us$^okHXiOM>QR1XqTms+@B4K`%U+3?Gncb#&+{|p`cXC|n^w6uV6>pzZ*+sOC~2jSVPIeast zv0z*f_SwXW7dT8ki|zWSwq;g>mc%=1nC?#1$%j6NMG|}FDtOPvY*BSyoqw_JT19`K ze--f1$eSb~Dd?|(25nn#o#hLo|9&aaI+F3XmWR>wrP8bdkB3+|KVsL#?=10iJ)AyU zF?&IAFWAv%6T+XC2`C@;$Zq_T78C$sfjpm!NaC4)pYKlGh<4j)H_~o=EphfYXezZa zv2dKfHmz|(d_f6af`M#6*g*Oi+v&68uMWcDh>JT# zEH|#O7DRi;4tfBK!HY-ZS!TSime2bl@SU~eIU&9RFa3o=1>??1^@Ii`l4u$UaFGf- zx)TJ}|4x0|hK)VsE(vRE|z<Fn*P zU2mHRub+9SDQtFI#uH>aC8~RS*clPdnP;-K0;d(R6?zh}G@S>DT6{xynrfGOuUjz$A1GgwkYAH}9%N>5fpo?P}255fTg< zi3ET1_E{df-K!7Z;(Tll=R;hLU*|}Q{8x3cqB}jl!SS&ym}l~_@FBb7CO(ge*HBtW z;cZ=?JKXETb)Cm{Ur1hnZv(UtwjQvrX8boC%iJdqD$X-G()Ip1pm03Kw4NBP$Ogspff zUVYRPn>q}xzc~^x4uhr!9@8uoqjgbs2n*DEWno|GyvW${nh?v05+>GKL~fk{mVLe$ussox~A8d0Ct-O)>|qoJ|s8i|nQ^8fu5d6i};bQ0tLFW(RMb4$DM_K>FU0!7%aO;{F zSXR6Kjf@lV{M-*crCz9gH2OA+E7CP#m!P2Kz)*tXu`d77U$qY{{;HNt$g( z_LtP0X62WC!l@o(WXaXB)hshk+X9^xt@5Ya}uE<@~^@eW}*i|i& zkRC|G#wC|$EU1_H>i6pMsGQYmFAcUaSJIDsVRL&#KtHkv+`|g?pszRz0XM`Zxeh=p zE{<&3LzG45D08wM`SDgiuCKKt~B1WF>x_#dQGwIe48}`rzKN>VuVeI)R(= zsps#}(Y~IcvMYq4^y?@`e@M448PE>@xHe#{AIjd9-e)o`=Frj9sd3SxOwA$qlzld8 z_@GZnognR%WmW4%6;p<@nPtg>R||sE&%aU9QjLrijU z6}+$Ov}GjTC9#~VEe>sKCElrHxu>A~?(ezV+KhdFE(_iG%S_=m-v?|iQVwx*CnAo}zf)bqfYnHa=2 zIj7e4z~4Mmwb}*p6c56R2?vnZJ)9VVbdJFRle$?gqfar|peHx{oex@8I^Leh<~YIJ zmgI;JyM2VuzEuS#8&mG4n}#WkK0QO{z=JMM!ScYf%@$6U(gJO5c0Ql#(i`vV587L@ zg-G~7hHhQ<&b5GsA6i<^$A>>lGx_?^NRb{f z;I#%E6gG41DiHdd+VzlQ1&{@t-?B{(bh~vJ*{iObdq{{@P*K1IC`QWs`w}BmHU*94 zaE$9V`rH9fwyDrD*ojy?CA+MmFG%80s0&ky+YWH#P3b@PzuKx1J(jY^XIpWM+oW6! z4B45|Qavz`|4{I_nkLxEn1bA3UKYLQ*S(vZxBy8EIE2xF9bsRJsg)H-WhH=f z`g%iod9OY)dJRkCMC6CYd0HsPiaWY#CjsNd%Qa^CqE7V zyc>{Aq8)qXM6yQkBY5BaaW%mes=0`Bx`!Czazl|DZV-`x$DiN+gIFrHS!8biwb{wK zTQHOFmPf5GPQRXqk^-wYIT7XCi?uaf!u%;G4$T*PYVKW4(>ppcqOX{c8VgIpp#dwt z%lq$%c=jMMqrVwPh0F*meyE^n!_i*3lDMg7m^jbn$q=htfpK5B+G@Ej!t48tabya? z0Ck&xXF^b9)u-g%y#yX_r4Dh%|Ju4sTy@!fkBt5Jt!sp#NaRQ3pSkD4zOK7N_l89L zK1#|vYl%?<*B8mq>9N~G*eBwICKw_r>Se6i3~g(g_Tq7BnEvpWUb~aA#m}ebyoCC} z7KxIS0!61ArWl=yi&gUwE?#~*K=AK%ac=yvvTmE{()hXN4q$8vWM%DCF*vna_Kb!J zqTg(uS8#ZCMQNEfP!svu;&QtvHrAuon_vTPpv=en1KE-HFxp#-K35-->E^c0^_0xT z{-%uQG^T`FGVuoPsSfcrAY5*4b5-vtC`AjM)>;^OC|OFlYYJ=fLMreVEvFOf>TE+C z7f_F;EFs!zu!2me&+2OPvS!ea_A!FYLj^C`hw?`DV{x`W26`XvFxJ{k`;4_i_jF<8+-mnhnl=0%wZw zl`J4;)Q`1)6*)a&1JQ@E0?M;3H~g*EzlnES-D#yhW|LyuA+Rn_BrBdARf#p4RdZXl z+xCgw@{O4fOIo^ueQa7N?6LsnNXpEkMF$#x>zP~>^huxps9JttMB6BE_uoVc>;2r0 zS;1EFX0IhD#U@DOGv`>(L6>CYyd}TJKH#{bp$EfcorHbb!IaFzwn8Jrzo43IJ zjyk4oJq5<9?0EMQpwP~JCn`8-gU0$1(#)u=<}+m_Dd)Sn@BSaJu*Ylo;XctHpdmC`3>~O4fS0w? zJ$dz&rh;0fe&7Yj;cN2Lif&ZUvnN+Lun;m^K+!ncjIDxZ#wGP6k39ju*#Bn-UUpfW zu*7yvnUfR3#G^bErEq(VIJ8Z6tZ~0>zmrL@UXfF8b&)!}?kN5*QU>*=<2*l5o9%O( z0_RVz*@NcJZ!UC;-&?teNDaCFM|84?^Gu}8P(Ob9YdscTH4r{}j8dKS^nNq^(`)CZ z?S7F=&oP{Ts><5Eqx2|Pk_ifLvG5CdFmld&+`0q>#BRAJi&%L#PjlJI%8QWU&^kDe zx^k4&P)L9!*$ibkpQ}+ql@_<3mj5u__d5@8H<9=&Lx6z_A%kXmy?(w&Seya#u}ix;?#CoZgfvDkomA3B zC^ux>5Zm|>fS~929}(gA(=ufGq58CdNe1;XEvL6U(bx9s^|2}-i#DCl{AxXKM##mf0q?iec4l)jY!iHcFqf?Fe?*jZ-V)&m ztEGCNSYZiItLn?U>fVT0b3KYxAodH;jT&mhRyQHf`Xea%<-iUo>E4M@fPL=J>1wJA z&+eq=3aty6ysb^S|JN?XTInwc47YYN_sMTP&g4_F)uL($xq z9@(remy~3nvwE7e*%L}00gd(5SSXmU)~Eb>B-TJxQyVkfa|6cNR3{==KeoVTyjHv) zTLQN6D z7mCw{h9^W^ZHd)}6*)?2v#1$vm=3W8goLM}QcmAAiz<%l#Ym`2kQbk3TnN()!74mT zW%IlkWKcYRLF&Hy-TZI+wI|t}l6Fv+IM1xsMQS5{RSoR0L0C$ac5@J*m7w$Ii$0GU z-qpBg2y*(cRWEh|`j%Ah==nzRg_8fEmJ)RqqUWxo$VRVSO73lI(L!iC@u$F|cD3mQ zxO>*O&5u2o=wr*Uq%)Z^+d!vE63CWkk-OiTJr~;tPJ?C}G2Tx4Hx|@$%;fV-r2F4F zw=LG}FMSDKx>_@nUIDP63bNMRWGEmxM^40LJLbNGrz)&7CLy$7}&%v4i@yed4v4K3kCKdw-Z9Wpu;gXIrxci3G62sCz9a~&cG8O_(Lu1`@L~QLWqh+Tv zEzg%VuOZX*SQgEy7X&ZIf6MKo>a|H!HtI4FWd7cWUfvFJzdBc33)=CC^nt25h7u&z2rO;TY#?_QimhH53fS`U%n)S-2b|r^<`Mm zmEqa!Bt&wYSot<~z1+;#TvQ`*$B_&zTNeZX9?3adUv;umB2}0!r!9I5|KQ;NUVCgMqSU3MZWWB)7**_ zUdhj60?*Xp1l_(&OCt=WTCXxYR-ZJyAW%h35w)hK?J{Hii~9b`~TjmL3f~(_RGauvlLO&(Nb4ryYWL3q^)trirR( z1hXzS-8!CuC7(fordA%mlVqLjzCEUX$U`K@balpYEK4BaNM>+zeYkU**d7b8$h`L^ z&M}HQEJ_nZwynYp3ab5;rPt-?cyC1tOd(9U!G>Fj3cUs78_0$gBKTR0!EX3x$l{bH z^nBN|2I=2a!>Pd#0<160A1swNOyRkZL$-4J>dSKb6meg_gr1pJ<>nZMS@l{;`~ zbJ2mr+0rEGqYM!H1*jq=KMdGyNn5$n+i*UyT2D3`ERiXDSgJBi6qq)z6o2~*T}N*| z`KWndH-%CA;$d#LAieZ6PQSOy)4NPsWT8HOn%)=#fiCyWV=$9@nMmuq-+64YX}n2T z)b_81{gj;-`+40nK3#?tKJBfUdOl8@tV}DF3U5_0W*kvJxA(e-Oqs&xBz^%j=z4oD z7tLe?5_ zAnpP+H))<6Vwa?zr%n{}({53+lLj(7X&#_~= zvWwOpz*9|&NrS2RwwB=^Zq}v5tj5xE?$f=LGzkDyQ+~AF-z-i6H(n2@edz#yQP$a5 zRqV5vPS6zj-8(pXv>L!+)h4=Q!E)fU(#ZXZk#EfDM5SqWyk;?@ET?N0J`Y|5qmPh? z;+R@IBc@ZMIcCuLoDppjo)&q89Scq)^PU=N4KX35lkffsRf!x4Ba0Jp#l+t|^W3D{ zTJjuj`o`|cElH3JVmrH2-7=NXnVvan|Xs-S%O|!V)s;_$f8GQSvh#0AE*ynfuX`q2GwV<$JaSbsv`+34 zbm3)lrGVf3ja(_>@5Dsky#-!ukxIz7%mg8{+^z-nt9-19n^XLw9;!T4W^;gjr&-Ft zKPp9Dpsg{Q5D+UyBGkJUbBkAK8gc`NQm62~V}tt($*Tf2?tcFJ7JHNbh@5jwta2;a z*Q}{o?wZcJOXB8!d0Lr6>1n!kx`8_Hemmq+>L6ot@O@NJV}TPJ!|PJkgQeBZ;H6bU zbob_Z*)`jS|C3+4OcsHlwO{l>1=c1-!h$>2RL# zP(uWYCcha(b@v)Lv2B7X$EW^Uu76{1C9o@IH6Ae{rn}+Y?U>{$&vA7vF|sp10c(Q9 zG$lgr>1=&u(z^@wF1dIRfNJ_q`o4g@Y*%N5oy-!OI#>XqO&qQ`7#-!n<^-{>f?99P zae$tcA*#=CLX|1DfBL*b!GiyAdOtF@+X#|>LEV=`x#0*quNi4+@IFqO>qm}}yA?l+ zQJ>HvAA%fO&^fi}kY%J@Lblv3mpGRvS0(~K7F|6E-K*=^4G=2$!Z_gx?_W8N8OT%< zQ{vS90#W$&7+vvuU*;eBov?qwhq~UrS9h*4J~9OoZcfWX|2gJu5B-ExM$#Tm9z{}g zrr1lMztga*za)?lPqrm*3uUaz|J`cGSPN1fOJqN=c`w_QXDPL};&fqG0Y(yzs%iA< zBIj?ZwzT)2=lDj1^IKE>W_o+;Dn)yihuxiNTD(MPMaTwTSPlQ9TZ!le6pzA6d@+26 zJxZyudCE|^c~MJIfVI#rlY_Y^7wKRU`DVF;)BBlwReqWgHem@&kv?}3L;J>kqcS!M zq7GMA+6a}_zSM=~Aw?541_HrcD_qpzNW;Z@Yu2?^xzu9*^9sxe{&ncgeXAYkd zF0qzWv=eA~Y%b(yU03^pA9u57>$%Aq&aCXY0hr!uKX?Aute{bU|6_{Z_=6LO!rAol zAERa@dks}az{DRKiUfQUayjl9Aojqh(?B>!>sAVrGoVUaP0ihJlDV9em(|qJVfPCs z8QI-IgQe#mf(o^AP#_B@vI69jY{vY!`RiT>{I|_~JgDt&_npADhMff72@wWPLCW;U z9j>2!*@OE+guh2+{Bz>V+$d7T$YA||tN;{EAHqPzt~c+?+{K{Y1EWY36FLk^|L$e;!nvN%_? zfLFgrJv;;ueR1@!*wOCII{u#>G1x;!Uk%+csg(+eHh+(T#B85Kc#fVTvODc2;&n#I zCSC@tSrVt+?b#_7#G77sO*P&%!$#JJ@t#9l;IJi2wc@L5&H_Fjshl7Re#aj8Bf78I-5SrOWW}}b$5TpjaZiEN^(UlGRVi{DuY}=u_uVE$jR1>#EKBT8 z2r(0f+7U@T6SVw9^(~o~#JOpykM4%mFEZ>BxvADjZCGCU|0Ym55JPL&fiOn@>+h#9 z$k8BM(@~#Y*Ifv~691_OKR>1JS?Z~nol2U{spfs9`xGzRc5=6dYtJoBJcR%aAA-SZPp4CQ$TK5JOB z;vm?^wQ0KcM_Rkc@4=dg^Cz5hZL+An;!icZaof8bSw@De%Qsqt5E_zM|LGe*rJ;6rbu%y>Tmo!=aBapnma?R6(J>UXEu`6gH>d)r(?>8LbfnC zcf{fu8Lg7j7BDK!fPC%2Ab~Et6%$P#Y7Bn_BYB_qVo~*TFj)5#tFW1xfR}S3XqxzY zIk;x?{_~RU(7!wgJmh|9j1s*8FnVTg9=-kGUvHfnBy5FJv96ZuAFTTIC1B68qE)RP zytuu69R+h2kCBDZb=|X+c9or^}n85K|iQYY+AWP<{GYsc%t{7hb(|tnCd`F+|kSj2~|Y90j4?`sc68csKninqQgj zoRh&A-|XN?;D_g{q^DM!1R4Au(RYh@UtDm4pb@i^yfa$`x}%hqch zUw0_nFmo^jn9vl5;o<38B!2uVnqX$TR-I`FTmw`{D~^w9O9`F@HUF0<(9lhiZQuWu z`l-UO&w3M%l^-DitdUwA(Az778KeG3RK|X=ap2(Ck&!x5O_)zruG%5>;mZ^+G1{mG zPGU^L*%)#^#W`WnHInD^J9v(3V8%HjSj{CXCK4OENd-1AlG4)kUNRw|nzi5A?f;Kx zccgl)ex}?rr+~pvXGPtnjWv|b@8O^l;0i&wuLrnaytr~g$U^n|EzkPEEjT&TWufUF8)uxraCF?1cNR_J>%8C!`bV5^>*I7N7l@p9n<>*Jc#HW|j*)0J zIt@yv9(N{(w)-jmc1QfNxd_T-q0(>vqTA=H8!ct7Yn4+k=m$jdX|r6K18y6G4GpEQ z3{Qm(rOy>Epe{aTS87?#OdrvId6%+wzBInLwRiYI-&l}`W1$|#W>@6|*nbBr=2~GB zYNot5Sl$WBS0~jukFIS)R^l8Nd*m4ZP(S=T?s!>U%g;x6rF94Y*qc#_NfV>~0xK>k5pUFk;= zglH|7l^zAF{6)ju#6Jq&$G>ew~Im%q#G<4m1nYjx=+hqiQcL2-M=$TCmY*k9uJSy?^LzC zNuhEznP$NRV9=-KIvX zin=dRuBrW#p^2GaV>V=l^$QD!F6$(|l4N25^MsG^jahqtTANa^)XU}7|3b`V;O3gk zNF7kBGbow0GvnB4n=oN(M}Lre$F08;PGBf9-7dPRMmoK;eG;YBV_mQ!Ku!h(^nop{9$hT7< z<4xg0%Wd2F{9(~wtcy(jKM($!h!y6kDS55WQl_56ONg`5ypzEaAvt^1?t#K_Ib9Yt z!0MjJ7qFQo_hq|=M8g@;AQgdMc~X?`UXuoBSf*=A(7Dcit9(7hL+60eZvbNM}W&2&zIOa~F>+iEW(>vZc!rBJFE!*Gs zlE2d)C~kgw0!KOyhOM8hZWMpAeS)&ov*EH%*otMO)NfASuibijjJ-xZ9#q7K-G+ExYJTNB>r8$El{lIp?K?)kYCLqKLpl zJ%%D>O{N+HhI4}tH^%-U2D(%3wiAlFYk6yE3qvwx4oOdnB z&^i^A1*C^r+%^Z1AanX|Q{%7hCNKD6a(u4onWoQ%dmb!o8NHNzGs46xX#eH`WPmdk zs*K^;>Ri7EvYiyi%GDm$-?TDXCW(@w22C33jlYie=m1^~FDRLBY@eQ(*NZU?!WUeZFjs1OZPXl^?s69>;o7*X>Ck`R3}{i!i3}~2KE}1K^q_b8oTjY z>ag+2Xx}Mf!o^049jiTVP9*~#ooOPqayOIJ3}(C3x2V8xQ`WVoYGurg!lnEJ6ybg; zq-Mj)gZAB7BtfC62#Gsc%r>Ec{S%1%JkYRoG>yxHznwm7iuKI*)*X6hG#A47jdguO zXt;i(^;+;Y`jQO4+B|xyrOd3^o zakcf|+`7tjUiUyWcQtXEN!HM}$ViMPvT?Jr4@?H^+&)9dl&U3Mo{1W-MX(OM&!gAxvB`C!r*+LjJE07n;=!{Yc?C7>wknuHdVh@4dOsChSsJMY7g20@ZETydlj{;wT<#8Z|20&6IqbXm-5{xZkY`zz+7fo!dQyf4?jY zf&^6!T(XMJ3a51h1C~aQPUm-hbQ&8JTmX7c4Za+{V#tweOboH)8=E+DS?)YswxP*| zxuvp#NHA&D2WtWtnhhO1`-o`4aqS_+FWK@6(Io>_xLCf(rVpJU*GQhA=_|dDOv3H! zdp0lSmJp6aoL4o0C*KH;`07KuC08^UWp^@!$}z_%Z5>zZYuP9b=Pkee?J(Cm2S^h^ zxBB^vi3DnOBOG*pJFoO;w2m%aLN)u1zVFHND_?Gp?cjs1bL0)&|ZJr&0F0+TQ1+?QTeQ-5}y< z_ByJ0kt?L(<>5n-;NN`RY`f|cDnt*UovIut;RY*ry~@KX)f3CYOWCO`d7>L3UO%aM zpz(ClTqI=EkPZ|4pdF=|y7=8NPcP9u-ObpY)YoZg^x=*am#xBkD?rS91CF^5d8wsX zeE|n#;LnmrBQR4)W#E1)875hJ1qdFbQ-1pbH+~W#!n^%w6&xITI9jwYp)}co&cnKJ zn3=fCk!0~%_~{&eQh=B1Z55Mi)L3;PiT4L3I|=hrd9IY)Yq&f_33>XH@F4QTjIY1F zT79AOtketbIQ_3dk>%V`#z^^}%ED}+;GJUU2iG8o)wgEH(7`^|-gSXzlLb+sC!vRM zhJN*cGqR;y!AHJ*3cG3@1~?Qd-Wq&^17;XhesH_d-kctx#Cq;nEU8K*_weIV2y&o$u)!|%f$xL6F6G+>;(c11l zk$M*aeb?Jbq_!kucfUa~;VaiA!-!CpRTov}yZ_DLYhbSHu}f-w#W5`WAG-bk7yoc4r`%u7NW9KNLx4}t8ZN17lmegA#E0n(RA6Eha&(^+Qv`Vix3#wjx10V} z^}7oZv|X9s-9Ld5tc^*8w@MZhqe2g8ZkrNwnmPSoD2YB=-AIt{kx(dF`ELGv`Hw~S zdxEH&-#fAcQ)ReFs3hrAO);>LivJPW>v)U)Et`M(u0Mjs++5ZolP0rde{v5d*kfF+ zP5;zL#c;MBOa|Ek=f@)ak!MC2E0ke>$NkmlpG^@)o13{j*)yYvXKBUY+`u-P7oRL1 zE8N5~$y}&p*44{pR9ah0Iqt;(dEb)B_a~D?zJk?p65O#^dZkUWRORqGPYxE8%rdaL zpM|mbX>_)H?*g;P>R#T~CbxHvXppCg9ooC5Iz!8}WBOnCwB`cvvKFLBlcEcD6a#A|;qYPl}Kv8+c z(%^_#;xB!B8F_CPuuOj^LUI88T7&qDliwK12-y1W#lxKr;Br#o4kaE0II+-jsh=_9 zzi*+-QoVRuSZq2APQz4y`{2Y@3yuP#lK&3*8WE?k@z~ch$}-g+w98bs!_T~H#sX=4 z)LiGa%Mlypvi%D#?gV^f%5dB|yUn(N>6wa3i1Jeyv6a)W)$SZ+)*YC`m z5N`%TJ((8jZyvuHQh6Hqd^-a!q93tL=)mtLv2vV=dxdWm9O5U zm`fV0XLy16pLG`YKOp&jwmblR3iOsKrdw9-DUH|pI@tX|^IAl;O#!Yz3Mc`Egw|m# zf+$C$-?wW0*Rs+`XYdAps%m}xH`tC8tAZ;mS@V>8knZ1s-pRKTO*;B%HpMH6QD27V{}y}iVmgvy4Z6C8HFzfZgxRmVE$~WaoD_S#3efi z*`qd;+u45YvvGEIZ|B{ep$kzWY70+UF1>|GSK||v-xrz;?N)g&Y)0tD*mma2*J+@G zk$gFQgspzgmrmukH^A)6%EVf@&fe>tuTo4(T5~;*XcQFw1{qb3=`S9+Dp?K@6gW>^ zP^~e_b`jM;YV`H%U@D(8U+zz`-&VJu$vM(-7ro>q3Q*)A>#=K+L#RQpZwWO2?Ln4k zT=eXZ>gsRY&lVR=N@5uJgRS* zKfuzicc^7b{OT?|_|s7c_dURUy2k?qB~5pHyno8W?w$G4*q@nR*9xA-A(Q*c^vL`z zs#b((CEw-({Pu)31;c(){Iy;0xt6f%EppO;YiGZ#JdJ)^VxE%gh{!wj zDfP@RNw%=D;hfZO9q^4_0ra%1z-x~h79WH^0;#niS7$gF+GqLZ|3RLp zOYe*+& zXr_R|J_*AO-yOlB+@p=>2bp<7NU;J^K*QeTZwngF@RMD(TE~geYMn-k+SF5uQmwkU z|L^P8z&-LH3FJi|o*h@;Me*KUIb?3~051|l&JV~jTKcqfRc2t4X-pX%%lRCTqLw#= z(e(eRjZS^En*3Vd%f~)Zh+m6Y_4P0{h2(?tHL2#sH_r^m><;YyHTF(0WIi^ldyM0`P-p$NJTXF{HWbEXR|cXNx=OwnX=sRL(aR*-mhY?z#DJ)zR^$AZsDd+I4wt|DdZDG~n`qJE<6uv29j=$t~K2MKr zQX$28X>5k2!!$HDx~BX{W2zwtM`D1IO_PnU`VQpp*>QUCirETjFKnecdmlHttImE?QS?itt(KAkqqJ+?3rn)Y04+ z#hb93g^pN3z9bE!79QUj46{!)#>{K}@&vG;OTks+A*{Lo^-E01r2~(`FDxYM&;LgR z40w5XqE-&h58^S9N-syQE>D->lQ&H@4TM?;oWZsgdiDubeF_M(9o~C1hN!A?13%$} zCZDj!Y_SP}B=a&X1%U?gK89iR*Iu&CEH%Vr3|HDc5w)ASTx9nvgGUZ0DV+dgymWqw zfO)n1bTX5O&-eaIS5WvLQ8+ULN*RQRYRyyo4o#LWz+}lx(P5`>hV@YL$nh|>#K>ny zU>0Qnud%zL&5pwoT)`c}xWr7Tu^y%Ea@qHhiQSMA!Q_s8PyR?TKW{61v(V6Kf4*Qu zOz?L47ZHeAV`%q0jg3@RcOO+Z{%P?iB*PYt*{gA!=>Eui&^}zg>9q&JlJW9mTn}g0 z*E%p8_5X-QI|8h&kL}$je4TPuK!AiB%N~C**)OKGgoG2Mc=#3PrXlko^sz7FI`Gq> z$cF_H+taPxp2bJYYTxXlT{s1KqWReZmwWhdP?)o?+6f%>AlSFF%hJx9?EB3ZBFFU# z;f78tmGD7ZE{tPXF3WEDRetQjm_KD=ihR#DM>BTB7AQ54?No-AdXRMA@4E!0su`&u zzfz-JonH>;R=^Ssg|1-jl_B;Y_IjY}{^!3E6!dWg8O=@Zk?zIQl`$|6*vB=QMfXj3 z9D$IZBwGfM6rmhCl=iPzjVj7Ip#`SD*|y4^hPUwGygSK_7owyvo2k>&H{@S!EXTc1 zNy_2~Ff1dQ$#dZ~C>{Qv8ZDGzZE%UPQu^&Z$xp9}x4wic{zeRXUj3IbU}-AK=B*Su z*>Z=`@|ciw<;AAMHF2pm9cP<_uP#2^UA}Mbum5eIE{XkN%FO0VO=NjZz}n9ZZJ+pqv~fwctc8zGGNF@K zR!yxPlRj_YX}rm>zb$^?fNg)KC5jSORe>2V#*K_tI_lNb?+6Sa6e+%4oCaMsV@aj-G$oSc3e(g_?a9| z_X}k)Y|n~{uPra$EYIyP;3)}fr80gpVDWDSCe;#&PW2ULG}Srv9Tg6;sqcBnq%brQ z*?dX!1-21~=Kxdo`EEnv<$TflFGrv1OmoME6nb+rh=zy`k&y0i!OJ7-v5Xkl20R5! z-LxJLCCbV-n0c2ont=ShWp*5ITCJ012F4kj;2;hW=)Vf0YMI#0(8^ z2=;Bceu2DGA2`Z%?6`1;_2d6-s!_FhKskz}sPqbT+?K;tSUe6xRcJ;bgbVNCQgcdX zAVQI@@yr)EKKIr;Nm%$Fk;|yO{|!O3=7}fYy`#ezNXN`?7NteB7T$i#dH1#0Xg)xv zco4d@RGH^bLaMOpS;<0H8=`0}ba*R0$vD8d64c0mJbu3xUicB6!nw9l`p=A$>nUcVEVkVSNJ z6oQ}GEfVZ+1mIb8ui5y~5GGmRR_i)4`S4sVO*L&1p{(y=;jADYXc6`&XTx3U|IUev zL4aLlj_1g|*YBQh$S*NyALrrY2{VRd7@3s;QjTDR2p|4WcuUY&bK2yhW7!~WY480Z z^l;o3ufy}1;Jw~1m@4lxk26bdM=~=QDgd2qhQ}bW!%L$715iP)zVykj^TBDM{(w)p_*?$~1<2L(=J-AE{{X^L z%8Pk%uHEV?jrY8ZYY8pc$G4et8lSCyHPF~vH~P!StAz!!AXBg{5MxA;oLSFzLRU2uH4Pa-z(Yq-2VVEd~V`MX1qmOtp#|-(O!1a z{{T1WQnxzyv4ay48ySB87Mn$?wNJVAxa_YpMRI%HQqmn=)JP%P6 zLP8a7txT_sHto%lNx=%camG8IMsv;w>T05~`1_+P^1a55GU{Gp8TJF3jchK174qA! zp{|vp@vM@UwzGKHoPi7hKqICA8T@nj*GcwIi)^6?m4O}hmilD&ABHQP;!^6q^QIU{ zHM9GVF1@s!Be#xKWZ!pa0mo2o6Q9d9yBkSHBSk6%H}vmys}$^ zd1CZUy>t19l!dQxyeIxpw zobXj6hNk5kE6e`?Z9hnSGva-B_EY_ft$rwYYgCg-@yEuCy+X#%SB<=joh!mRG>LC| zZEgw75^9&lVl&24L2L%bS5V?V9O_;+_)-4=1k2TQPm9+VseADg#x`Cd@SdBh%QM|B zoub*l_KvF@H!RabZE+(lvNqKdvEBh!-rq!i(;u|ze#$-=_$BcBz+M6Ht^WXu{5SBk z!#6fwEZ6NN)vl~pQjX$Tmcq_$wvb8Y#$;=TF7C_&Aqchj?c=W({7?O){vlm!{{Ry_ zL3iSR6Znzhu$KE!m(7}Aw79p4aG@4Bl*)ET5=Rnb>_y4X<@p8zzB(%~#6dzF?xU5{ zl$2!o61LGvM)!Am`>UC=`V&3O@VRW3T|#n{`4l-4vRuh?H{G{(yzQ=s>OcGy6U6#w z!`%nrAMBUmBjueF!JZYp*Q87<&o#A^L8Mzq%5u@#Bi-DrYzSHI7DdA-HNyVa-xKEW z{{ZX<`!swB)AEh2_)0$&YOV~!%CbBKT1vWv2djX3L~SLx>M zfAccFHtM$i8~*@;UHlr;SC`a0Tk#8C0eJy5_FX&w077rhy?gfK_=T%}&fm9Bfi*t^ zBwug%N#TzHT3bUe`mZ-hvKOWoz$I2ixhTW}${er+fn2BULExP~#6JRmWN(DJYdIRP zf~UK53dqdfFxR3X_WH0x*O6VX?RoGE;!o{&{{RIz_}$>2g5uXv)@*MtF06Fv?o!@s z(kJo@vq@W?05TAcmv`G?KAr#>RPvjJX7JTOP?2bo4{K7tkEr$ zw=CL)=8b-vyx8KFQzJlK!Ye9lfVv9uuZTJ&t+)IW3*oPWUBqATn*34mJ?z^~8KJV% z(?VubocxAd1MOXQlkhv@KkQHNH{w5tKWJSg{4L_YiJl*mz}_0rZ>5cO87wsmm}S=W zYm0c=LvL#Zyb(ce(z3L-4$H7E$8R3}rTjy?4P~vs#135UlLQYP3wId zZ5U!J(Zf@zkHpiBEKI3;>uu(>w@FD|=yX5uRV(|8@AxRc#eF1WZk?-HX_oPfe308{ z+HypL*aMI~{p;F1EuzeR4F1QTv!t+RQt{`+2l#_?6h%u7I`>(%(=@3Z1CTCs{spK?Mm9_(?HXuHuiS58g0C` z+HRR`Bcu6nTB{)quorP7>{BMy%MC*lh|6eUBSKM@9vbs5RVl`sf>CbCrrVb?(@5*9 zwO7Yw*bK_9VJgvoS(I&UZ40SMyDcwkTi11|^OxoE(QKo#f_4Owe$esDI^Y5$l>-G}MSQ)bd^xv&i#jKR zb!(;G=q@g{45~u130M={?SrpO9CxpO*S<4&CrMU<-d4YqsTjBAcXkEOA6!?2=$hE^ z2a7yq@fL0B);c^^a4tY0uGm7Hk3qbjZ^pT+(X9+6ECn^o8^$ZWlUKj0PUkf&bt<`y zT*{=mrD?a>UM(wnHoBhOt$5=5!(XzWkA5@FGA^s*-ADTx&7>G5A3KndkO!2ky@1_{ z{3G$C*Kz7HSs9#vXKsKU`?*#-_P0_U$vOA@zg&N9%l`oQLOvCIF!)7igvn>4$*)~T z@s>uG@tNZ;qp%Su=NvXWoc=ve2ix21%bkqsaypPOPSN?-^QVegs;!y(DEu+~cK-m$ zR8Q!y4Dy#lEcTB_KhysJwsJbJh8hGnlgV*@rZ_E=%$b;H0FZFJ0$7d*e`@XiCV1Cc z(Cv_TQp$UuEO}QKR!V=+M1$qGjl#Csih^a741@c$U_%<+@bXWl_;}n|g${*U0LnmV zpX7r)@&OGY_FsDK^e7?Gbkwki5!yh7?j>-ZV-^{h`G5n5B$1qvo~N4iyg%SailHpR zZ{N`WYIcL);0wSpP2HwImtX%>QLc)Ea=8nHJ^2-;I!9$d!L(P zc*h#*R7yVLdrI>4SH1MLmAy|Q@o$HdNWPjuZf#aJRJga2CxxJp{Gj=60gEm2yk`bC z=r+3J=Fc7Ii*0)Pj9}(h!(#vg$Ib6wp`Q@`A$Vuw7lh!tjyXTI^kUJpvICiPy}Kn> zJ;WSK6ESHOP^NZ-$HJZ^i+oaB-!2+0(Xl-@As~_I$j{K%^RFCm>&r6)=_^8K?>rmC zKVQrFYkYCjUH)e)ZK~ebOJQ|sa`x!>4;s5{Vc+*elq6$0#&gv1Uqs8`&jnn30gK}K zx4Qy6hPS)7wpf}Kl1qaWMlHl_SH^a;XFqfo3E&%E>g}n(Tc2;%lD=YLhL_pM5zhM(;iXNF#xQ80YzVn(}+yM?{a!ir(H+ zCP?|t&<{^?FnPsy);e{ysI@XIk~Pj*cMhW&-H%MyBXy%C&-#_X1e1bsp1pXlW|g9> zn>L%X_l_OqmR8&VQ`3z909ta|yp4{v)x)DHC65j7+x-6kpJ^qv0f8XldUV13`&UOg z?k9*{4rlC(fOFRz*JlTVT1XN`lWAoHgA3$jbTrQo>8q&gHnxaz819hzD*WGvO6=~n zB{wn{+4k-@;PlO5QxM|pj;uXCWp>a1)%$zkr^bKyBuB;(Z)xMJ4J%0TBv3^y-OSM~ zq^dAjJG>0=`O=wBnC#5wX*f_g_#a>KFN1EpZQ`W(Fwb$}2=(Y<(=BeNd#j6UJ6Q17 ze{E*Zm2~J@1D4tfjIMA$EB^p#FA{i{$G;zZU#;*vP!WKps{5&`jh_v1i1Kl7wsqUpTOthzNcZUX;$#WFJ`!qHOjV}2{yC1 zL`0fpk~8*kle8RUX1q#NAwr9&rF}Z;eu*mc=H<4XPw6Y-$H3nT>0bkW5BSSpj%yq5 z30pjE5N)bb{?zYlf`0dwpq!Jz>F9r%Pm6lb#Qy*Rd>--F!CiXk?W{FlgZ}^+Z?9~M z#=>XQJV|sk*_ntg3R~UFv?~BapD-j0e__A3FN3vji+`}E#P0?89{KIH3%?EDrk!^R znI7I`fy8nlKX|WTM-}{+d|dsjei?j3@Fbo%_}k$E_Bz+YJI@c@4J^m@+jiGsVSO^p z#Xr%exKbn-%19$)l6d>y3Mp5_({3^6N=r_u*|l$0+x1_6z;T{BjvJP>XUo5)==tOR z3P!Lu!;g+KPB3l$Hu!8?`h4C9wvUf*D<6g{^V{|^*R3=!h!N>m@S|GlsP>N|1!A(X z4;D9nyN2}ydRNzw5pAJr|6OBc8?pe zXx(MEM3eVeLANc57=!tLZGM_IohsQU*{*HwBAOXclF1}&2xF7fGJwDkG4Eg59v|g3 zMwfMvPGU$I1T#qsvxE6Ate(qV7?VV3ywo;?@ueZP8le(Jf zUKjACk;=A@rD}RV&5guq^La?e-J{yyLO8GIUMa>`!OJtk$-By%d-hzf^FMastR)Op zECwzyO-6Rvb$s8U`!T9qIUp`nHUJ;izH{-jTCf;c{;{d!lL{B+f3dnxp*c*3xWN#dq3Mb4(D=l92pEo;TMmOl>sF$qa#2#FZk3&|9rOEU*1K*4uAjHq%x zToK2&{6e~)N$~P%dIywYiq|ZnbwMW5qztJuZbGG)fsyZ;;=CoKc#Bl{&3)sWxY{i$ z%F(UF&V(5*BnYp!Ic6X!4q33g0qIpfA$XEesC|!3Qx&=cZ*E=Luu~&?0fh`VVB;kB zIIW#D;*pm=wmffGg8u+hRJfX9wlH5i9E>Ooi~FaM0U5#Imf-L+Ua_v}38Hu+&q|!f z9G2=PEi^ltUok@L+8A)FyK!X$AY@l3;NJ~Rd*T~i9b+z67e)Mn1FU3uYEI={%;yS0 zJaM-=S*vTQO+9V$sWnLni+KynvIq=cwe>LY3r^Qhw?1j;-KJdG+Pi zVn>tr`AJzrY(?q8!yI$b2N>ry_UDH@Eom)}*)Gu+aZEcAAVx(vZNDnvyO6^e6pfL8#h%lzD64umm2n}PK;?s^u5b$@pjfo!q8)XX7NiZbgS?hY~nD+T+z zn;H3=0=ryWmWL(NO&5svU6GQOFh2Zc`n`CUDvri}Ff1AugRoJdU%P=P&FdU5W zUOhJ%$xbUr`J3WuN}OdE@qUk~jq!Wo-jVTt;8%-26#PN8g6B%s^!Z@8j_To|P(mH7 zClSbD*g#vv*~(8^QHibAMfHz~Oofw^$S zMsPpXpH2HkwxJplNOO*OAmh`gKKMSyzs-aG3Wncm{ki@sYBGmJm&Dh5QPjpzsziMNjf8>s2hsXkEg+OUk8)fqYRK4f`a?w@;L`!3>SKDglh4S4s*9}q8w{uOF-hS@#h zt)w_OZPJ`E&uk`6YwHOZO%o)@vJ7Jk0seLQ_4{e-_OZ#Lcs|1kYxLe8_SC) zxY{KAYva63&t4BAuTeWnOO<-dPr+UMkI?)sz{@hvF!*+~Pnq_#b^KkI%=k}EGH;a= zy5}Ca>Hh%MsV?1^g@s55>-}n^E#%Cuc^&c32Bep9a-C1De+JT>(f)!;I^3k&hX7;i zP@MWzsRrN2nyc23mkO7e{YUgtz&g3*Nv@G!i@)f8 zhJMYr&1)}=EtOthJ4CUNq+qijChle6E>9k2dRN!qv$fP#dPcE7IG94Jn?NcH5=S`T zk>8Qcd}sS4c&k$IkH&pMJvL3Nap9ZsyK3)7hT_%HfjtWEY#!OKw0sG6Vz$)ggn-ku zPo6nDn6n>pNy$H6Yo~(e{{RV;)ARh>K4a+}hf(7kV_kV4{z((brSUbaf3=pM;p^sD z#NJxN!90VwDIksxNc^ku58}R?Z=!r$@lKA>xEA{Ms}S;?mkO7&00%r{(;V08SB{qM zf7&9?#c(?`uulssD-cKr0F+)2<&j^VKebmmD~g)A{Qm%rjp4E5S0859E!NKIg~IOK7U_3msKUD7jP}U@cdt9vybUaO8PG$` z!#9${_k(Hv9DV-)rZ;;U?R3pL@$Oy=qT%|5Cj^Z4J$iBVt9IA&Tgr6yEwH+-Rj^0R zl`2O<2Mb+ub2J; zd{*%$z3}ruw(yprad)5}lIKTReDAPoaRw4xqAY6PXLOLdZZ67>;g15o!m!h|4S!9$ z)htA4ntiL@&eA%fh>)s)nfvbS>an7N&QBZyNA(U>pHjuqSc-C1^*@tyo*l1^@* z?!T`+kE}mnAKH+9&p#E{!oDuL5cogEQ%7&97tYH|k@8yV3ywc@PnSPU(6J=qFc0+q z0B>_8vbCMU?0+*ZGmse-*Mr=A+#cfxy?+zUr(SCMgXwx*t6i~4Br&S*#8IwgXTZRX zN8e_`fIufCXZN4>T>ZYh585~D;-&54 zlHI0|s4}|;~;tf^T?#Xd2m6T2R8KsndDVq6L z$Di<9Z;Met>)_7;XnJuxB$v0hckDCJv}qlxegXdg8vHje@K1>GO(z(e;*z*{{Vu!{@h+9fW@Ww zj>AC7&UB3?+7<)QJ<+>4_pg$Cd;4Pk)?X2~FN{7Vc%JPT5K%Qr?GS_5rHQseJ7BGS z?p^d?&MG_NYeFl^?fyxAC&c4^k{KjD+Ax&9NiWR)?sZLnM)2$(XxFc8wD|kA-0@qF zu4HAe&2QRk{tCtWHGB=x?=vD--$mRz9H&g6TT;E zt*L8Tw4vs4`?N$+h@gfbNQwxQ?TR2y(UMC$JGTMlPbt{ti9B=GzM0az6E2rxEPu2e zt=(itw?=KdUmLd{xE062?co;|_UIz#40#Mj2>_l+!OkoC7Y~_X@buLh zFtk>QEfPrl12M>Pc*@ScCT+@EyZ$GR{22Hll}CGXYS{3~z0-TA8(($tpA8u;_}{P@o& z!5`VTR+B-xHk$tciB+^sE=BwCNpl-{_bdMZUhr>lPw|u;O?{Y+BREtUB;XHo(!ZKd z_$Z&pzxZ6Av?b?zu?jb-^<`CX?!c0WDe$#w5z^19pMhJ~lFY3-Y2N)Z%clq!rU45O3}s0N+omZX&@Le)!*RKP zy5m2tX?<$?+|F5ild_2ba;J81MLA>zPaQ{Eu=ZClj#qcmsL1i#HYOoC11-i$rO(Z| zo21s}M!EA%q}wnJ=IVC%&rBlAA@=CCvEiP;`9Pb@)FHaYnR_Ja85@PosCCHQCH zuY}r7`}{KS7l~~AMRz5{Fh{Fvx6s_c_d1QvcC0c5Lw=#g=f~DOUHVDMMR9+m<}ny{waayIs32k4G`$ zYM#MLoFfUVURHN}wzcls*2?xu?D^C7nD~&Nv=_#2h#wT*DQzP0uCJ?KBvO=Vpo37B zLlkbvN;Ag`Dw3qG<=~NnT^H>^`$G7G_MQ0Qt9)_s1%V6~f6y7e@?xnfbF0Q1%n^0#f zboWxRcc0Botg^GQV(LN3IK@*-IBj`Tf#^;KewC|lX>oAjB8{It0;xIttMl6PqfYWx ziuo-&dLN)+sA6i_Nx3~QySwk|aT@1^?0iLiF0Ze7m)Dcv%X4)!liWyDZC&$9%_CrN z12F(_Fe{jo;fICg3es8wx1hgyL5M^E+5+&PdI8XW74!wa!F$`Z^6u^yHuVdhKR@TibXtGFttJc$ z`<9dsPn7rUGmp>kuaV4rZpdk^Yq>up-{zA1PolthH=RyO)RY&SOY+z7K5s)Tcq+gH z$2@%i@AwL{4Y--Sv(dA07>{u;B7Iy*zHWtdNl6ReK;fm?ZtIgo)_@$ zogb5BX%r(Q{E7w-dibn=(jP63n0iu|t(WTm03+?O&*O_wdFRd!;dj zJF?)8LF>=byi@j2(yfzB&~*u0DRrvdYBH0_*&WQc1bzt?E9YzY`@~fx8S}+7^!Hb% zTlYPB9tdO+r6j4nRiC`|S5M2a`APo(1tRe+(*D(cA+^%o{GA#L4LTN91r=ib*OXVO z3*?vo0D+#B`KUW}4KpEGf3}@d1x$lblH};M3dGVLVFB$lz`BF!HYs!Uz`NIoj zX*YH&zE>c6lV3e({#4@e)F~d>wbsW;qVmj@lqzeTPBQ zq$A5oFyzkVh-7v7nSkVBjc@#MOYakF%RSw;-F7(Rezzov3r zhZ#o)C{nMasihe0%K1{h{qF9Pdb9cCI?wR07Y9;>+$uRab&_{!e6F{BG(6YBI`*e; z;V86avVtpJPfv|5eCdijATuhULZ3PE-{vg9WM=^IJq!C1#w#zf&RCtMHDn6J24lkT za0teF;|FhCS3BYT6I{}@DE|QAB4ae{k33shnmDcG7zv%KHz{62ZuUIap=rWvi08U^ z5^(`;}u;dI5=RbYIcq~jXN;u}6{{XMh{L76qd_1wPSS8C1{a@b5 zd@&uKm*R=+br|E#^DUf_uJWeFMKK-;BOn$ekC<=|0Au0rih6a&hx}uxcym(uu5D)7 zE&&70LFGtKpf4L9I+N2C_OFQa-x7HRv};25j)j%{u_oaYt9e0k7_^xy@-T1Qc%c=T)_e{_Rk0Sgs~%HzJIZ6yPvZmr=tLn*6)Q3><#13B@mVUz$Hy z@M|-a>rGAkaohYh{O)}2Jjc^7BUshQRsme#Hgo>~>-=l5w$t?;U%=WHnR5PJtPmB2 zm=n4enn~m*Cn25W=Z;1=u6s(;Wz?;vvVufLbGOW8;1URDJ-%K^&Ooni@YbiGYc}6! zx0FezL<~c6jznajRT(~+-RMPsJ>ntK$tc{r2>Az3n;!Yd?_Mb) z>NeNs?GU4>IV~G-19EUaY!T01D@qkywL0p>C1XDRB!bZS#|22h;F5jMQ}pj#bpHT9 z?#U!)Bn)>wKEM5H+VLi^t1yn`t#rOO)uJLh zdtw5{2*;oU3&0=b`d5;jjCVU6o=*p_;Azm4)>7&^j$aHOi2v64yWwa27Ng)7z^Mm= z{5#=C@e1hn*I#Rx+j(*s(-9b;d!#H`n~2f{5U3kL`Lchajbrvy_<8$Hcq2pjm*MRq zUjTTAQPMu%NVR81dG_Kvs0v2|Y%RUzgcj@4zlhKHBG2rl;{O1Nel*a2E$h%}dY^{$ zjZj@$*iR{0=D*Y5%!p047ijIHkO_AU%3Tp~8;bt`Kre;fG|~Pd>YDGu?NUiJyPF#b z-4fr-jzxy^b4#i(0JBas7>|U1`@6 zMvAhOSr$mhm=el{LdeU&=dTs|_l23wOs5|VOZ#mvEqeZ+GxP5ea%kpwDbQQ4z3hK3 z-wXc$WZxEiLb<=yyj^`gj)A6mh-HkQHs*N9EgVYUFP5hv||&ok%ExVbjgNfP(sHFkhlc4Nf_0Ith6mT8p0nU-rWL05C9FTHm?Bmz~kGV_5E)? zF~m!jaF3C8+8U!VCOlm7s-KCPp8&q&uk2l%W!npcRe&CDnc z9VUwfnl{1NBH?61kG!YXpVwdbCuhSAAK^cVWAKKPcXNJ`!Kht5#n+Z3l3S1&$yk*U zRaI3$QSNe0en$Am;pLaZPm7m2-1m3-b*!^B#e|k6QQyu!V=6p~V*)d|SqA1H1HkM0 ztN#E5&ha;kJa_R^#v1pCH59zE(X^-{wTd{BJ+FB)hi7F~F8*9XbI;Fs-Irl! zmSS^?R@R-H>#dc)uQS*1dmV{}I*wOSyD8q@m)GuJk^0=tn~5%D`>_H~rxoJ50kzaM z?NV|5RwecS6YrQ7@ zk|b$SSSZ`LtGfg&cM@2y2cFy#IIno|H1gbAxe};cZowqvj(}s^>yFs#UpV-(z-kTl z2@)xXCP0NgRV$J6J~`dhKqnsjkzbh>;;pI1X;USjTA;QYx(5q?Tlf21o`galYh~JUSTTQhG72 zw!0kOh2Mo5U9r;cd_^RGy1UyHPr69tbiph~?=Vn5D4=9*>w+sU#oiCTzmc!*r~k&M@<-pDRs-jL6=aHu7S#~JPWpxab|)ZxJ+v2H7uBkePk zSW)CG`DjO%B!u$AD#W^}UN?D)GC=@zQA|!(jf|}uMs-Ak$(6wyBOwVVKiv!q5sor; zob=8?ud6?2{{RB&J|6LwJXJNcQz|dn(51{#!tOFW#%Xu2n4}Lelwgp*#53Xk42r>I zHx{A~w8IPUl2Y?T&RKk{+gjwbpWQ0!!28M$E8RXD>e?oS;>XounZ%LJ7)Wec;DMvq zJojl`B{^Tc<*@~FISjjptwNIIyEtVgosa1I;4~f^*E}C<ulyC;!}gy6K0e**QR&iYwpw?J;dr5yR2ZPV9XMuPo8->mc7S?u zivG^N1zd@AJ4?7hvo4-wku}0?RO*hQRzSouo%@^REO1ETzm0$REuOOv#jhUt?@ocz zb8~uNhDfB0hnC(l0}X*!91Jq3t)ou$5rzg$N=+ZFfAC7*iZ+w@-$;hmD5swJQr-y} zfQ|_Zh9nRM<=P682t6@h(k7{GXW`!*N2tG;S5splk7`a4z{BC>`ND=!KvVMPIL2%I z_kZA&U$wQ}wYP_@HOt4hzqXnH#h8#Lh>khhq~UY4a8Dd{ukV}SuDu7w?}V{qA20CcbP zFz|+@Z5(#7##jhblwl_1IU@rk_sBUsj`jSxf5FZl0IZ`v1pTBmn8Ms@Nj8P37Hn@8 z==V0oRODqxmOxaQW`P6>pjR+2htjEM`Go5n961UoHOt z2HE_3gx?@sW08|rbj4V;cqbhTF#N$4jAg*Yef!q6rP3Mgu1GnI<;QLa=D*a2v_H+M zDI?H<=Od@RK{B7bD&b2L(n6(k(`kN&+?x|TQ)Nc}6BGoH4@F(&yHROFL^pVGe` zziH1B4-$N3*6txcySKRe9C$e!El@;od*(pr+co;raensxbsptF<6 zoC0u8IL;}`_bC)v9Wd@bQ_%LL?snJgpNM~De}w)w{f94nL-5Z|SiDo=9BPx?5T@$- z0maNxGKMS>GbCy-0bC9%@wI_WrgM&%?kmlfAvv|x9S#j8a>benZs9@Q)N%E%(a(l* z3pjVbb;doup4?aGpN64Hwd+XaATOGyC)=OtU!Xo7xsKON{q7Y&&UhHl{{UXUGw~I< zC_~i#()uG)g(+fjGj`Q4xi|WsPX5j}cNZTYuPrWHD114nk5B_OwhjkgK;!z?=q8ga zT0euYubGw1k}PQ;JxcS19-y3H{Y8FUe#v^H-Tu%zl&;&1dIjJZ(taevkOe zP>)XVd{8T5IN9WEFAT&k033mX_}9>Ig$E425or8Z$82iE@kT1NcDno$JgZN&v((5VmsN+)5na2v$>%~rxlizt19AFfcCRq;W{7lsPwh6N5zI(s>J|EsJ7IwYdsn63#*sF5Zrj3Rr8_blt)cHRm>hY9^YRM)w(z7r6`R9TXgUq=_TkdZzhk>%Vx3@- zT02F$gd-|Q2+4G8NWUlwNeleccs9~aLPbDINRC;sNM%#BZph~$xZ|AT>tE1U?CoXa z&x-mkmp+FGxtM9QY7WpUtn-UyNKK$vlu0a!BQrx8Rt3Ib%EagN&H&;Cg~Un|smm8@ z-Rau@0Mq=B%=|*gC5odOkfx<2ewzOPU*vsz<6ns$3er9v>c0>E5a{x0QFww#7g)Kv zCR>KKv{=k`X(sml+|#SgH$@?YU=flzpMep#jdk5;;EaJJZysqPYz|p%zmm+_9+pqDiBV$#UGryVPNrk^K{v;HW}`snB=Ti8kK$vip(z_m9M4UM3Z4E>x|uvP)N`yX*624Y|{`3wZSlRQ=7uEU}gx?2$t* zN6^=l{7tj4TS(?WsNj$n7$-l6KhnQRJbn8n{B+g6F?g>__+#MRTUgSQTD_9e<`@K4 z#@6O$xwjEbaM47_kxGa*u>)ydszo^e0Kq8#0Bj!}-a~b9X`*<2Fj1dT)YaWly5b9H zE(Y&$AU%N{Xono-I7zCQoJ?cZ&Px9PcK-l@(~aO<(-}fk@Ys1u{drx!{l63OerWyr z)_jp;oLBTG;UD-RcB?JIOXL3liB{0Y2h9$XrO56+lU>|w{{YBU^>2ay0N{(ivOkB9 zo8qlU#J(kyApZLLI|TX#)8d7(^?du+pPBt3xML2Rg;ziD#s0Qmh_aufw-MF1wTxiA z}4cyvk52;^7P|Q_;An*V@obWT-+P&9H(C&4~L{<+JZ0H?UN5N$|+G9`) z$-ij}u~-r@#eY&B2K|CRV=n@T07)=S5f=DxGgd_SOgdL*>) zzMX5QfC3`gK^$BTiZ&=X^cDF2bM%kHPNSX{gq{BYnWgwHWA?0rz)G~KDtMa7TeaWk zWB9i54y$3{dr@lfTu-T^1(GY63=0$B9iBvk%q`A$Nwi=B5UJ-^;k>>v_^;wiT@y&P zYn@|PypGDz3ZLHGTu38`9ixQYL1G3K0XzX;XMe#({v`N2;=k=*tm=Lc@~&;PYfEhs zd!)k)tGgDM&mdut!Zwr040$Fd>=wSF{{Vtbd=LKsgs0*TrSbDyfc=BQs%&)a%ksXZ zbrxihWO8lop8$i6(~x;yKT^$j#u!c?d9`?Q&y_nj_L|Z?I@_kl=h=S=)WqkX+4QPj z3QcIbE#KU`pSGU}J{M@81wI$(UkW@n!b@EW(QVAQ0M9Y`r=79Yrj)y%t$J`yGmdH$ z8R#)qz&Pv8f1j9)MC(>!4L19mEJ8W*&)O%7hwX1REKdv4YOUosV{AJhL(6l5$BT@@w>q z{t9RD?S3NscK-l{U8Be((fkmm*~o8}O-f)I?fPT)DY6wg$Ub1jes@_;mr{9Fp`E=H z5`U4e?T-hzYx^%KRch%vzVr0^EAH3Tv-umvZ1!1}Xz0<_H2u}<_MhLU$q>h6=-JI^ zM;i~l(zR_gjbiv_-qusj)0HdRIr+Bn?_2i%2JsZiOoAl(u|LRH-D9}REW>{v5e5GM zZ!4bun<`4=T4CI zcw!Pa&=q$L&x`11%3Yjm3+-_7&vN|$BDm`V<-68 zFU0k~;u+`jvZq0Mxg@^}zrf^mpNhT!(V3u{^h+va6^AkqM{I$N{x#-0&&Qn{KbNKH zhTtBpaKQokFDD;b`o~HAkh}$E(Va&VZ0ZSTQ_g>g1EPX0CV4z+h_Hj`|T z%(4>_K&<7MLon^h;P&FX548T#8pIndY-f!WbySW#`~1z@*l}N{8t;d^KNpLnYCP%_sU;{MFIFWIN}O8_4ciy6utHF^;_KQb*%n^WhJG*IpRJ z7S_t^Z*QBaU1LGj+=TU@yvfz^i|-Fc;OHCqFYX+e>>thu4pZevju+S;C_l{azU0jU>How6mlmg6ImFsu13 zAUWTNUlzOx@Iywn@#n;i7gg~NjUDV3*H#)O&9$Zci*QQ*=4-IfDN+MA{AED_fx)jp z_<8>T1sneWf?oK9c{=C7e~A_wK>q+qpIoxL%Mu3>+v;(G6l9(TK;UH8$se{X{syx6 z2dhou+l05cM~U>f)_t!%`k&oLv;tT%mo71dP*{<-HTg}Vc)ek@MzexBQYRsCIa8JT z1K;qkOASIYh1qoHXQ}-&P5Wv80KpD*8RNF`)t`_2Hf6|>mq^vFQ5AAG2{h?$)MS6W zz;(f|Pxx`+FZd!?#C;0K;m_=^@rPfq)uyzT!%VWfx4XQUHN(#s2_;VE8Nch4`KOL+V~7{g?bJH;T2t2I+S? zox&Ie#iTD3W>_ucJCpq)>4qVQNUbN{`>UGLbYm5Bwv?3HvG~X1i)m!m{7xs3&yN&p zP>e!G-Ip^UC4nFUr~qDh9Zh*wm?l>-OmcUIIQ!THWMkWe&tdJ)It zUu{BCO3zc|>qYankG(zz`1`^i4!lESs`%de%S}y3!xoXo_s)7xB5k|t#!7XXYO zZq@F88T9vzL*gwX3n`F8xS~9MTg2poxF}XZjyfE0Yx7sf-T-fj{{Rm(U+lZ6t!^|+ z>zh|;Lo&6!_*s(6zbVfs_~f;aK20ayAj3Yw;`NKDT46d}i?; zxu+mDe{PwfKg1&}N)mhI5P9jGbBg^%p5Ffe;I*Ca#czueOp;z&Po>$|M>Cw-iqf@UmfU^Xhk2dUm4|TTf-=>K4Z8H2YHdOK}Q^k7djyw_D0d z^55nt_{Ch)rMR%v?=)wyiWn{TNgURV^6h6odL{!7G8mD;=e=q8T05@`cw*jd@)$KM zyL6d`HWW?MpeKWYAO(Al54B?@$J#Duyb^Li`>n{scRcgo)9|mwT)E`mL-%?R<$_4( z{8^}8T-`{avkJJ3GV-bj<8a1QAM$W2jpv0mYfl2%d_$V**w|`U652!M$%#maKKXXz z8$m)o_C{2Zjz$i@sM%TFN>$nJT<-!bsl#w`NAu}kSFY*$-`S#$ZIQ=vvP?z};GcPk z$slKp?E@#1>N_aH?Ii3)N~vgbM^KAZ)h5(Du%cNDFHD643On|%T(bj9(nRtG%-kc8 zISRak&j+dR?Os*kZ4!N6>L{W+06UoX+t)u)j-I%#!^EB;d&}mv(^R}L#ucy#!90A~ z$v*hUKT7nH=8B2W87Q{d$6nmas$G(q`T7ChKF81s*d7v*fMuCLJo3yB;yYE{D@ub+ zjc315(y$vwasgqUMrf98Ov=R?MFgAxxd3}}S-mz0%d0>C*Z5!d4fw&M{1&v+?LHs) z>q*kAY%Za-x3ev8G;I@)DrqhEZ7(QI#D;cNKRXV``n~;`H9JcUD)4C@Mb(a?eCs`1 zQ4K7zzmggqoAz7Cr;spIXN|b#zs|Se{{Z|G{{Z3?3w@<%zBskfwd;6{c2WI`=|0sY zt^3G+)eA9<0s-B%lfM`>`yZqK0Kq{&Z7{StJu%+l$GpWVW?N zw^x}Nx5i1sd692oKo5Xa{&!y?_SJc?wJ-P|rqj%%q03r4r|);hpZFw0U|VYuK(YUe@1~h;0OPE#AKQ z{{Vu3ctb(>VezW+Uk7TVS+Mafyc(sA>axgWiggMJBlv(`ea=r>_%lh-F0VC9a0l** zX36ME44-}u2fcq_aQ-?q@UX=H05VMeTjE?SsN$!GT^aEo?M>s$Z}>zT;mSf?e%}%q zvz$B;nIs)@7_$W!{uOMKk&o10_$9mA>i+-&H6IhjZ5^bqJ-vkK6l)r^){&6%4ZODE zbCciRzApW#w9kZo0Ptj=6TT~3Pc5mE-rLEt$DcLBHptxLosX^#v3>o{S^NI zf_7@21=qi4Sv8M_U-AUFz8_`<8a6~oRZ2QH%p^w)t{a`%@Lh}C&DWf$wW{5kaRi!SC#mPk(3h%Q?=1mxgj>(6TQUxy78wp#7=!Wi>@ z>m+y$u%hiaDtaHi(<43l*P?ih(jsPlJhHAvag2J7y|bUfxgP*P@!IRG?xjwkWiK0U z7<2=!I`PQo2aM*wfadkOKd@<}pkwOg($uKIj!fipyN{cyB%pKFiMk3RK^%`uA!K=RGp-i9S(YA4^duKrJHMb3<)6$<+ov!g2RLI zfCwWTj)ZzwbSd9Nb6s0Q)$JjDI@ipXWvjGgu`T7WZu^~sbJTFY*zZ_VX{S=Wl1RX4 z0^ldiS($$4?{_H*4tUReh83-)%(syw(6LDq_iE^%I7I_7?I4neg23c=1P~2Jb0z#X zw=X5Rc398Px!h!vgBVDL9Wj!!IO+f_Gfior+Os^m=I#sKOkK)`NS|WJgba~_q%%p* z&@cxh7$YYqKU91W@m`Z<<;f-r4e58fY}-zu$Cs4_LJ*++!g}F`IUgx#nq}Ou+z3&i z#TO5@Vz0c^2;U~oX1Hr!ubsrGjY6))^jRosCWqZWf5nTew^Cy`RBMv1X zf#RlMHA@S*BV={J0c6jaBfMGLg|1$=pKw=D*9M!M_nd;ZS^6{?L*_R^IMdE|pII z0Nwed79~arQa22B#~o|?xBmcwY5a1&@WSZ6HPG%Ob(RGx%_qwwQy^Ey2?a{$0F&79 zTzpGHbZbKqYbJVlc+#sG;iDw<{C~{;#toZkdaK#lG;Iv4=0g;CQMhvCbx;@(epSgi zBc3bs1O5wr@X{}gpYTnu7Hc-}quh8lYi({e3e4VJyM?$_z|LAHVUzOuk<@;wcsBmS z#XbYlyi2CURx<*H68`|KO}uh@bmu*5N?n`h#D*Q&daBg33Vnu$se%V^s(|!rPiO&5)QMH5)Fa%qsIsX8@Kgh4h7HoXu zu1^NPk34|OMehWbKw>6&Dy)k zYjkI}aradvTuO(Yg_-d-Oljwf;pwP_@}LHSMg4`@mi_p{U1$my2=&Hsn*1A!d`QN}89Xi(sKzeW z>VH7+Uxv6kxL9KGNy*7x*Rt~4{Lkb!_V4|i{{Y~jpYTeJLrc)LIJK_{_*=yGL}~XV zSji3JM;I3>vO=>QsyeZW$~vhe0;PZlVksT;L<^AOdsi)8Ep*7XJW(ZCM*QymxX2N+y{RBa9u_40GGlJ^r2k!T6P^ zH1-T-E7t&w@Cd;p)B4xPW2tgilwWhyrB|8?UC%f8D|ZFQj3T&?ECE>wQ_ybXp84m$ zt$joLdHf&nU&Sjg1nV9rvc0;L>9>|kHJp(vAc5nQ1dV_Aachp^n!k<`;wnz{qTnG7dtM zjPyTW;q6Lr&Zw%fQj2X_`KG^9PyXww2(9AXDi9b9r2F*SI6HH z{0E|F8kAa=l8pz4{71URbW&K`EyFItrMmeSCOe26fL14E11rz$UM2W-LKGe+2)ymv z^|AQ3haMy1E7khGJL3Lj_Pn1<{(pV`S=1v-n3g_)Nar4(tyI!PZ*rE5j2VGa-^4%} z;C@_ot1%_a626_OFOzd_?+nUM+?qBiB9BlVSbK`eR*%e)5pl44?HK9X@b|CgzH27C zAJ8t+e8yZcHj(06#7ObxG9Aj<8+x+?#zx`E>7Jgo`-%Gr{8PK|PM=||S}U{`7q>RE zO$4SJg@WQJ?njXS04~W_YXyHmSH9u+qj|m?_aFe%2*R8b&j*voxgx()e__8ACx`DZ zZYR@X)$Q$V?Qd>vnU3+~OmGGc?#L*r4JJYB*0*)zFA!=P_P-jN z&1P98xDlukDU%E3qLd18@kPgt{{Uw119+bCyvv)N z62#p6{$?jg7AZ#ljq}EGM+D~;^g0r&v>b}~yY>T?Nk3(60tsLi;@a<2w2mN4Y>M9c zVGtzpNJbp@>+C+(j4Lk#HT-qPb*W1<#7c~>7bj-5PgDB8g`qWu)NV<^ z1Ddl5bAY0t-MHaLHLGHnAc5D9t$eRCXV66&P{hd}C#`%L`*{A-m;V3;d;>4T--j0j zc-O=b6^dBO?$a#g%C*@CAWGmAwgexWG7wcTdk@-w_Qdd?z^f_zDdAryS@Dyu)wIs| zEJ+y>-jyQ+4psm|zZeE!`;cNUA~`y=*iASd0)qmiq_u* zz7lIc5oT3;n@EIKF)_!RdokH=pFrO`9?GQGu165>pL1W2U+_?miP8A4;tij|eLeSC zcx8O+t*ygq7YWL?I+$P~N4$b4_GaNx!SeJW9Ps7U3G*wK0p^Z}i%-xc_+;p=F?2sFK7fA|oR!XSSq zBDc2GF?2IeMUg_IW#>0 z>t?;Q1bLDu_Q=g+YI@_@SlnFck*&?bLnFf$;5^3)b|!mnYz+3UW;r}HZ0zQE*n!C3 zl0N~-{Hpi%rLU9)ta7|ye+*|In_`n?!*lFk*~?Mz$HUJFN#Y@=S!$Z)+yos$`WPKTA5U5XC8K1A4d|&Y&gLTgo!*k#}3u_=DBK55Tq!P|sJCP#-H7YV3 z$HNji$T|7KOT?CQD6&Gwt1&?5r&0q}j5>|`a2h{Py3BF)%8ol%;+z-ZRx2#dDrWe+ z#;tbwl8l>CN>8f0PTee(yC1W0uZkE<&n$*pgv;sCw=LwgZp)*xvVC=DW-cX8cm($U z063{0;$6GR>FN2`6yokjUDpz4JcSCMmH{L4r#uia9&5CUJ4q`s_Q)jv04n{b+9UC= zW^^}tgDMnzgpb_d5&HiCO350DMjl*(=#%#VVt+hi_5CWidn-70A1x%?k&%Lb5Gt}Q z)y5};7TR%~WO6gu5lftv2Ij?~HPy2oREmiWg9V1lK11MFRQ3b3=Xu-&7 z3-Q!^p1C0P@m*K`3Voni>vqpRkEGtiYTKu>pH#Vp=TY*oGuuR30zuB`<96eWWSZ=T zPm7wmiO)P;2=33MzilY22ZFR+J{V#}ePB6Yik~R7pP!U+0VBUAzC`eTouUO*xwM8| zsk8=(2;cqV^5C3)rn|q2fACNrg@3ib!~H+u*MPnf={^O&@aCYJmDaaquj=}}#mq9n zG%IXk)NP_gLhK}GM=T6sc7RQB9v_Y5w|0c?V8PhqfLLc4{4-vJXw;=mLZ*zUILcn` z#utrzG2vvg6KVEV_U)G3BaCIZ@6WG)!oNyC;E^A-AMGpQKaM&l?AxLE=T+0Z0OIcP zygzI9YimP1k|d8KNe#MP$n35j=XyItGDi(k*=3^KN@^xE}t*O&l<&XBtYD2dfKZ>S&F~* zoy#uK-H&fRVHEHd3No#*j!Rc#=__o7`Q`o_f8cK25Ftp@`hu#gN`htl-5w3U?GqK6uGt8vu6K>f9g3^>X|)@G@%K z#l4mHT|HL6GxHuU;A%OIXyTfBTbV8Hx~=+X^m{8V{{W}H2l&t48mELV;qsOhyg$YZ(8Gzi@pbZ0@Y{M@3o6b^&hgy5wNp#y1lqURI555 z5yH)}L6$~H#yT9==gas?=3K_V5G9hqheH&$QfxiExb2yM!8zF6_O3zfd@JFvGSkA+ z3GPsr`$XUf(c6(C#_aDbtNbrE6?h^tMSabFHNw%P+o+$WxAa$b`LpCOTmx36Z-!Rs zqe(9|)NkmbZhp`|89Z_EZ^SLD_^H0mWQa!&nwj#n>pZNHzzZm5ns&t34!{@0VU<8t zPZZJg=hXEZT`tj1U*>Wce5wfC$2@V;s%uji=X=Y^JlEaJ5=mflmM5upCz5lV*RuZ4 zpA0@Bd{gn(k?`l??V{XxpT@WFXxc1t6im-_;Srgia9F_`+~DC6n`>if{98HAKCx0* znd$m@AGPHeKV{2?g4pPFm9V$8)3tkB>uc4L*5>Ez@cEzUuCq&T9C9-%DVGtme(bnV zLX*2HJc7mUV$D?}0bmd~SwXu0GuELT?#(Iu3mLt?-fnHki>#ari24F4YKF--VKAT1>z41Jb z8XYP{`@~5m<~z2KK<&#AJ6JE8mGwJ#OZIPK8X*oQq@Su*wj-AGOSFcHXzNZa& zUD+BDJYF5}ay_XL%X*tvpe?u!&Di}bJ44e_%S%b8;#WJl>9_p#sHF3}QLec!`-r|o z%AJY_A1LD|Kh*kF?QZ@kBuzRJHiatK0|aDq>GiJtm5e1Tl{Jk?V_zx~B>wXpFyk4) z_Rmq?xWg=pN}`+&2h;PdoBNA-WLs6ncwON1&sN4c0CwV!>}!t9qnuJ^Q<^BZ4G%ky+U7QWGtNJ`<#V$zQXNj=V{% z_t3+e^f~OI)P{>S%r-Vo<(7il9-=1k0I<-LR&gDWxZLtDm-bLmoQ83NBO`XxlE}1m(i}_(Dcnb=@$0Z>h|vh_U=(9NLE5nH}R}(*PI+z?@#;_AN~s8@D>jh z!Qrh+$)=wALve4ac^_pE&lLO&yX&F#_AG=tvGjk7{u+El`0wGZTf{aR3TW592ZQZ$X)wTHc(m(@$eGdx9%&Ghk{jj#eJkTV zQuJEuX3tJyqT6fPZqJh%{^B-loNZXX(IC$mQH+ycuOIMA`BLNdV%09RwJ~Xb z@R8ix>9KsHWwo?XLO26CW(@7SCvzUy75U@)M0^-wMj1b7rs}c)Yq-7^4=$dcsHs9Ql^8Om-7`SlYWlipHQ@# zSl@iC1G5%g#AlqIa(+?i+ov_i>ZtL?vD`^1l12nEP@t*kHo9&j*8mK0jz)ck^k5l$ z!U)gt705j}QO{1eV=BCWrPJ<=?ge=#Jb-hYazL)CP}Ir0o;$5M%yG!R zQH_;BVC|1Fik-@VKwg`&a1`~ zYuKd=Eex0=n1Q8D>uRaoUKv3#(EvHtS9#0TI{{{UW|q~_+%JF}U*r`^LMxev!1 zwsDcq$^pl0RL=~raE(W?(&#a@p^8Zx%92%GvZG2$@;0_N1Z;z}{pH7AMSY|DC2Jae zr^LNGMYNU7bE+tN_M- ze2jI#86&4sdkr7NT8D*iW1mNwSVU|#!vQg_;5!J}SQ4a+j!5}WY-LTgAey>p{bm0E zf}8lIuD%|4Z$Z>7podfrWlgiiaPh_!(=8M(UKenn63Aj;za)jQ5A&Vz9@IRco-zvK zaVK*z$ON4C1M%tBzWDf=@iz0~Bo?iGZvmFhAsa^=O#~6E5Ls3r%eCBs(mJjrU>gC0 zSH=D`flB?PqOoakTN{_~Bxl;brweBro@9OBaUaPa#!DZt{9W-yxJ{~>t=TKJ?jkArr471h|9<3zcd+C^dmEFkWMNbmcyF`n4Rb6?HJ?K`4> zV1CdZFVh|OfvL+NAOXa?Wdi^Xf$)FEzeN84;Ef*^m%?8k@4g*gOjbV;+J=NP0ET-~ zNgU+vXxojWKPlvOuesv>@zDqt)u-|zhPe=4;y?Ng39y#ZLu+x z>~M;552(%$Cz|#Tjs6$0)_x3l(?OCs7++|5ghg_4cU!*zDPD8vN7lb9Kj4`k6y~+p zkA!boL=rRf2H~CFqYsp zefkX7fANP(k3#XK*M%)k~d9|!*4dWpC2&xUkO!#6iKS0zVWM;gQr z=O-2Ue)~Wb`WOEI1q0QybUMI4cOwWT0CO%CX3m-wA zPvu{osU;SElNfu*{^7sikoRhz@J?Tcc7Ju`pIelJ)%R+b7;k)mn)}yR)U?K$+IbiV z=TnC5jN|jq75SI@HTcy(XW!Wi-~`%?>kXcvqgg{MHU=cs*fe7sPDp7~eNBA@;yp&? zej2k)E2+>mITAt%8guaCcGU0=f9HMoxD^lMjxQGlGWw3SS4AQ-`@;;GVx1JT)i9e-B?f2GV{Z z>6cm?jo_qrBr#NcWMDd#=cYz${N=xFe}tbC{s#WiUNrb;@rO;h()>rGYd6qe+uX*{ zy2&NXvqvTylN|9n%?pCbByJg5hE0B>;ifWjuO&CHExzaeO&^T-tCiEDoat>QrT42p z#Xr3H&q-h3a5K-9+drJOD{Zan_f@|smeGA9-QjEHPX5D8zvhJEQ>A*PY z>t9Xy#^~R(*?GHMXCQKT9G_3~k9zns{t4&e7O?oItN3=@nBOMuGEIeTeikqGI? z1-hW4nBrxHZLZ2og1@_!OXoQRZO>lS`BP;&>NfsN zWIGYlj)S4c*A@D0;aOkDpR>n_^tkf!-)MJnZ2)9EadpRH4>jG8*Y-Ju3Xe-lKbkzu z_YLjv6{&ZNUxE46;_VXu09^6J9vJZptaji@b7IgTC=+7<3X;ssjH3gx?%RMdSw9iJ z6?s1buk;8l$epcZm6*!1>~j-t+kidE!Onjncg4EY_AlYD6*CceXB<9pAKq430o|Nq z8`nSW9<}u@{V!G3f3xJ0DdU*5CNjZ^-H-z>To7yfUcC=*hn*^#rkeZ@;(9o_P^(U) z9jw#)nfQ0(i@|^5RPhSn4QkHIRw|_7UgiZ^Mtwxl$WI;X0!08BiEOR_0Fb_f;C14& zwQGXD9Pq11*<+{61Du@gx4oTHp4>456}JLKbrfkPau;s^XFk7=dj25HY0497d%xFH z`%eud;F8&C_i6eZuZX8v+R_FkNin(Ojmk>@03*$Qt^UGVm94Lh4V>j6P|<98Lye*G`^;LIYPa5kJ@C?Zag^B>AG#a2x7gug4PQ(aGqRl zMpQ`d;O!tWY!XQ$zAfKd+AuLgY#QkBQcS2;91w7FHu`>bH^ZrPNcBi`nRhjt&i2=G z#u1gAZ4tR=c+`brNmpfFhoX^Pea4^RtBcjrEJd_%?`YckcXE;{F#=M%hWjYw_+ z@m~iGux@cXq z+@;h@a;W(L%YQLMgb|#JbTW4jUl&2(tqaT&Pi+*E003qMn33+p{dqO;pY0Rj>kEI4 zS6AYA<2HI#l0yqOn38#~A~=mk6L1Z=$Oj`F9C4)?LBD%rKW#bbvHPv?`~C_0<3AX< zlf+&t)2$cHVR8MF4BE_#CVp_z2eylOC+1@#9P)FT`e(xb0PsmU^~gNc_=OFzT!PnH zOq&KtQI=c2IRp}W@y&lM8ZNV}=+>e1onKMYf&fBUUR>J%K)@j&krWZw& z&(wd6e-}JE@I$~Co;&diDR{w-#@L^p?(Bo*!h^p*G>4*~EOB4Xey6EwI@YIit?Lst z&F$>cTwF@u^3qvYyp8y?5(i<&YPaKWiGDozp{tJ&c>7ei*6z_tq_ehGHSV#X3mj|H zArxfw&2o!2qxa*89Py0%eifIsIBV5&(Hg=+I`dFVs6_$&;{jA8k5O`b@)!M_;GB96 zNySYLk0hCC8Z^#x^6okOc>F0joZif>FNy32?%@vY>IN6S;lX4xFBId$GfI^Z1ksCAOiYLS$)+h-Sd3jH#n zhI#e`n!KrOz%`Y-y84L7q*oare_fH%uo%y!ZAO(^j?sxSO47?q5Nj^sNxU&zFH0^WuJ}YK}_s}XTFI}J@9sN0|p}Ukw?G$Ewo_Hkx03%<`7mL5( zsQ&=9R=){NH&)WLUi-A|J~1b_4RqHy&q8tAk6QV&#b34O?Fr*6lP8KicdlG0A1X(A zExK|BK}%Bt9(d$_Rp?XU3`^ElL_B;tE1%ZZvErR8Mv3oyM|Eqb#kY3*nWKYf#t0?Z zdV1H3>i+x zu8>P9InGRuS8!EF?jhkq<-`I zYw;)i6ldVhn`#O{cF2Pqlati$CC{+Qa~j3qaCzv?#@A z)UTRT{`Z!P2yQzE=ca4<<+||(yQmUxE~S$^V{yRa94O#cAV3CiKPvVp@Rk-f(&Bi@ zv-W$$pYT%Oh}ZErPZitSu>^$C^rcJ#+kMEmAc5a-@5Ow#;?LWQ;tz?7jbmNb^+Zy9 zwVLhVjQSRi83Je6NBUR6ptOn=j()kRHk=YV*L^&Oj9au(i^R`n&v3u^<>EQ+<-W1C zvx?$l^GwnClCUJO%P|?jJ^T7sHK=@9_=l;&BDmG=QCWUvM%X>Ex1Qd=rFobD$uERJQMy+dXU3YT@G1mqh{6St$Tl;wp>eci^y4Bq*6zJn?L8;tH*sTXJiLp!5GFe z4^jOqI1X?+A45vmA1Bip$8TT4x?`c7qMnTVxAro$rNH*}ug{BnO9`p^Cee?0U4 z)%tbLHw1)&-i75rKdvj|FBIItcNmvTb!gvqP!PpV00-NT&c3Dp0D^aT)52d0zi3#z zckwRz9T!2ew3cb_E^@HOk+61Vd>x=D`f*-uNBgJT?T^II$3KX&K))9}TVo`FUOjtP zU`m2`)NT(#a6@BnILBJiw${GgrAK4tEZ=3Bq_+g1jaE)%MN#uJI}EmR1_vI#74c)n z+6Rb!EBKSe9vYc$EN``qLh9OWtcBs8`g!7Qt}+87DcVLxc>vaChvPG2cpaRebC%uC z-1T5M;Pae*mGtzmE>xn{_B>jXT+H=tejOM_mpUnAX!6G(3}9f44wyfu6_=;$;&RYk zC=a|SJ#sOepH6yp{3`I&tu?!7uk`t2N$}*w9}9)g-Oh3Xj!8c;Bx1PBsp5v?d*D0d z6+9gJetcHc+qI;O94)cw(CahAh>+av2oIKQTaM_X9JL>$OGkE44!(P!1~wqMgIT>_s#nf{0#V=sQ%Lb03ALu zctgWpC}>WB;olGHdbAc-cFA!veX{;#xY>zj^5Txs5!kFD-19Eg{$T8NJy*mQLR}I9 z;PIHof+uR)?JaQEbl5QEQ67jzyqN* z$(rGQu?|No9+m;}Nw`A(XS#mT-XpW|hsJ-2-X+uSW4ujU#oGR%4a5z;Io5q@_E_ad z95Tp`tgXl@z>W=k&bP2x+C0R0Mn9c~0m&S6@9SM%^~}~0L1iY=WsQ`k;g%qV8-2hB zy?Kv`FB!;q2kzZ|VUFPAra!H1m%`CdM{XhOdmP!g`)%W0%a8&Fy7i5h?Wm^SAV(S#BCnHqIDrA|EIuvpR#*ImS(G7j3s> zNV_z48eZ8SYsvSw4o@Iw8T21c^^Dr17A0XA&lwp%k2L$6TesR|x`-<00Z9#leLIur zNv3(0!-md#@+lC1|Iz+R{{UnE0NIEB6RV~8zwt3*p29hqt?Vz9vR+Hq%$&0+5QZf3 zes%*VZbAN=J`;Y|J}>=`eggQ@aF$(EtMl^4v(mHb!>< zv0!kae=NQiYflb`5WDhj6&;%8w=&BKJOWbzHtoj)ft-PpUR&|YTlnweFNcfpzu{E& z^F@4a9(GxhAsk4!l4g;%l_d@Ww*ZpEoLA=)#Y+{5jH+#a!2L>|3K)DpBvt8+f5>WR2u&!2Wc0(Wdy?N0@-QZWpr(^FI;(#y_+Ng*0FING`N^?rnr` zD>bdm6BSfYc}?oX;N&WXC#ExBr62H1kJ>->RsEE{BwKjP#8w*Kwc;H{-YY|46b#bZ z-iCawtZ*^7m0Kh^3A?g|8LzyJS`_f|mK}3G$oVWEwsq34r>=+k8Gpeq{vGH)0{;MJ zy*A_fPQ>3_!uOsb@>x#F&RxD>?m{DOL6OjL$u;@E{{RJF_|f|__;2E0hJG$ekfaIL7+`%U!=`VtoAF@}+pV-bng_e3(mE*lD&XQEqOpNl} zS`zm%BLd(ov4Uhw72U=$CerU% zsZhx{l$`VOfXp-Niu(8Vb+8NjKIz(Sn|B@m0FirozH+uf-#L@MNFX;k{5Y?iJ|TX{ z{{Zk$AB)Q*T(QUOYerav=$hC`Gc&?|OB7#WbNd>g+B9~3TatjVQuckj`@7cS< zAF}@dgnU8prZkWb>zMs;rYjdT92)692adO^F zC6613BxfhL=Uhg=s7GNcSc1|RfL!1CR=q?84~2!XbYF~br;BP0%Ska5#B-0CROTf1LK;h2SvA%b1| z5gLu&r#b1=;=JcncrKJcsg3FLWk@Z%f({4=C)^w!)KX6OCQ>}RU%rl686Dkb3RIN^ zfPSD6w=1|0yx7ki*9&!HHe-@#+0QuIKvl;jhFIr0$iU=x_4Hp6++3k|Nj5IhrO8l8 zQ-P90ect2KCzIENW|i(8C2-6#+_Lk*=jI9t_3ANR%@pl(jlE2#7YMrnby5ckyp%b{ zNCW_VfIh%;T+X|5c?v~xpyWE^doRqxfC3f;ev6&JXWF`JsG+%5j@3adcV}pRaq}Ix zZW#2zPb*Q`ux6~iQPu}6MasLOfwq^2!x{|gRm~+ zxE%12{MbD)j(zc3sSU%~%l7ba*vM`;JyiYQTm#3~y>gm$#goSzYSJU92>x-xZX=9H z#1r@&4x=3hQw_wir=EoQWNrkK+yj$_UBr?y1_AtQ2+_09#mimDyia3m8$tF)mvqcy z5rO8UK#)cd?Ftw#%*f2y!6v--#TsM`gfRKbsmaOR>73_3=RTF!-_H%j%wh|PCknqZ zMy$-~v^NpqwvrD)*E#1rSk#dug`j3tWZ)JVU=9!6&gD4c4TE0Z5!weGSMMLnNBk5M z!;j(r0NSs_Y7w^Gt*y{32o2^ypt9r)f>ex;PPqoWgZ6dt^~b_*jh+_xgk51xBU87Q zRwM4h&Lf!q5xW}u5B>_ZZeabU?crbH&}=?zU^Zm9IRImU)9|m&OITV~DV*+BP(9cX zEB8F71@)Y=F>3dXzeDqEb#G;hl-EXorzh-V;+PVSx+@4!B@QHT>_D z(ViOzeax&@dwemrYAD3 zhj&m3e(Cq!!XTo_YyXvv2Q9y&yVzz{B z{0?j5Tzc2#IhJ!geOY5FD>=6&mrJ{!xn%h~Fc@lhI&HNCmEQUaNbAy`k_3=38;5b& zf;k!M%|^K7Qk}m1SDQMD%=D4{BYYY7^1rsn>~rC-gnDkZ3_lb+0QREZ!c0Q@l4MqnB#Ps2%&rS*1@Xz*+_)YPpzS1Qt75$|4yP>!e zz)s?*`&7ugkep$Nuj1=K@fN$`ZB`9)!5(9Z>`x**9&m}0KF{IWJ)9tpD|?&M1haboRhl*eqEXH3@YEa6* zZsTJb+iIV_A%;MaOzAe|R$ds4jhgKTAN5<_G!LGpO4w-JD^&oyT3(_3Gu^_Z?F!R8oFNVN5} zp7y=Zt5aQeyTp175nJh(7xyb+6f3qgw2n_A-B1ny z5XP9t72x_)NN(551)H$PTK>DsoRta@slOt>t3R5t$}X*EN$GcQk@Sb`&*CrmLB2Zp zUs{y0id(^E_+q(B#Aq;a*fXj2z%}|O@iFI?=6hJj{Clxk2_$7fJv*Q0#ePo*zR~Vs z)ik%o#gnbVmKnps!bLd8KQfB_P56VTLE?+#@ivD#i)~IAZqZa=GF(O6>7t>)7%B->|sDT@p+ zF>xLbUw$#`Uza`-(XO>e7Pj$^EzVfJeq4RvIvg7Nv*1>?bW7pkeNHlNrHn~x>Zb-+ ziDg21=NyXgE7faQ?t0v}i$6Ox8($vYPkZ2~OWs^rU3n8Ly6y=CtqQOl46sarfuE_b zp}rJpy6%y#%c^R3cQN@fB-4KFR9&e*Ma)i@~(F` z?IkC!y*!WUd;~qjnUOu6)xt$;jgb=>RR#zJij0f{%WWKV;=e?{WQZS7_Z7|Hp^r?Ibxbl(|xU*Ybe(|E5>H*;LZ1WM5?ixVpnT(6lK$pi%;o}ZuC zU$M``9|ZW5BNr(f{HYH^i3nsa22?*gAE4t0)%2TdJ13HOZC)tbIP)YfaKw-| z6O+doz&^ay{{VqnM};-d4@Kj@6>8U-d~w8JMs>upUMoC`M5?Z#lx#aTZaaoR1D{s# zKZg7@d!tR_?-kw~okuMphTt+o6aC^!Df4cmB#)GNa17<&j(<0fDJ!2%JUia!IpQxi zKZZUT(`I;pYwc?2kgTW={7ax8FC^_8XPlhiVz>w`MYJY+i54-?Zo-4>jPu9Rx<83` z_J>f?w9P%H3uztWI?6_Kbu27}f+`}CE%;Q(UgVRS=irLgFv4^@w&gZ}InPs`NbB$T zS5zzLaMh!Cp&Yj`K>kcEk};N0PCb9fug<^Q3s&%kkK*ki>||Egjk!0G6)UpSBwI6J zkQld^6lXhkWDYa>A9pE^JV>E2fyW$YoQ|fyHos~Xjs7z$frUCiq40A>T9O!UQe z;OgF1k`yV~pDdpl>QJkTX`zWSTbvPt$9(Zz-lO8XtBJ%i+(S2fFCZNB;E!KZUW2Jv zLm@W`kqmQ!pp*I>X1M)M+SL@p3{iu$j#M6c^c){fYuKa8-i6NR?4SG+yZ#B=`(J+1 zo+P&Tp{K{G{5$w|#60@^sU@7+i9^JCd~yY6h=KddvxEzQ36QyH{*w5W;7<_zJ@}KZ zco)Uj7WaCdnRfA9&uur_W0pgQ^A>g_xOHIJJc4k?AlK%1{1d;&`u3;swKSg=Yr4Js z6G5`x*85SsyS0zY1c>8;=6NHC0ryrURR95i0LOX%00kWVyl%b${?@)O_+g=VNp%a0 zL8sh(rt19L+p=9)!y=VPqLxk=6~ka1q=Q^@%PUr>;|F*5A~6-C82iWD_YAV>yBzJ2 zP7ec+KAr2Bj_U*>oyth*$s@lWwfy>cv;GQS`%Y^SzxFxNG?jMI@x^r>o`n4@(%j?NoF7A7^!Qs6^piY!W|UL1 zKd23BShCY&d(B>WtYBRH_j1Pw0CXhC$^rNNYv&&ne%W8L=Y{Rw4Qt2xG}1!E7%b;E z)0bX0h^F6nG|C&Z(>F5$hL40Fd09jHj4O?8N?xJV*Zk1uggora6zontzBq zK`6=n#e&&J20HJ)k`vd_N%pS<@pt?cZ^d?Dzu^t=EtFYL-OH+K6G+3^qPBr!&O3R> z>tBXbd|>e#hDf4>D}T>GWIa6z0DV9e&ENR{0L9vnWV*ayj0o6>h%iPEMfCh@*rUR@ zXzs7`GkB`+Z4c4!AAiA1e`{@e*u|#nHaZMB8-|sq#S=GtZ@X)z{D?g{10udP@u%${ z@w>-T%|DBLX{=v)$C%~5onyzeP_jfn1h1atmT6mbtY6)UC-tNR=V{>AuS+Y#!$`px zw>PY$Y8%L7<|3uR$YlUyL7cvgLuc1<8^1G}f^AOjb=vG$4_|7eNb61Sxb-~Os|^*b z9EtC#Gfz7lgYA(`XPLhH6ZJe%l0XNwJVxYk{{YvkP9kk00D_(qHDE?c`eX9^=~mj| zn1x=4=~Z~4I)qye02Df&20C@C>t#AFTw}kn6{{AM^2Cf3g2#?}j@bJ9R8G1^PO9j0 z+c%7kJ+V&_k(_=N(?_Gqe{}xp!R;11)O(}%)$7cq%)SLNI@=y~VU@u>RbI*Zyq z$0hb+Fzb(OfA#4_vpx<#4!>ITX#6uJso4}#FoFk51&&!%w-qCiE`fO} zf{`9OpU{14>w8awmz&wxf}1|+1h(Vvk<$l`m7;zgcymeABGT{gt)ad-WQC?hk%lv$ zDbC}-;{*9t63x1!T6hYw*2h)*CipJZJRz&=$rwxj0Jj@fn6J&`q%p`5dtkc~ao--5 z^)>dBs6%Du$|GWM3hX`c)wu&5O=bKv@NLh6yg4Pc{iWTH+O8gTTc{?0ND3l{kIqEj z9~j=5QgSo68uVR4$)>kgjeh4O1B{II$NA#EM<0pym7K176fo^8K8KIqXxg+)HZ((a zc>%GC{dNBUf<;678d&^7)b#INrFr1c3hlf_uZM zUhA5#!y9iA-D>wUYIhB;-QH?w^IbomBS&X*Bfyzy9fG85)bZNB=J;FtCwv&bO=jYqs9&pQKp+{?*721tfcop%unH8{1bon zv-qQSvHUaemx;9}Ai5YdJMj~3BktVXMFpwnkUZQTxxh90De&L^2oL`N1t9UIlFQ;x z4cq8VA?xA|Lr=11&VFKTCDdO4@=uqX;1SOs=*MmFtKuZ_$8Yvjg4K5e9o5jkR6i>WgX@AjR~>#Y#Le9#Z%-w|t@S^N zFNc5dYySYn3!`oS01G4HR)b=vZu=s?pndwmMX9c?_{hZEV*{GmWfnGq>Nb^vkCBOX0Xt4BjBW zymfauvW2=Zz}!(>kUcZ{*P{4)!F~<3v+}gfHtzW0xVc#uIqCA9iZXkiM{4sb@dhq` zYwkOKp+Y(+{s-bG!jJeQXZ#a`;asVzX}<~dy>4ZZW`a);>Ni^J#X14B#+r<-v7Nwu z)#Q!_Yv$khEyu+V0DjNE@K1eH;3t8tZ1ino;;qygUA>;4XAP>pooPHb8oQN-A%$o{ z+X47@D3M2b6TYk^} z9PuuV;S1X>X4_YXQM}V7zX&bkkQflhac?9Zdorw!g@KQ9GB9IY_)Kh-Wjc=LvY}4N zC1d=!JWn7OR}1!ax{^lQ=n*rMh8%|riUmh$;bv8yOjtN|;=*982fPbVkyuhsILTq-VTGv)ErB(>;yJL)6t471MBMg}~% zglzk>$0ry*q>iVhXFj8;NvOqfr=PON2r3zwWMEsX5z}$}agHl$^45EaT6Z2|WMPIg zh3|qtGD)nfyKPoGzq8vRljJ7|fM-1Q9P{{B(ZgSWvPDfsM4m#7k*jMHA0LN@}r%iXG>6%&l97;*r0Z>ji zbtLuZJBs9vw|a+Uaj{XrI}cCFqA--TOshT5|I+?5yb=3G_|xF#nQxZ`EzU4ao}@Dh?mUidME#5XkhG7ECJi?8 z;`XdHpY1C&Yg-AD2?dln`6j!R#17XSGAow~pc2G{9)~`+`#S#49x&AP{{Rd=fu(q^ zJ2JO&TidUnZxhIO$rx!J_shZa6eB1g99QE10K?yo{yy6QnZ5Z*u1{L>(ZB z$Y`)*k&ZWvW18`$j#qd*Dz)#v=esH#$yAy0XYKLuhgkiYe`ufhNc=tF+g*3VS9cfs z&D@j9kjtmpT{XO>dsv<~^JTSbVm!4dLUI@os(62enqHlt{95>d9g&gW#7(DbQ?!yZ zgEQLwkN10y6$)9d)lrP(^(MZD{jUBZYoE6V!OL$M&8N@cNcO7ODX4ljM0|B zXyw`%o)1ySGbm5lxXo&C&GOAON#iJPZmzAZY}9%1NRme7Nnm;C4^U19ed~*`yRq@r z%IY>NDm2nGe`lErIFd2*s;)j@4t{OK9C2K)#+?txy0?b6Xt#HIGtw zKqHWqVj3`~1P@x@hWIYNX*og?wY!s=@fAEQ6KD3}uW2_nF1mE120VhUpx}K*F~`3- zH2w^zdyv9OEBS31>&nIWw8 z*e&%ar#(I~!pVErNAL83-rqLgAL}E*DNwdK$6MjjE>pQ zDmcY?b+}_Np`__O4w`xH7H>%(QOk7%Vo3xnKV~q#5&4JdB?94;joBN5DXG4Bu zeUSE0Ufh`^c|nM6MUk*v9)!!aqvvo1n0&zFu5;DL;fSvxX~e3q;07$JV%$bgW4jr~ z6o3!qUx_~#f8ewlEP4i^rThc%we6;(a=uNzp{vgeEDA{7C)s233o%iEjUZ8+oZ}Vw zXX3BhhxVZHZ^Q)njd$X^%S-6hRy4D3Cfx#z0=DOTTd|LvsbT=h;=Z3N;S5z(?<*co zD?g(brl-{3@K=30D_`1f%I?->ni;e!SioQTX7EiqNQ4fDBpi0leqw3C((V~p3S%EjHu^=3IZ1Msa+4OzuK*_!J_V-)2oozLoL{t4ClLSB41{hqWfD^1hm)GjTo zEE>D6k5J za6=C)n9vff85@H)LHQ-%Z`os3_yzku{4BBYGqtw4uV~jA#-kC#G!tB1nHJoTc=Drk zQ|NPD=GOx*=0`lUlgTHkHTZ86Wy*$QTAPiZGyHX5gFcHS&1X{y2`OEFlO^#kr>=Of zSh(>IiLYnWH4C>gO>uK0OEiw(d7aqhnD=AU);uIyD9$Uo(ZPo zdtG)vm9N^;Q77#wvF1*sqGh=t)6+<$mJ$dlOQfj9pX%6_Q0>tj}?Byf3!D;E}1Pf zgw>!`3@6j>VVMs&BWtu$WOYBg>G;>(w$`_V}^-KDwW=8aUu)Tv2YEw-0r z-rpnSB0%K#t2T*gHQGgaA%-S6cU9h0dhN*1(;~em;?KbU0E?dj_4M$^#t#i@o*KWB zKQ%4&OF8W1X!jM~f@DD=XQReh4`w;9D#!;Tn)bP(&aqgR$9P&IXWC8|B)Of$YEjs>0JQ7S{4PpEw7S zI^wi7#e69qCK{S_o%-MWk6YGa7TTxUEtoa5 z#a&r>T#z>boE`|TyZ#EvXX38{=+~OGc}oZ<(qy<-+Of5x@3vVK;d0Q)ki?AO@m~$; zZE34$@IfIvlwpVfU>=wQ1L{X!^pV(VR;>DLa=3!@Cz`7xta6C+x6D9xw$66s9A}EG z&nx=OBN)3heD)78jVVeK>G~f~>i+<=4wvC=VB1?rv(0HR3bv|WDg0rZAZ{npzO?)#m;c-pW-0CET0@AUh}KXMubL2EF3@ z4-dvx0ex{CCT>c^9m1C&bqk(=^~P)Z*Zqk+1K|Gvg?|ooZw~lE+ACOYQ6#%@k22!n z^Bj`Qyv`IOce2FW7F;_`5vc5xypTK~EWcM4DEb zTSH^^QxdAAQJyVCds(7=&Sc7d^|O%Jz^|0PIC#UrmR>Ts(EK)$&2QmLXeDMY*q2gC z8DofJ?--1jEJkv!4gfj&@Ak_5ptT=_{xa463-~V0)5Xxt`hDEd1_IfJ3aoBG6Hd{x zuyOK4tO&`k$&GU6$4{}3!nO*G%OINa=}A6(nB)RRN-l7jP?{EOPuR zO4i$DZimV6FAm~rVqRyDZML+(p7MV!75z?j$t1X*INi%Lovg<>BpjALvdSf`kj(uYpC^Qhan6&cR}bDaK_>s}j)H2pXW@0ZI{aWLovYtS+G z8De`^^WHY2acAvVYd3UxC&mp>-T0CnG%+H=+9)|7^ZZ@#uu z^Bo>}RFA!w=7Z%P%n8p2rF;{r>USEpqc){;_vdoNdS{^g2Ney5p?{^`MX70$yc5EA zM>PK~DAcO7=c^HT~lx0f$p1vM_&PrX6?!Vv-#;Y%bEp_cy(QTr?M7db!wh?(T zBy0pqWL$`ufNX=8CoI^&u8+mP7k&u*F|li}iJmmkwC^~XH{D-bLZomAjTnYwy9YS* z0C%t9qhI*B@iWDGM0y8_b?r+2${+&4b9-|ijj-}HxtPxlg~>0KHyjDpX`Tl zQ{bhlvPMC>yHVz@@s@s|WDUa{WD!p_%c#(%y_9Tgg{dh^R(&)Ag6u;#192y|Ys__v z=eRy{3=mtt8+vo=T|C#bPXLTyVSsFR1Ex>&t~Xi?CXj?tv~?bx^P2H!J#KmvS3a)& zgEjk&JNARIL{P>|C^-ZHyQ?W9IRuV5ubjW&yVts|hx>f$+Lo1Y@C`#j&}Tb)=H68A z&l@Y({{Uip@3Vfo>CWSG;SB~-bJq=g zpUrF4$?CtZCEVi7p&m;!@doR}I_=_lS9g=hq@1aC$3DP;SmXv`*yq1L)}#neOw@uF z+l(GM@n4`Mcxp*#hyVbQ>reT%?LBH_w|4~NpyIBLKH-!zqXCXP4Aqk=h|E#hMJZ4R z%1`B8Ew+VoD~zcb^&Ag=asGals}F}}Z;?|6JPv<6^fjF3lQDJIQ-)Zafsu~ijVOWG zg#)d6oL&*UmOShrNzc%JRnc4gJDN*&+7+@@fyu`sj)SMJaac|LX!b_`())=7@w?VX% z)2M6=XPox-tY?@)Ivo_~PI{jwri(L-x@-(MCp>eGp1#$w2ZowKx*YS!&JX4FudXNf zNu=pE%X_Oy6^PshZl+j4WBfn7AP=T%+Pn|@F?i?3g4%BgczagYgsgD-R-1dMsp*`( zq%x=-!#UbJ5t`(yokB6@Iw|FKC-AGEpO)SgmDLa~c-yxGk(1m10I!o<*4_-%W|di@ z0zCf!7#%*K*X_rE{{Y~V-?rb2H4zq>@UvUAf<;#s8i=y7gbd)9`$X$1fypJK9SG#{ zUuF0Y{s=MqSL-oJ{{RVf$AvsU60!2w>J#71jDSl+YiNs={vamDOd9g5@lF~l(waNr z!}VpaDg4#J@cR2|?6(Q#V1_C==cZ3V>C(CzFNRmaR#>CowBzI}pQ!$w>-z=xTmJwC z`1oT`3I6~Iuf)kOrV+EccymgM*@*;koBNy7kPZRz^ScKe;C_nyJpTZK0)EPRSxc+` z00nrT#Z#aN@2F}wNddzt++nsha52Cq1xoXQkWF}4jxeIMif_>#g+3hRUi6YbfW`Q0 zZK6PK_42lmf)vei!X)H{`|q?Kc>33Uq5LCVLf$*OLbAm&uad}jMIxSsWB`Wr;Gd;` zr9Z-N_$5!le}#9)Q}}Q2XTn|_x)OPU=`E~dn6M{kc^NLHWx>H-I47Lf+m=7G9=w)B zYI@FY3 z0P4TtA=I?%Eyo+Nw%S|CN6XvE2ZP0bp%vG`kA|8I7gl=z0FF$NKg`j<(I6Q6#3p3q zsKFs~)aNy!;r{>vd=a*{NIWxPeShVSsTI@b9C8SaPY0$k?OYFuaV`6)Bchf|hl;vQ zpT)0${{Z0MZ;LlKHp%fb;wOf@Jtow7t*mVBbyqu+aY$#9O+G!!4<(`_)q7Xko&o;= zf(L%eT5a-Nc*Dm&De)|P2Dh|bKTd^#QaE^^mQXhp4t(vyCj<}x{-o}%bpHSi$cv+R zCUy)8{h?#GDuqhmFpJHRat?A=fDU_B_m^d@z>?}3eukyf4rI)ZJwe|Jlj+GHO5v%T zVx{)~0I$3CJ7I^T1*Ctk#QZe)dH(Nf z1Gqtq;8*C+!|&Mhz#j{A$n^gJhJF+9zl7~#`BE!eOGtt$frLf3iC7LX)HVR`Ur|l) zBf^@Tl3rg&2#!U;j@m3L3GuY zo(PPSfaxG*lt<-|j2NlLf18 zqUv$arC332A$3zC2WCGvLRjSTYy8;%0N|$)X}ZVl`SF(K;w5r?W2&A`2rDr{eo4P((*DY>`Zb>AG9SkwV2S%$Q_VQ_0)TMh*@E=}=3hc(Ue3vDYu!>Buai(&Rwzw}XX} zJ-?Cq`q!y3kb@K3I4p2M9P{cx{Cd-*wU$Ly^7zOkGN}ZgZvDBh1yYT-JH4w@addm5ik9!#y#P$F&lNDRF$;BdU&T)NRuGamG3Nj%w+<)nrrV zMyT1wSDrn8{dMN|UNq6XF{l3kYRz*k#0`>jbf&80>^`EoD&V0h za~rAXdWiAF0vR?+%3B}|6UKY?1|3iKDTVSrEoM$nlY zA9Ls9RB1Pe*F)XGN?40wr2ha9_$c{r!~XyX{wMgRDQ>(q;r&ALC=@fv9khiM5y31* z2Oj?Q^cTV(f!`YZAMrCp)wQn(T-{oRS>6kS6miQJKgA@A5&OBroN>tMU!`6+@W;eW zYeb(>_*JdlSZnqQ^2Hhg>v95ZFf>x-2^|E5cQ0ZA#e5IpkK5nl*Ms#((LN!C)@|)( zUDh{HMFF>Xiw0FxjZSlbM4PdU<0m!dVY7PpxJ%fj>8;OBSyc=rSjx1YuS5Du_zQpG zO*2>1ul2n~D?2wwxs0HhkJ@fF#;G%e3A=A9idk`ja6110QQxzz$A-Qs{1dwPE8=}w z_01CU$s;nnNVjo7r{uWvS)+_c8}xYj{^;vp!I!|__$e2Syla1BVc;)?*P8yPZFO%X z(nEU71*m4(zD9xq>Wl#Zw%wy3WSaen{gymM{{RI~_@(i3Ujz7V_8$XjR@N69acL#7 z)HM4+cOu+b$}l9n*%|{GQyDFi2tauF+W9qjR8x~tYr8(amO7PaS`S0@bHZAG#n0JC z#eOOMmNjWDKeD`C8(7{&BnbAmmbWg@t=u_2K)50D;!J=*3~`M7MdHu+Hb3H5h310q z;Gf3`?C&E~j%L@aRpb~#4$F-!*~$)fftGBHocjL3{>Z=ZPK&Eg6kmAz_OJ02-x02E z?asxuw7asm4-QCFN#~MBk09)gL~*Id1d(5z-|$#3_$ar>583Bd(yu>c-y6%|9|kyG zGHXMn+pYD!pqVVDYrA`;RAk1`z-9Oh%tF`F;jzD1pDin2uV#4IoII7??7su!J$L>I zEB^om$@nNFli@bGW#gMWNcSUYUL^A_Wyd=b(_FV?Mkj?&wC%?w#~^GQ{i8X`VG#3;@fRe-&@me{FR>G>Ug5LWoEW{AUnjXFflVG9ANbW^gH6G{1ub- zi};78PiOHD;{O1QwR0MP!qZaJEQu$se!=!A`RYJWG1~_f`IY-Y{6y2fBEFmAIOe;S zZARY{O?@lePb7-_H%Rh4j>PR8vlEWKmHMBDxOu|7xt8es?~E-fag?3){<|L_YL;Pd ze_FD^{@C%b%_jC_Uh3t5V>cNV*y7$cR3*OoY&|_!jFo68u9d6o}1xI zvo*rosg@Urhw@qQf;O^F55Q!gpUI@~6fA5%CYh{{Ru( zcyGiom@aMG8C>mUKXuUYlwr!B!1Itf_=q_59V_)u_Mq@4nd4ssUTer5wM`%kfi%RI zcocr|7aqrU4<9eBepfs`RD}<(7_OX;o^Vwb!HcEIK2qG#DyU`ol!4D92h$`0`BN1* z&tclNEN$buwU#z0n+E;tGC&=NAPk>hYR(8_lgnd}4^iLT8vATD1?>~I=a}cvxo?y5 zbO-s@?U(!!xAu7WQ}I*c&Hn(8{wPUrdmo29r?S&#h3Af2qIZd=8w_D$+P+ri$$iAG z05$m+@V?XG&F6~YgW{F-?w_OWW_fI2ONm1+HjR=HyH7bdKT7_Q{{Un8e`Oszq}Y^oR!OpH~i&^E7{?;+dv zrcjDmj@27R)f^6h3jC(O(O=Dq-s06nPYg{PFd1bl!(;*1uUh?J{{VuR8IQ$(586tf zB)T$|Aaljb;~D%#etlnfY71#%5zQp+BoQ%+Mnb5}na2Yoj)3!D(%d72r!2CEI=-K) zKQPa8Ia0;cjsF0x4;7Qbka+;SJ2!4%<0Bv(V0OSh)z3-ctO7x7o|wVteJkJXtkOGZ zpt)SB-GPi}j@|GDbN9M*DK5x;`S*>ZW$n&uLMO8aCc%tYI#}P@# zLUy0=tM*a0c!`9gn~D?OG6NPb#V)~-oGdy&tf>F?xr{t>%ZGCKtb0|Igb4`Yt~k3(B!n3_1L%A{VY?0Qqh zDpb^JX`$2jZ%(+g(arM$W8P4bM+Hw$T;Td*yUT4pIpH%BT%7GA916SPof69P!t+~8 z5EDO}wo*?~vm6d~^y3tGekrum{hxk;We!Nm7-a2(>0b3Et4pEqnS|V=>Zhrr;7ON9 z@phemD{UzzH+ND&3UT!!zpcN6BN{%FG)c6{6iaC80KWCNbP>+U#0x{YgfJ~N1*=TDm2 zBlG+I3w_{?E5kl6@P~{v-6GZwu~RZ zdvPOzqyV52#g8}~_4TjmOa2SVeWp)o@Z&%e#%_(~t$83;+(cqHpvXNohLJ{jujfj` zPixzOZtw28F+ zBkfmVWEr%K2YDCnvW|9-a(MRZQhZ3&rO`Ae(p@~Oy;O-9WUC|~=NKQt+zcMP@m&eJ z)9r0#n(AR4y|iXvN_BE_+j|VL`3EPCJ-&1CF28r;O;z+?FZl%-?uDN>AMp@ok0n90beZNTQ90CR!XixxyPo@!iYuX>#!Zlmc|63%&81X4*q z=QY(r7eqRUrG_2)&~uZT*Ih34utF2hVOlVFdi~o44hLQj^2a}gbbBP0tj<+Y%kAxx zU#b59vVN;;seB`kO0$^_)|-1VxQjS+N!UpnZ9PK(z!BUL)K}+?r-QC8vD5?ihy>#u zYxEoTaPYOZi}22CyCm}N?{24ChdE-&3vG#hiU~OBl6#u+Gx|v>{b_2xXuY3aoUvlZ&a2s_It=w=K)$a=O66?anC&~k<_$b z5oz}vs6C*8+u3AGJeqm9-;YU-* z;~htC{zv}+1&#QvqI`4xtrA}j>Dq;d{3AXe()1g&)SwVb*Kiw)$09w<%BvOJW9Gva z<vj<^KSJbokfcAA-Iu>6af7{0Xc0+r_VNBe0VDP_(wtY`$@_NUh_I?q;~|WSlUJ zN3_}QQb{cCg4?y#(2rxv zA$Cym%8pY20;`qR1C5P5Qgt*-t;dPMI+T)v)c#7Xycv2_dC)i@_X8&*^~NiH)4;lw z%q=KFfj0~gN%?y5{zf`xzn~8T{{X=ce{ate#4qOfc@KqzYy(N+O;X2CMU93_D7TXN z+tUUxeLXAhFNQzxZr8^xYD-&>9QfPAdOOOI`PRC1mF}*7X6lza>|@3`1jo)R*{{X; zXg`YnhYda*t1pRI{!p~; zhw|&~U)*=YPxv5s!JJv!nmT}o?vONkHe#IrzQt2T_Mh@K=L% z?P^Je=w8Q2zM3Z)=l9D8kO=F`oQ#}fn*RVypNIbdu%E%Ng4WW-;Y}C99u9<^#w{|! z2;Ezj`^tf_f_=?;?Z<%r7}+Y%tN5-aSV9E0NaWI2CcEzWu9J-Dwjrx?|@xevXa zu*oPWsNKJp`~&>&J{^C-8NY7-02e2pFT)#+9t8lr`qk#2WSet64<7H4ddcLK*BQ_XsoiK}0EKWEW=9cu8#yc?TX;y5FcOE@`fWc;HYE6T0J zxT!rFUWcI!o`P4559I6Nzx*4;;Tzl!6!_~|)EzvwbkQ`sTU9HP-EWP_P)0WeB<&*% z2(Qqug}?AYAK7Qa&mG;zfiFB$ETKe)#Cn|5z{*Mc_F;qrBkyfNp55#Bz7G}rP5YWo zfuPy>c?r3TvE#5&g~!+Ot0w;d<36*9KTQ$T_NmPH4 z(N2Z7ibuvj2|wVL-wJ*dM3HzGz`hXhqyy&LVWrzC8R`~85Fti7dXvwu($W6TULMe` zmJbnluHMQmvKL5#L>(l?Lx*6D00SQ&$G-%1*Iod7v<>2YZua>{0Ly(eTeosJh(bD{m~5o!M+{5bh_b2j(Z#oC@hQ{a@i$j>~W0j{@H6EdavE_Tdmeh?#XUqBrklaJn>c&#k1SaZEmb1nUI6J)?{&# z2S}y}!<>QhbL&*OUX~N1=FJ;#9DG%11ilvVzMZA3{n2$dDi3m&;PdNK-haZ{*DQ(F zukEa)aniyG91c2Wa-$=jNc<}r&*Gi1{{T+7Tg^T=0cJNA?Ggd&<|8^4^}`Q()3Nxc zXW~{8XG#fL_zwu4S{Yj3B)YYc1zMIu0?5!7KpQ<6X* z(AP&DyT%K7rd}{ZrtU{@0UVlsqT02Ei#E$>Lauzt(8n3|{p5!O@g}}n@xSeX`!RS1 z@vS~1d{pqZm2jLjsMTPaMPZT5aqoQNx8Yqeg;{9`xAHlqC@+bhRIRWvgDi(=9n|nW zJ#qNen|u2L88=(W#_K|hs#OnAfo3p4u({5q0Y?0iS6_=;r&2=v`5Wmyhz zLQQrp;=$~9V2b!(;y?TxL*YAXc|YM-d@hpbfH&&b-ayWAksYK{0y=T$4*b`CA1cLE zx4LrA8%8=dkLarNL-4xEE2W>=WMe%wFBe^zJRfhN_-AxxD89>SB+jZxZXWL9WM|_6j2Rz8 ziuhmRH~bZ+_U!R0-`Hy37Bw4K3}t3mTE@;rNj_hh?X94aJde8D)wB0VuHW^45xkm4 zamUk^k<|V9{{VuV{{X>qJ~#frU$kTX&wc{%PlkL&C6puU%W|keKoVGRG{dAu0~-jH$Q`sV2Qzbg9vb zzKrLo%{>x4+xA!eoF)CG{vus?H^g5Myfg61&eGK`Wz~FDGeZNbN+S?k$G$H*;DEl_lBX$7q z?O)8EkN*G$9y zxBmbHMDTxrb)(}y9(+rm#a9|jM=idUqSy-Z$gd&$9C}rSkhL*@g!%0Z>jq>%tF_iC z<9*J9hA}xf!NI54-RTLaHNBF_Ev@MCyvNVVk{qjViERkl#0FKBm>xG*jY^$YbEZ|P z-$T}{JYC}%Eg;t~XS19Fz{RI56FiNAPT}UeCOFBCSr`-dO?if`g@QGAwHzn6EVM1c0cP`bNUGcq-Q_Kk1)p&lWa%b);Hyz`9q=uC|sy z-e0=9xhKsrnB(N|QXztW)(*86sjO)J8P-!={?4@1V!u*MGD{Vtt{>!>EJ`e<8&4uf zEZx*{D;CdHUkPcqej8iaXVb27up4NnV3RX`?4+8FywSLmIB6PSssP|*nyFwU_GaH- zR*%-cpobx$av4min$CkS$KvmYRTo&E@Y8& z8)p9iP`WZRl2+daBv&{gk^~@(@4>G;)4XqIbpu6Vd=94(V20Sc-8fJ;$K_egabWSPOB)n_9G~!C{{Y(4O#PxiZTtTKhu#pD=4}JvTrc6MCAyui zCAid#i8ik%m|gxvnmBD$H#e6hys9yr59Wu*my%p~ihblWmS#BnyMKrDuWi=+cd2W4 zS3V`x?;i73)Mk?A^7X#-*EbRwX1O4avq;Y)alvC=1>*uGFwE>3_BQMh&Q5FV8Wa+Y@e!V~BZvU)`M41FlFu{i%hCIpeQtkw_lDhkE)>W1U9Bfxy7Q z&$U{%v}p)G!cNe0&V3I`t`YVYBkR-r;=NZv@R*-chC8jT92j}pR@oA+H#S%TNL2-o z7{ECoQ%)&fPh%>zHj3(fDf=A$${!JbYOjrc82A_P#3SpTB)&^kmBKt0ah<5Fhl~_;LRL1jqjXf>rnn;pT(jX*>g}>rJBR*0u*wp3e2cKtv-Fb7UX zE6)B1d_eyIg>B>8eQxE7+RP2S_$P?ikC@}AQUS+&S9|b~K)(w-Uw?2Fmepfs8TtNX zWn3?&c+Gs?E#0iza6i>BRyi2wr@vnHH7F^J)X;Jtor4^X#XHaQ?eF6T^?MvcLZx#SrE3V8!2 zzf*r?pZF*b?V0f>~4Ow36ek#+XPqA3VeW@E`HsE~OZX_QpaEviAhb-JLAmiw3uP zqFuaRREBf&2T&LgLBSRH{rh5kLWB0#{gS>Xc#Gm?&7O&Qb#HO28SY2!EiY~HBxW}9 zenm_G7#3tFs@<%3h`5xYq{fx4ZVVfclHMuT!GgI71!M8GI(=E(kwMeJjf!8 z%Mvi!%OW{A191v5j&on9a2;B5#ki-fpXh#h#uVy8imJQ*jPXAhS{pmow_vJOitsUz zFagK)?_3?g!ia`>>*?)XKC5=ulSdrk{K37xY?Fx_IS13yxto)M6}dju`kb^@K3f3> zqN5^{tFPSXdF1-l0*&YF9M+g(%tLT{=e8=vrIOxVM-)X&j>@W-B(oF19lu)Tl?5hx zv8&3R#`uP7rPFlnDo^!FjWG9V7>xVi3<~-m_C@fO_M_vAuNgyeBoOG3jq$<@G_7>5 zii83FuL{iE9zu+F9tq>SWwr5L-kyvZEZ@ikbix6iJAB#WwSAxa6(`yBPZC`M6v=UD zneKM1gb!chCchfu3f%STNj)!bxb`wwd$4VJ{raBy;!8<|jPXr>GDUE$69;lstBf3i zcmN;s#ePz1x@5Xup>v_$pWfO?>5w=jPBIVoSaLrK{fhAd+fJWkkS=l1bdheX%^Tu2*7sh`}qD=+&RYO#yxOryMu8n+RB64M}BWf={ui4X{8{& ziKcTJ1_9YdNK$ae2aX0!Z0q_7(r(+yNkzU$jLHV)1Ha4i;POU0el>?2r+99d&I*N&dD+ZlumMbaMo8nFSMzj`?{G3byZ#mTU&5c*C-%1Zli?dri9ZT_ zEZ!l}ZnjAk=AU_EA_LT`B$K45>P8t_FhLdXU@At9RW%M~{I1UjJ)s(u?CNsATQAK1 zt$*O9UNHEp@VE9K)Vxpd&cpjNS@6Z~nRxR;rs@-FBo@~bsZ3@xSpzX70GT9=9Dglw zf5BP5Y-=ef;xB}CPXTz>Q}9l;wB6|!Q`k&UVN2P{*6eZaV_jKJUdU|5dilsOF6vOt)oMKNJ-HzWazMh}@4VbTk`tU_A16GQR7uglr z_KolS6hrpJzZh>G>dBw(Gzfn*foqk&{kVT?jY&#R@e^K9c=<0{Kg(cW>0c|5Ito8b z)kd6GL0tO>MDe$c{vv4By6?m-X8T;dmRpFH<50Pl=2eb8hhTYGil{6|P(bAI&0N*3 zZ{fYSxSG$%-xO^dmLy7|F~|g-I`rs!QShe0Z#)B{v`7OBi}h!~^^A=F0G~?hJV5Bv z#r}&uysHd*q?SO@%0y~6#F;xoA92&Ae?V~iwhD^=>OUXNS2ilEec$GCGooHYvA82H zKLBz+&wN)as2~9D#@)HVKGoPuFp%V^{{VOsgY~TaRtTa51=!gH4th7Y)0*~=9+d3P z4hY4`A1PilgOQBabKy-E-tObhy?u}_=^csNwIfnV&Uoh>_V?*r<*YAvapgtS1Cjo7 z*Y&SY)pVPUE5LI>WdXUpy+pE!007A)!vVE((`W>ePpPbNQ*og=F3j{5UTV;Xc8vW} z-8>|dJ&Xoh6B6J%77Z?bPELOJM;v1r@5d3OY91r;RqfS|pucOliNi_INCW5Y_hnY& zXJAG;S8c9o&7|Af_-x#Lv->x8={Z%BQhsB%(LU;v$sG5uF|o0@wedqu+KN8k9K=d+H^A&cgZM6^9O!-dsJRPK-PhaI;%&Dd6{tAmszA=mW0=>fp!;_uN3lIV485un@UT1N8rQ1xn zw=!gBxX(Q|b;t7hSF->sLDas7o04*|^SGCGn>ePlSXqGRGmq)beuMtOzY?VIRp*Id z*5sONIBb^UD@7wODlrM=?%Dxtka9xcWZ-1i$XZ5^ABf#;tu6R-^I&9vypnqK_4TbU zk9XP^!S4Yl!&_-yFA!_(6~*j89k$wY5<*JAXUw{8On<`)ox|o(O8M8+;Oi~Xcx(Rxqb;zry4ZXqLq(1@h-ml<L`nv&*$J-nzJwF~f$ZV%!l^Mqs^~#MChnltHZCiE=z}!I# zpJH*J!;1c#ejEP)!2y43ZEhC5@hm>G}M*L2S!zUTOT z{{RN;+NGf%6Fwr`#@mx;_AZ7klmpO-H7Raalh^lrsq9E2!NqZQ8Vd1ue|g^y8Ljr^ z{eCC$xnc04^t5tHfj}55&Pi|L1Ot!4y$??KFLubUC7T75{{XCz5}<#*zykoDNzXO? z)p#@h2tE4|cyc|SHN5c$iS8m*%iL=1VH=hJ6Xj^)*x<0>93G>lbM{Yw{{Y~dfACK) zhu4d*cyHmiht2)5&)RNuQFlIK7&*DTk)I@Trvo|q*NFA zgiqpIvti*qn#AyowZ4=Z_H{YO5L>LSdE)^|&lUCm0K>2NCYSAF@fB9D~IJ&*P*stsEJqmd>1oc0K4+ek1 z1;1}k6k8^x7lkz4Ffk?L)-{-5ILAM`vWobU2;^iBt$Ba#3Huy)`}QXBK9BHo;!Xaa zd8l0J$*mKnd97}VcQmLX4KP-bl*Dqt1vwl72(R=kUx)gXsUdsor$){;OCm8qLaxw< zZa~gP(oKIOKk#0k2t(lS+b=`5g&8An2kE3KPzYgjdO8og#y(Mu5tEW>VKbaXdGb|y z4!&DP5psgnABT500TYzj)03ZXexCe_qxMtr9vXxa}8=Q=7V}a>kz+dd$6rb?9baf=6 znTb4r1_37^bUau0hyMTs?eRa@J`wBB8UoLGY|9`6a)TB#o=6;a2b%FSzj>y5SQT^Z z-A_{0{28gi4}knRr|Xs`Bj!&X#5T83?qP;#bqjV`7dxBfl?cW*n(91Z@e|`4$7?5p z;jpz!c*4sRcIzv%fxBkaXGK;R7(#jFPp3V7;+r|Hn*Q!NRX`zd*(Y!%SPn7Q7y~^= zTIa9)Nw34^Ute4`pO}%AlX*E{aDhYa+t-j(oOa+>&B?uy>hi;}?VcIc{wM1;!1#y8 zR+d8Li^FehrrJeU{tdS=`G*HM<#F4N2JGJrMf6~JkHs2{%D`|X0a%aX#&(i>p7rwn zp?PPks#@yaOoB$-%n09{eA$^uIPb#){_SFE$i5zdZG09y2$=e>2Yd_dL8 z`MQ>v?=jkQVInrwVpr!$85vk(mE?XL)mdNNMKt@LQ$Dw$Lm60$w!{Q%9Y%A)kLoe* zDhTy`1=Gy8OtS!Rs>;OUr*3oCo-4x7h%}K1)b&YavxS>1W{Hz#(Ut_J9ciXm#XQogN9Ern*GU_5x@%cp z-S&~mSogRY02=A~wAhtYmZ#E}0zF1AGT{uwF(o|@Sio)LYo!=K zSh~c>N+(cA=L4>B>0cS^9~Ax}ToZ7!cFNCsEST~0r;I)TGT zE-*Oxab7uZsQA~%8dd(G;w@VH#FipEWwi+-t;mtsw2p54J0-xDJOCHUiWW|H?o(GB z6Y5LFK>ZukQ^Qcox3}|KK*|`aG|Ho&y4${1Tn={RSDI^o7XA|2httNPr)s){5&5y( z%{8=hTsc}OxUSre27RmI<6&v3cv9Ny#1Puu-7W0WTP^H%x}=H11u;#mNn;{~ zIotvnfzDZpTIzg5;Ex?ycml>7JD`?!_7Mq+>J0I-2{XU!4`X*0!cG^@y7O>2egjtV z!NF@LVLVG}$sZK{(Ek9p$NUrr;Vic|IuF2?@SdJx5*zk=-D)O8Lgl>cQys8mT#^@d z2M0Wg_y@0M$9P4O9BC_218f^-mbk;OF#qm$|j?KXckJ`4WBj!4#wM9*o+jkNzb+-;*RSCAWvP<)8O5Wm#gmwVPElNQ^RH z?5dz1QykasOd9_Hgzda(Wh=_RYP-FFi6gYIfu-}2ytVn{mhXT@(>g!0uB;=NfCFIY>}f}oJZ$uUZV!No_&a#(c9z^%jvC8%$*DV2|4>3 z_#;Gv#0#Noeki?1xU`iuZAoN>=ZhH1MRTLdjf3|fft+XOLPtMwcq`#o?CtQfeG|jF z7l8B~H%qo?CAWaVZ6jN?zyW|48gzRbdz{9X?K@{1OB`3G>67U`C(tzr?1Iixcro@B zOUG!WP_D`XYTIH`FcTp&#~ICH>*GVV@UETYUqNRCOD&vmU(GxW1P_Hog)eLo5m;l( zTUI#Z8*^TEBCO`GCfU-dsJkQ5zqE&htuK5%avMdoSSEzV#yd+`j3heYUR_SsFPG&E zQDiv6q30Y9lXY>c_*&;e_75o)Q(B@Oe`XNnIvu6<}l^>t>8#^g;zw)buSw zNWRlNX`$N34YZP5-OY7tZv-JXfRkou>~|`L$>R0c0hLltD#J|n9w8nJ@usGi`gWz~ zTJFAoG?@VNqA&K_twB^|V|1}cgTO{+0=Z=qO^z?cekAdpnWI6jc!NuVZA?JX+}Yb* z>2~*%%V@-}kshUO4b;w54b0QWllL+(Pqf=emzvGDiLI`$n#%BQl`pI>t|D0OlO9-- z`e`j)mG*#%97&!?AOLtz2w1~!d#!j^$hEU)jyt&HiKMq_)-q$9FWLV9vI}haca~V8 zUCYb1aJHTw@aKqpAZ`t%gtxkT$z|k)5|)9FnF(25+Tv+;f(9UwV^hbRSQAY_hcVvR zUFg<2to|F*geK$4F^FT5*49HcxyWa2E6sJ>rvfnmCA#Mr6z(rHyIm&c-c31@YbgV} zT%?T21<`JN%h%HUq(%g18_DP9>Tq?B6>1k=8nF10rin8Wubphd<&kB#ZMP5&mnJnh z41B9-Bsuv@1I=`I8rOq7GvaG`?WcwGn}kqq5#x`@7fK2<2qBMSBW(;f%RUCxIV@{c zOw*sSSn7Jk#9C&#rtH0zJe)-s9$P|AKoD6@v8$&7SuWJ$?y&?`o&Kre{{RnOTHH@9 zyt*!=s_PR*lBkB@{F27EH|yzs^0`%>JjlS92&1-pZn zX$s(oQ=Ss8lboruhIGJR%c8Wj+!m#@YiU`5^ zfct6|3NUd)jg0wjbeA_5x`Zj>Nftjc(sK;alPr*lLV>k~&D2POR3M1shjtDd+xUI| z0E2{oYinPM-?iZVoIDeoe+m2~w>B>UiW^jyS=3;P=KA2n09f%2yTNG-6cVX7Dp{3* zKiRj6d|hGiOGNP}j{I~CdIpE2UTROMrJ-~d_Ll8-lEm`e>Ttw2F4vY=V*nnwHU3<_ zKKO&;zl&Zbzwz<{Po$#86yGvT_rex2$yiT7+$3FvzYv#mV~(ZvffNwb~+ImR>C*Y3Cc5tIH2`KkWi-?cWAVdG0v zH^ct`3n10JO@@v%xQ5Y#!=)q4(ZcfIxWthe5Q{~bw*ts(@rU7O!(WI$7yNUj{5|+@ zqh9O!#;Z8Dv$4~llH%s#W(9;$#)`$Af}Yq_1=xjFVg8?A@Jqkgw?X}f{{Uny1K{_D zWwXEW)KY(Kc)7HY(@k-GJ~q2b*P49GVQ8RQ#*)O$Q4bBfpOfO8>)2BEG2TahS%Fn5 zDzjQI%>5eFJX@&Ca~FxUTPBWMHxXPLR8W~9C|hx74c1YU^k5sw5yN|`%6{wtCv9-Q{l(KT_*qYV=_HY$~ zM!}8E*dN3g^{(l3xWa)4i!|Fv{{V&wZuJIuhh%15Uj>(;BZ5a2`4fTlRzLsM{uTZg zXg(tGez4vX(j*bt-4vGexgs$f1i14LVV(!6KbJigfvI?xSkwoG;4(oRjUiPaZB55- z1cTF{0GjD>%bX75y=O}i%{LaTdvnW_VHr7HA1J(U9L6^4rw1O@ z`+NHhd@T5Rr}!Gv;|`H-_E!>Eycc&j`H5q=kNsqk$o~Ml@|<#{A3%BHzaPFP*)^w* zuhVw@%&(p@a-%$UT?qT5P+tIGIQ*)r#Y}&j+5zwR$JRABTQE_=Tp+YvC<2<57yEvZV!#po1*mbXR@Sp5u@n2ljUr^W9FAP{fx{dvuQlpr zxM~<09?ly6W?1ZHTuoIsyU^o&6Z=4Gz6#Enq}oS&c|EC$(s-niMXSxbb4LygZs-Z} z_ks|UivHgJ0N{oChr#cI{{RDL(L7V7$#;3<+eNn3jmGI>HcXpR&Ngw(U^6qbV|%eU zzyN<5@_ZiCei^&jX|Q>%%8RxjyoB^Q94PkUzkWa9l7H|}PYirJdvDph<1L+>b~kbr zu)ch%vy;R!Ncjjz`-Jduj=0alaZYbMMLtS;M(64LBau1{%C@p=?ml<_0D_Eu!V!MN zKeVOa!p{rOHU9vE^$TUy{4a9lH%X?<#YMY29JAR!?nyk9F|!tCV_%RT8*XIN#F}sh zNMmT$H6=!6Wi0%G*kZq|@Axl%oqzuT3cui|#k)Ib=CY4c)ciiPNQIgxE?^g{^Qq*w z*+^b_BpebkU&j9c#7$#b*ZfI!bs~9|;#X+-;~UDT9Ch1_=N0;ogw#E<>o=)huo=>G+xVK1vVAzlbQkWwJPfT=E z-jWf#Die{V+gi*@NASb_GwdsEgY%OnmS8NO+ytm!FwgB2PPvAP& z19NJ^^9&B$XOE{My??+K4;7uxr*K3sHN249Dd2@*1IRc%2Lsn9wQ}NT?5Rql=$Y42 zYJ9N1_FsACz96~KF7<1zN9}vsdF7i5jKOy?U^Boyf5yJ}{fvA=Hihu-#Y0WEw3uGq z#}tooY_d3y$Oj0`orDvf26JB}zl9_FLd)c%4shte;AcI1A7U}voLAVtvlN<-g1j55 z-RZFfd9PR`b40QuCzB9AD*^~O#z#GHMr-g)T~|1^`k$gu!}7G#JI!YDecx`jJDO03 zI0G2q;17P->0Tf4Gf1)4JU28tsSh0SJcXBo8=)*8gZw?g>t3O)>YjX%%<=hZxDK!l zi-VFdx#4r$r?q&tx2XNTiIO_@@eZLV4|lqkm?c=YN4 zuc&`!ui9_oU+nSXB-MNetX$babt*OetWH)tgp8lOI%f^R1zR8jNaDPI$C_rCx6$2c z(IvQ);38~Q+5+H&R$>MSgwB4flOL{p(JuFG6T5mY%4RXWk}CV*Ye^200kxRPMP2z+hgK3 zfu(NpW!5CWKu&PFjB(#_{>gdYAN^|k*Zv8C`#}60*RC|r*^A@;meVKpqhI)UR)LB& zwdCB%ZE}Htk{Q`@xd#q$jO?!_{{Vu$_y$q?YiiT!a;@BVI!==?^3ax0VUi%O0VFW; zdE=oRQ}H2Ft(4NPv}#)UT|53q%5at%czn`$X02I;*1X5*MBF0H4xrd zR>|OGV<*!G1Y_~_s}Xnt{$>%VD%d3TJvwyfp5H@XgD5Lr=jv+>KF6D9*P2{pXYj9R zgW-kb@~m+zn~pduj@*KKS3zs=di~i$uTQ;c>lt##fd(XZ_xkljx6K`JAnoR2L^kB}aPeFv>{J{tIS zsrWa+QpC_m5%mOhGA`Bmp&_=Oamo&~y7XG?8gRP160|a)XoIwDr)cCyj#OmyJog6` z`+pAAQo+)GR{M|2Gu%6;iK!cPzu&3k=?bhn20mWL?_~Gs&+@FzRLe94W5kY5a(KoD zG2hy?-raDg0G{KxtXt7M5i_nb6^Q49*Pl^eMIOh^Ceue-@Xy1s-F>H4z8@+ujxK~3 z87$u53<2&q*p*mA>m!Jc^sVQkFWEt zufn>Jx0)XlYVrR7qsG|YyaTl0F$<7Fo&dnyeQ{9uk4Cw;gZ&oY%#J=;KyE-h436#F zrUx}9tt@(tyb!}`ZJ`(i7v%+)ryP8%(Rlj%VP1UCm77qb%C93ERM*o})hC|Zx0GZh zQw%^Ij!y*o``1a}?GsklRX$TA(X;}B1a;wSrKL!5)!MgHzhhCG#9~t~_XLoU{?w(0wvYPJU-#;{k z?&k7v#{dBj7#wW?59znWm--fkJK9`F9n#1h6p1HN%;W{#5LGJe_eSCK1<2y0ve#k0 z%o^I?= zEQXKpr%Sxlt-)lt)L?tf?X<0*PVA0B>w-Y8*x1E+ zV>Q^+E#R6)GBh^uw9-ga<#%J{&N)9R9-JO}oeCXr2Dj4mJ2_&QD=e-MWfYzr?l(Jf zLB<0feqMsNH47L7AK3cy#$>#3scmS+cW|I>Z{2`TIm#6P0071geCbuQHMc{G()9gp z>{|Cz)J?s)ki^m{NaEX;XITnh?o+@35HbKAtG3YmBjLLn4>s}db*UzaS!IS++5_`| zJffs!PU16x#ySew(mX9bseATs40vNwzMGi$-bNyft^gzn2@AJ@_d&)$42Sb5ZEPMsxh??g2P1a#DPQjoW zs29Qn`GG*NtV#eN9J0Q73<$+lo^OuUR?ju&m1{E3^Q@LPN{Y^)0sEq`6b`sQA#st8 zYUOn93tp#vrT7a~wt@#XaN8+dd6VpNFgO4L?hgcg(0#$KfwZ}-hrBzM5t6MXm0EaKy_b*Y>__L(9O*NQ9#?P+JU&eQ`|CzAE@D z#2TcRaq9_tA&kWo2yN9_7XUJ;RAq8W0Y>0|0BV)D6N^W-%d6@ZYS$K)>1S^umJ1dD z7#Kf0;0$xf!Oeble%qh$PH%uew%F6WSMfqkTS)OPoon_e^?e#>P0L!y#p4Yf#CI0+ z1Z~WW>A80{8w7*(-nZfek~rkB5?Tw8+f1sZNl~|~yeP&q_edj(;H_ZQ6=a$Rkt23v zB3uwdC>*RTuDiCAg4iw2GI3f)PX6xX$u+6|eV@0F!Mk7C3--71r|j9IYL{LjvGA?z z5L)Uti3C>{5!~3@U6zqxNa6(=c8}ym7*KEtHS(D8Br|Oi!y}(i${X;=BELib0N}5B z{(-Fj0Kr23B+sK>L8!$Y)r>DSyR3>Pk4?5oP_ps87;StLf-p|tMSg4Pt7Uf}juJ_c zv5b;Weh0sNSJ`3Hmt@ChoUw2h3M zl1Bhn^jZC}sp-?r9;oSQrY?<6w$x=*DhP8Oq@HW-=MI=*o`;cKF0-sS@pL{O)UP!7 zB%UR6Z+|VuiGI6}$YYM`-fuBDB(l1&Z@um2%Bngeg`LlC(7Zuv*3(|cZDV#$_9o^j z97IV9xmeXcX91f%dmIjH&h=}(Fluq>sMe1x%&HVZ>{(ACqfpM#lwbpO-N$_5wbh@C zEHr&vRlR*y9WGd$s$5wtvAF*LS03BFD(CGkhwm$SB0xIwIOHUG_3aAFM6kP`QITZx zjBwpWe!g64Krarz9kh(Qh~_pf4haMrE{=#{JDt7fir}%=AhXr&+ey5+Qe%6llGqh| zuGKRZ40rw9WB>`{Bvw79m99&p-)gCIb7N^?43e2HH#Ls%Pt3QL*~*sg%G-JE*58Nq zLE(QC9Ya*KxY2C1*wN%?w$*gFz==$ZWSR*MmhqBFW&;C|!-~e!HAU6@C2`^{JX+rU zy5<$1H=Drlr{1kuHauT6il~ z(tH!E>l$D7t)kse5?ZW|@xZApg^ZE2%dSAUk2zh4jzAwSMl!6t4Xwm1ye;A8w0pOa zu5JU!mi}zhoB+p2ytA1Ys)AYUPX2*8LsoG7d=x zT6l@aW??KmCGjJOy^i)aSno7VK1+Q*6^2`xOg7>sGKDWBy0x{v$I8T==`bW6%DcHe z?AMd{i^mq%-cf60xsp#hXVh@OjbjGD6&V>*x^kj$IN2Q4={xBG*5)EzVtYvwCNpt_bIPd6It;IVYAhtrltP z);j+HgY=Cy`%t*Jusdc}GYM@~T!P1WOMO9OQoRev6Z@@$isvSGe_BI+Ov*k?#sADsR zwbhKM;Qa7}8?fR&O4ns9WeM*iina>0HE8t>L&Wpyem%4Bwy}L}b8UTlEKrFhnj5gq z6G*>j^KET>*BH*&F57B@oPcq+;U66QShc#ni%{0}AF@MdEUGo8g<%p#LTxvie43(# zWIRA36EVX9!4;9>FCSUyUmJAo2m5N`#`99W5Zqeo7n-1X8Dk-QyIo2XV{{uFM{6Vw zGDu=U2M@qr81ZD^60da~FT)R~y!w8lZui34!op@j- zJ<(BW=M3DF>~%gC_?M}8$4A$#=8oG?mdfVUXNN$%XKX`ISl4?3-_lx`)r)sB9UlCbc$8~P^)^>VMu#yyzs^;eC*xks>(`BGZ z6s{QW0G{{8{x$IU@SU!gBuOT-eQ~GFWY@ahp`@%fh^m)Oa}D%U+71B`*+0!AHrJ9f z`^Cm}ZK@TMHng2f!S{a6JOxJKA-L>YsM_s&UEDdnp3$gH)spC1cyG#8SP?q7R z5?jfwPvQp*6om&*)|m;D2kxS@YFFm~3b$C8$ErHLPFR<~;@icxmNrvaT}f#K_swk= znG~6Lo;iiCp(AZAfKp_SjCq9Q)!22f75Iz6nqIx8PkE^YhB=yRL3Md#B(q{3KQb%1 zAU|-15L!Fn?au7uBf}aFx8SW4#Uj!v^mNqi*7?@qWbj%%u`!R%n)gLnV079-mA64B zA3Cy-Y70#t{u8fQWs7X+ClRh^F`%oQ%| zZ9Unm;tv@%oAD~e+f}fRdyAN4k54n{vCSkifTwk@lW#hrXC$*rAi?=}fm^ovKB*sx z?sR+kB(;59y_BD1wY#!gmydjrjrxox8-jjpJ6r@nNfDBAIA0LnU0!@B)1dJ#qHHaq zd35WWO`=w)85DXq0oLti9vedjW6}5yC&vB(UmVdK%HOku0yUQ&jTxic~h{cf?$==0t z^BkJ9{{RTZvwTB~#aeqxP0Oo8CAGea(*FQw-6EGr8&7*pMhMPEWpO@0`?<#4yDbC5 z8fS>^yfNY_ZLVRRrm$G=F0Zvu-!>Wvqk~=?w5 zC=Sw3r(9lJBtjxcjUtS*bygtl#IQaa@!y8oyc(9?A=GqfjMndNvG|fmgMY1#0cQGq z@w99*&g+*g%Q+l1Qp^uX({#`HN4#Nq;U5p$e`ea*ZgKl$sA34M| zA1qg!F}1Z_MU=nD>If`P7$>!P{D{ckAn0+Nf1b6}UtC*3AC>`>K6v!@=iiZB;9%qr znQ{HnI(M(q;IwDX&}DMqE*Jnta&cN#mo}nE;=>nW7bpWa(MR{Q_|+0ruGo&?M{(cN z{{XLFW`4#00JDy}`&ItY{tf+}d_8#{n=gztNVItGO{9{vK6ElXlkA8y{^}%l47tI> zH_95h@mP;%2RrO^!k;{pJD<~c{2N>Q63qvQwLkbMhs3K3y&FXE%wO5w54*qeAW8KL zu*r3(z_+N0JTOE>yi-9k1(sRYaTWcU&8S4P+jyHmw7AqQ@0F#J30qK>B1yxGmQU?# zV6K07fFda1=Z-o500#JS*Ws3l;lB-hCALjVLh$~d4wo3Qv{a55E#gS0g=2>PHG7Em zZyiRn{X?W{!+pnXJ;1|!6b8r$pHN!-J?~KbMV<3IepT%=L zbnz8kNi*#5xJc51giQ-5^zRKwZ2Uj}03fYZGG^1(&MS+U!yUUnvu-UeBMBoF-S%c| zZUm0CV&_5ln`?6=sFKONwkwl!tuZ(_$vBof?4!`_(wG(coybGYkfv* zDW-wtYJVSXuua{C$mg7m*{)_wm@Fh~8LzasUJ_$y%V`QwuLVdNf&q@03co?p zyyrV^ZlZtx*8Ww0WB9(!@R60 zF5o@QxsP(jzzCXIK>*=L%)kcg$8pHVZ>4-;u=r>2kHg;#z7mMFKNxCo10l7HNY;}` zLv1W~29%OPQNY37fsFj|@o&Np@N>snCBKNLI;OD*O+!&~&Q?NY+=;o)?Cnyd{vy2W zG$ksMsV_SpL5ao1VdG9Z?Q#DA7W_A$-{}wI$TcSt`M+g`)?f>|#Kl>>%;k=9dU8IM z>K_TdANW4Q!P-BI^(`D-Tx!uup|1dVgDj;3B1pSW-Qj?4;=BvuJ@1OVdtiPl>f$(b z`EHuV(%v(X7*JfUJ79HUa>v*T`xoI|h12{l(IN>iogJ>^SI#!Ovje#Bs&kdcP;1q~ zMcUGLI5Bl2Qm-;rx$H~vQ^!zPjWWXe*G+=ds-!lSkw%+34WnxT?oWTpv2@*A#@fE4 z1&@d)8jRCQ!^-*>Dsc6pDWC=#-VYM=7YO&{{Vj-j~~*#ci?A+ zbp2D}#f8SAo0<^v4ZsjgYDc!*j&gIKt!bI!BTkI!HL_=RRfcu3Fr>P)VR)y&UM$sY z?libGo12-foseY4%o$rdOEaM=M=D7<9Zh&oh&(CdZ824K9UfarSgs?L(7QHxz(V=U zboqyB{Y{6(8YhQ7A^neE#8#K+A@dL1jzCC4+2}HHyQt4RRQ~`FJP0o>B$gD8<7osT z0}cj0R^WoCxI7VGhyMU>UUrh){)?lKB^4;kRDRreH}-q@xBF=S0Kpt}{{VshD1s^c zZKY|k>pCx$a3%?;>FOqmCXy2*DZDPnM3PPzgT;Re(rY{7mrT5f#TpVBPs-?{kgtpo z$^2R5FgfH``Y3K=% z!lOofysaB67M^a!#mbgZk{3ALfxyqddi1}ApKI|JyMM1s3X=u6h7~QeM;ee@8RLb= z=UzD4im4AgdB^bj`)8bc@n58Wv5$aezqHl0twT>)B1o;G8^UAFl(ek4ehUQHNa{1s zYVx!BsNp8xZ_Mq==}X(j>i+=M&n$;X)OE4s>6TA+aKM;iX;T^FbEw)6r+;%^&)`i9 zQSirx4xN2#Et|=6BZ%Y$TW-|=jFLb&I6qqZVjl#?q{m}p3H72HP{BaSP{HJkMiIGiG! zkWS*mJdE@liu0*A(Cw2)3E~X}yiwuSgpl`ED15abWWd=Z@8dpM3P$XnNE}zna7YdW zM-9}G$I`yTx4YBy{VsVG!I#LDBEbup9x|xAj~ubUUZ#czZvxd?s^F6zeyA>{!#ZtF3<8YeZ7>TQ22n#HtrOZh0AS2RI9oco^cj z*IDgnF|hfR;Ep=_8teQgc69x6I5I!de{|W~-~3!~dmDx_$7;+Z-6T&M zI(cMBN;7WSnRAbqBw+r6d{zCJv~4@$73QbmEkJ92AJ>&NoflJkgK+m&7lSbwWsvWA zBOt=Xn7I+F6P7snWA+Kv{Bz--iE~5XPlj>)N;fezrm15rMZeH)M%g22Sov@)iVh2v zEW?JtBEPoJ*kk?)-}?vn)AoMw_lkZb%i<3TTv>^9{{Z+!H62RM2{hpxjO!9h1ai+Z zw)}|1wX6XVYK9=5}z^s%A2bU9TORTH9Nty5HuXk^GZs zf3tS0cN&<|8BmP!6`KRL(mQ*8Rj+OPI@~qDvUuY%vBE(b@0GHDg%@!-^aDMIO8)%( z8T%H0!8g1(xbVlrc>GT+Wb=6q_N>`chWYN?L!9==`eMGF@aOy>pAMn&J{SBrvhd_+ zKpRDdm{emtt-4JpK7iooiu?*$^dqETewY24yPv@a!r%BM&+Sd|Emq^epA39WsUt7} zpG`JVkbC*}q6om95Y19p0V z-r(2u^KNOFMfxS)1f6S~DD+5 zpZNa(75Avud~L2Vigndz5+g>rzn_-mVM0u(c?9RO@zit0YhE1i?Unc2FT9&;p)uUj zNRz%6JQ(-BOnmapNhFLPep;E0d3~beLMvo`H~#>^UH<^!oIkPW!SC2JR`K8L-=^8x zc!R>4zMFk>Ji4s6_m>)fo8;X@rb%Q{RyjPtfmEHmSMkN-vvVb-rkE{EN|9zc!7Utt zhBzIz9uI2%{(s=2z7y0uZ~Fv%R`9-#@Fz{uEcWPgQNc^Y7TryQ7Q7u~h zpZpW%-CEX5eMRlzK4O86hbIHC{{ULEFKig8MqS9m1mm1{^&+&jn|Wr{8Pa3rlbjRB zAKk?|#(QLdEbq?j0(<^H{d)b0GxBLuyE@5J?VcUE*RMWlKo|$*#BqVR;A1=r*JceP% z908G%O+jU+Uh5YrV`m=yq;2 z9Xvs{UQs z^_dPuQyo=7EjqKzKj4(#@J{_A_Ttmy&&C60;?_i$MbWQijyAcI@1^X?F3?~?yP-mV zdg=)v1tz`bT{#nnn{6=G4)~dZ1WPbIP<=A{3B}$voADG(CvEnAYeNxXwxVuM% zT6wJDXl$c}V{bS=)JjW|?EL4+IA; zuOy1j8=dT|#3|<)uT{~08+eW>Ep1}cG`|ek#Ssz7cQluG%z?LXes&)*BeR}$~i>1xstF{aNk)zM#t`uQ^cB5v|o{SApz4%@6Ud|)^kK$ z*Hcfv@HT}o(sVr`5oEzbahDFLdWDfzJ_sMcGuyp+JUYL`&kjv^JX7I=e-)5ERQ3}l z$s+_Gkx!iQvu8MFP!2&j=Cd^~+5_SBpfYOr5@@gmNTz%Ej>U4@LYa_F7UpAu3w6N& zj%g|>YSEjUXTP5r-s*CjB)N(>3OtDlLSsA-6rzPya570@!R?-vpC+qqac-|I;rU&P zu2>Zyl^~T0#BTl^aNW&$Od8C3wqH%ShT6?e8_cI@UI{)7h>Wott!E?_w7$6?T$JXO73U zQq??Akld_-6$q$VH#;WaFmiMJWS&sq9-xkcjJUVE)Z~qA1=MVXveC@T9>oNMg&}$8 z895lQntnO{(SHGc2ir{_#m^mCXpQ&0kF-ZM;K(>p4Y4zFakTCF;2PGQ8Oid+HbOqD ze-TfqmM?L%OO*;7NUqJ)as~+KHw^a`$7?!&gRiU^H4QE$mNK#|rrjrE+>%tO$sKZ~ z*S`Q);|GmD;ITgeyg47+{uOv%#2zWRxWF@BM{lItov3jo%SiIb*ldA_i}c(FIQVbI z-|$!8+V{nAli&*tAHo)8ivwu`>a#957@Be-;B@`S1D?ITC}dUTy0@rH6&0n==@q;_ zG}{Tqz2aNM(s_vXsBJbN8QCW3RY>FIa&eA1JlFEE{{RJ>{il2%L#hhpP%w;UIU)ZVG&ydhvZ27 zqYyxIfB@Tp?ZEotz1Q}9@qg^+@b|~7;_r*z1G4yoVG@yTrO&C|SlYDgK=W=qoyOr& zMivwMvdl|{+PM9J_$mJY1OoWW<85R|@w)EILa;}Sd43$8Eq+PPLm%EX#hJI*zuvZ$ zj&cCUKW+Fs{s>R|6!?F`cY246Zne#3cpb{z&#mgd6>eoK3@Z)fNn^i|gqLmq0C-@U z%MqT^SBux=MI4rC^GTz^{uY11Z+r*fX{~$>;E#k_2ZoK%5ui<7?c-QgfF!Sx=Q0k2 zD=P91dspf|!r$8)$3L`Hifa0;v>pr7;*`Z<8k>y}tMKzkuc&3;}+V*dcy)8Gfg>rV~*9q=`_od&ObEkTrT~%0UMJRP(=>QIL-7*k%F@?fSfu+L znzorXj19!6$+j07m87>ypm03Ih%o?Q1JU|#{4L)Ud>^pUwaW(9?(bDvYpu|DhfZd< zO{94j^G7N~(ww8@x!dz#*PY32&h-Je7C-Qg*jZXfr`hQ;-b%7aMxAq@Oj;khe$$~n z-|qvQgmxRgaopppA-qfDMbNcdyNmlfRgNibrM=R1i*3jY%kp0b)kK&$F_p;xV{bLW z>Hh%moZ5U(vhZ%Lr`ze)I$fegEP5^Wp{BHwV9!2mm8`8ckZ`9mZNUZFx(^P>vs?Iw z;s&>-YFcYf9u$vJv#{1;R=GQ4=Shk0q=Rd2%S^-!jXfeE82L3 zzUa#m0;;Y!#s^((Yr=LKkHi}tKVGx9GikQ)rOnQvrRoqk+I;DAsMxexc~Cu6g>jRd z;1IhLO6M(Yt$)I6;t%*i?5`r7G=+*eBGEO2V>-Mrsd(kme8W6UxyO{Hff&ib0C_dn zgDpHU;TxS^-thQx@@Zg+3MJ=(t{!=0&gQwgd0=-%4e|)Qw>bo0@$?$Ws%!cWiTps< zH+qe%vC4}IKA!;nKmxzwOJ(~}5vT4nsM7tR+IL{e91Qc^VBA4v;V%#A8gGa73wS}(>ROH5TEE+^UIY`~T6o6MR^n}-3kwyuyOum3^=$jL<$D@E zMR?ru?H9&g4)C|bjXp~g7lmfIy;hDFwY%_rz(}JfCOZ_pT=Fv%s;KpfdxnT(CH`EIRmBj7H4%W@A4ppZb}^;xXE58you!*;Tb zC2!`sxKpRK--DKO8mwD>$$FP5AUnS7jJ*Q@a4WZ);byNd#w#BYXnGa5`aD*M`m1PG znvKvg$Cufw+h1GH{9Ar|UFvrfBrw9!vDA{bx)5mkr~D??9x&6s(z?VD!*;*fw)%Ozvqn)g@pWyeA~Cu718J95T2e8MaK2s4 z^W-nh<~PiG4-!Y>4~G5(@b-&+rD@hh?`D=)y1c#cqx`Y$CJ6LJ@>re2_ftuR>s&E;S7*4HhM3 zoDndUoZ4!`%mep_Y2W-LFg4KJ>FeQthq`UVwymNt)WP}Un?dEFV8QM@DsG|o{qmUD z@xj__uDsK&^?!*v75utrzYy436}&duMy(o+>20-TKBUCMkXPo-ZmKk|6P9sr zbH=Ffa%CHz)>lKvf(q1Rx2p{De&<%A*YQRXm8?XFB52u``FkL()dxX zzZQ`UN2&f2IwOs!ovw4w2?<@eo-;qgn7mQpD;cjeyA5wtwY-{N_?T(7n)UCMb;u*^ z)7msB2g)Y$gv>`FkIZsBLw~91z8%o?i*(eq<+-$LNs+b4^z;(Oa1t#$R!93;CjoxY z>cnyHUrp-J4S&Si4!fpVro7SFL?n3a^u1rmbdkExt4V39DPVKEV=e|p8?2fY<>kl3 zEpx*2%7x^Xd|6!I>T5RXW8JkZwJ0UiPCFRoXfwdT$gZF59dF6z_fAmPZ4KXwuD&UYvbv_Z zrp0j!c^8stI@D_n;PPMwJwnnwt-Ewgc;{;=#ddbS2-jfvY})LJsrX^Eq=drv9#)*G z9Flzb{{Vz;YrA$IHzeAa$0q?t7~VMeS#|MN(@56zol1R9;%MZD%+u{Oonp*S2?b=E z?TMvoOmY7JKJtNp+%83L9KxG9XO~TEbDjycZx49M{5_%1YokXr*A`JnqsM>YL8;qKr`+8_*2T@vq-^yl_rObHTb&f7@(>Gc zM^$jd?O=N8c)!9r--o^@*`z9DxVRR#5-aQag^kt26O$0rZ=kr+q>b51u#Vz66;3xR zVT~<&NtfYo#XD~i!J}V8YiTygX0`Cqvq%zrtcwnzqRA%WppaVb79H8t;MYtm!R*dS z(4SO!?}9a77WlhMzVQD5g#190OQ*>d*thWWrOoI<%#9D(d_YTDM0SD+W9BwM+E)j? zxbWTnqv5#^k0!Z)?Cn5F5Y{g5yfDx#>+-BkV&d9aG6^yV@<}6bMjXc-)!oO$pZG-J zk#(CH<(|s!%Jak;XNZNcxki;IzlPvJsl>SfQW#{JNpZah`G$YO`6Tdnjb!jV6Frs2 zw-wc-Eu-oBpN8&QYf@xYv$EBM_6#;|GF-?Hy8RZSxDubNv!&KfA(siF1>DtYYjx@`OB9b4sMJ~2%rj9F{ zcH1GE>%x}-r7TU3Wsypr9>Nl@GP zk{GUFffxWptl1=llpmBDnfjR?Ch*Akaq$Av$37FV(7ZJ)*C~3U4L4TOh>iYBgL|k# z*A_Q2Dm$`AaT!K%K6T;8px+yblqkX4DTTuM3Fmn~6{e}-PZMeP zS~J`!zqU*Gt~@j)xrYlZw|CkUvRgAcZNO0!g+cjb8QT8<8~j4PzVP;&b$vTnnLvO? zcdf@}i!HX@rX4>}xAOOAmEAJSj1eItB6QQVjT^*~_`gE%WVY!$!F1O)uxe4DosZ17 zcXqPf+RB;bl)}hI@nhx%x(-Um;b;68ANKq3AHl!a!e7~c!l|ctkHnuA;=hJ_*1WXw z6n0t?pn~#9FC>QE@;i??(ko#)85P;(CDt+r`T5`6%(wRt>Ocb0Il=&=c5{^+-~-d! z*1u$b;I#h$wbj4vf%|)WLHLoRUOu6q>C=x4=#nst=F{|6k{#^OBx?R;)R$7DU}h$d zIU^_Wn^L-H{PdeA%sAcGzCr4JtM)z)%T*jdJeIOPJ2$V-QstfZOA$tWAVo| zNU;7O032j@qC($w&*CXUlEAh=J#p#Zt$k$XmWQ5^YA6|>843vtxbOye^pNvZg&#=o^!jx?=b#2U|omJKULp3)O7jf|SK4kerSdv77SSb~V>jD$;hoE^Sl zbNqSeF7RrRO}s2oA2vqu~9BQS88^Oc;q~sVapA8K_@hkBvZO4!g3ov4$+Yb96IoV>w5~*WNsjI?Nfq1az8mpv zrK|XU&fiqiEm59Jc%qgaCLcM0G8x+0uaF^8@~MpmayIfSnm#$XzVPVOb>n{<>Fkk1 z6tWvjDIpvro<^$_@tFdQusOlw%wx@X{Dbtts0HVs^bTZ`LBjCa%ArpY5M zb!0*l!FPyAI3olOzmj`beG&EikFo6#1rtFmnx*8iF*&x6>{4m- zt3(FiS{&eml0|wfJ~Q#nq)TgUpggg}0wd3+YMyH<@C=)*-0vfn%dyW^J!|LfQ^vk0 z@qM!Ex>mWa$#)L_`Ur-O32z zByz?6>V?1q7#sup3k9?Ih+F7-Y)^4>b2N7Lr2-E*ow*I1VB?Hurz}7f@Q=p*6HK%C z)8Ws7z7S+)U03aTuahD=OK|gVEM!7*V?}Y-fsB*Eub#^B^Pe^K^FGfr%jv_F^j$lg zhwUHm(@W6231NBR>z}dM>8`fpPcgP$S7NOz$VkZw0Qg zdVPvbqSzp4p>H*WQSA>PQa^g%#U^sy>*vqf-VHlN_-S(-Wdq#Z!mG~X^1keD>c;?+ z&3=@83%b$t&wySChf#{=1YHWi1LesiL7p*|GNSCk?~*q3HSgrrx#7&uD>kH~h_n|& z%U((F-Uq-GLS*~Prh8QEu z7>}R#eweSKVzkmO8B!s}r4Z^a%o(_wpCi#uIf4WkSg?{S<4$l&Lw`@{Lyws^Dt3ETTe z_$T&S@b~T4sCaW%xA3Ql6HJcY=4kDqvWYcjRYnqd&pg58D6u+1tCktSB#*)A+O8KlP&Jlab*?RUq(fyZvaj$$L{iw7noiRN4wHt-AmL?@ZE^>}oa0ouc`g32R z{{ZlCzrk%2jWk-eEChJPkvTRGXtT=Dpq z@c#f?@!q$gABY;1<|~~#c8=aFR6#K?;4F_cgazjqWdP&ezes;34MkG8_=6q4fk~@}dda&O#xc8-PyTkAhBk;Pw1!Nq)?iK0$~b?hBj@ zo_Y~omW^R=KCr*J32z_F<|P4{7nTI|=ng%2?O)jwb4u*~KC4+@k?Vg$>p$6t_E5C= z)vwEU<9X!K^m{l(%+tiN?l~b=HIh;h8S+Ra8-Q>D=D&YF5_|)>T{6SL{{RVm8K~;o zU8e~Sm1lQyUvcEM#4$=sx%rwx#FLV1`aJ&tf=m1h_!;{v{=}aSt+daFx^B7SJ%3Aw zT-SBk;D&kRy}yW0g_c;0i+gm$Q7LW99AIF#U($d0Xm5(T+?O67&|c3*QqcXHYj_w( zv4AW**;V36Hjq{~izH)_o<9}OxZ-uHx^Y@3r}>|v<=i<6P_y5?&*rn?zx)#e_N@4u zV%GXk!i(J!X3pFFe^ZN2w*;Os6Gro~AN9+hP;2rp{t8L^JrCJ0_JqCo8S#qLU-**O zLeceUX3{UnwTj;O+{Tu}&9zpNP}c0s{{SlOU;xfde|LIcjeaTJ$Sk}=;w?h)ls}kZ z@}u5Vk+?OKF@yn>sq+cWa0##C6aEW#7Kv&900ju~g}hC2k;S2C7GQ1??L>xI7c9&f zoEcv|j#r+w;Ze@&*Qv`SzT|bwCWIjEC-=(#0Db=eBl8Fk*|S-9T|(316}w0?mzv?2he@Tdouh^z zzc78QN%?W<)lChZv>$4Q<(zLLX+n$v9q>rcs1@d(E&*(@XE8D7nEL?zPuHj6#d>hY zTixYH2h@^x_LBSm0M6!*!S5Yw9}V?;rSM&@k1YD!f7w#csp(K&+(8ZAVvt$MK7Pw2 zh~-{XLga118+rL#;%|kIf=g-Fc|R|gZoC#MoChRz7{?>;XEp7*09kxRrGd+>)GP=m zc2(7fuckQYYp(sZd=GJ^d`h#|H3AKd&W&L_Oa{{JnG!+l!GO=y3j59n#$L0d-(R@- z&M2WDcDH7Je5JS8;almBWX9wkgdB|5XRYa)W|^zoYJNeU846Ex0F;n&+sf}$O@Y&Z zIV;a6w?A+^$qdti6;48fz#YAR8rs!$tF2GNmNrJ+*eO&071-~lw=nHfz5ww zKd~>0bT1J66}g*0m7aT>*r#=7Q!rPFGN~J6uo%e!81O*JIIr`~{gs*G|wXBGYD{{X=(JV|NeEgw6Tew4}P-Uj5PfRitZrq#Dh{+C|OE$$k}Nb#Ru^ zr$e@7RWOw!Ax`3R%78Iimil(9;yqpMZDTsT;lre%CXQugs63GRo_mK2PLU$A@xC_H^oRh{q8%6j<;@xnYo83cHzD6oX42XCN zLk+KJWlMq{(8RSWsdCbcYQOJGx1Z-toWN}|wN$@+u7OH>YDZ8~m zbCsUyov;A}m5U^t9_QM+T|Y?pHQ~5pg2e^kRmYcWZLu&qb0Mvk3R{^xOf8}DoZk#I zTP+mZ#dJQ?Df11@t-D-pm;z46k&xhl^9Fpk9G2wo-}q0!wwfT+JY8dS@gf;b&)K82 zQJfYfVnR8FF2KEyY#nnMRWMq7UB!wfW$Ri@Z)iz%XSoqD_?^m^wJBgxr z&BdIu+T0N8bD1sK9lbtoJ+WWq)A46dxPtogMVv6VveflEILHb@HMQHu0d5#&Q||IO zPuJjwE}^{x9a3dUvr$FPTBgb2cn@H%94&$%7_4;3%M+i1^In*OmrVt@E0uk1nbtHQcRjP*OO5otavf>t_~lO@Hh@oK7&i;1s_{gfu2%ue4f z(oLmEbNkjieshR8h8P(0%{F~M4dhbBH?6FElm7q&l>Y#NQ~v#3mTcii%D*}Jf46NrUGU$AG)pfGcv{u$bgeQ3wY9UiK^zc6 z5`{?PSjGaXkLzDI=^9>>;Li?Nct65AaMSe7F5PWyV6;0{C}RKw2cYAV-11MYZd+=1 zR*|zu1fhH2Z^-O&Jx91T{N=LVRK2yLnY^uI=sQW?+I@V^;aORoYu| zC+VL`i&*%P;9UmR4O3dW@~y5Nqe(o8o(UaSC00U#GC$wz$0ojdX(QM6cV6l}LgQ1n zlNRd$je_vO4<|g0+i-odYGu?lRfx%NE)Fm=jsf)R*Xj6In18~_yi1EaT`NG;ZxZr7 z(A-<360S}*hY`xs##nsAbYtb`0=y&RXUAWTdN+sl8UFxhUkK=*5Os|-4>G`N?`<-a zVkX@TutHp?8$!NZ$G5oW0|)u9_=o#&{{X>WuKZHg5O^j}2;0h-iWKmb)&nTXAKEwb z%L6y31E)j9e23%DiJ$OQUm9KB*?6w+;)T<)qe%^n7s3<|hD8abL2j5~(GkhVP;1wQ zP^GJ>AH3$N&Y!~m7EOU^IL7~&8^;B*y~%6?QPNOZqtUxd6`(ROpI61DgOWjZT+cc@LXOi@t2A;Uk{>~ zl3h1gw7j`ta{PHoePq(bxkq({kmK))>4pmdOu&HFJy zq$J+~JOQmmtuT!ums-5H(k$bRU*-KMhCO1&Qa^bt%R7lC;ALES;kd%DLVS9ewj!pke|CqV zM+Zs@>L<(EPr&^v;pc}uIpE(3_%};`APa^|zYhNZXhgAK#R-o3-Z|~n*bT&qH)9)# z>OEgU@s^Eq;~U5z@^tMpMv0ZqkKzea&+-8?Zm@v)D=(S~jrcgnp%v6=dSiH!`&IGw zm30r=es8i(;cMTr-A^NYu!M(GiU?LkLAb`!oB&rWI?dv530%L0EbRPgWcsb%sOs3) zH2b!ZV}OQMTa6R#5yK-$K^bw7rvM7zoSS!q{$phLYqn6gN6KYsF=V6-VlF4o_P3Tbmye+v!?YgmgJt!ET%FZX?wpzh_*S;Eqd8 zeEI66oPwQ5?Z#nG4qJG~UbOgiVQb%1VDh+z_1g6wAU2eIf-#R&T}>3U^D z^I=I}@bnE^!L6(K)(g)PX+`Zc2mo8nUqrXP+OUHxj~=9xtf3jW{{U0C5uPfqguHpF z_@Bm8YW^71r@hg1=$2H2O4Y2j7+NUMtYvKCiKjct@}}U7_w8IBgW^95_)Ek3hPA8> zJ5AL!=Qi<4HI3hd?q`wP%$^x_Nt+BOj$S?RUHALtLfTCgW;VQ!}7%igt~;2UFJ($oibmP zh?Fe;90nYWwo#wv9RC0b#^$%Ac%#Ig6w>@E+8xHJ1>BDqUKf3!+!H%%@2FT50%+O3W5vmO1?&lSa;r`ltUpcn(~ zo*nxnv2JXg%@ZN#lh9R~`&(~`8gIj`KUuW4v(e$Vj^j;-QM0?cNhSUBv)SqvS6AxW zS$6*X76g(92D={->DndFkF?DXTzhG*^!r$2W2xBcHrF7`#oHG;Zp}VGc*p|?e(}Kq zw{(9HT-DE!}-w6^O z3_;?1Tg%vEk!{Q#HMKraELud7Y-YeJk!o%i9Ql$1Ai+hSr`pBPByYTOTyltatb{F=(9>3Hy_{GJndez^Bi>I1Ido+2L z^3n2lgCaUe4+n5L9eD8GoZlBUj~2g&*3CXm` zFa>j3Zoe$Q8(+g79`Q}JuAQYAtUs}AJVk90K1v2)`aR8zl0z<7sZn*cmjsY_9JQ{o zVesbrR`{FaT_Z>EWjV{ z*I(iv!kte@)2|}&Y%&W~OI6fi@q#`i1P`)mao88Y1-@xK%nYe425$JK={_I$N5Z;( zpw`|M6Hjug4x@YEwrM_5wJmK{_TGKKF1E}Gjz)uYjF z9!rPb1ll&C1>QpJ1}Ox-X6@M8Mol>;#{2X;k>JI=_}ORTdrbmsm6jVL{ibwXHvBdt z10z*ih_6q%k{smxnHjG%)1>fkiFB_S_$vPZRkQIm#-NwbzK(V2b^C&>b{} z+*vRonT&lUssvmv)x$qxo6apqRwr0nBp< z^10jqH719z*!b(<&HQa7J}eLRDHn9``1u|<Km6E1iIkSl9dw z@jt}=A@JSp_K5*)KGG<@-*=^3c#~aB38K6De!Vba&c3 z9zD~%GvbXOO}Fs?fi5E#R9DG4j|A2~!_p~$&nyAt{MGSe;ZBFDXxc} zYaKdwg7F)(k(H2O72F67H?th+E8QBNAMiz|gS6ijCAN-qh}z!g%ejk8i$hoPShBg) zv^nOPLl3&?3djaaChV!pdzro_{4$2-(i=Y%eUknLE4~jI1Xf1fh16G0nGOi$4)8k? zNUVov z`WwdHEWPk;&xM~{)gbWooR*CMhJ9}L!z6i5302b)JG&`L`*NXTv;$7dwzl9cFEB%)> z9T&t}7M*!LoY%e^)<5AMjyssf*`Sirhw10~pBZe;z zYtky)OZLADNvqxlft*Jz)$W$*y8#}>T}N*PiUQIhOab#82E1g z07e5&@dm4-kisO?zT$9c-(b_N?w7=V6^38f6SrP?jsSS)e{8hTKAu+XBVsz0t~+QaD)UDBFQ5PH;JY5PU@N-H(l?)BH0o zyQyuyh|V5my|)SxTm?zD1OSk(3#^eA4gnoG__cP^ykjidqIg%scfV^FcFAiU&8DF= za^R{@XJv94;f{V|G*YV{n{9LbYAN?EVWjua^?!%D>fY#g+AoH+czm+Wc@*9pyN_9E zS}X>)iXCB6XrlmUX)Kv147Krp{1hwW^k1?U>;>^B;^Z>PZQ@-T&{=q93(HI03TQP= zbt7E40A)P{$ZRw%IQXYhSpy`Eae5@R)8mkAjxZ7x zyO4xKhh-&!9Qt`hDPl1Zr)}AWF4Ump+5EI?Fse?L2_s0M79GTb#dsT=rdgD7PjYy# zD%S3!p88dfgjP7n!Q+qVU0$Q9N#z+-{O2SeUrhcSir{C)O8`JQ_OIWlr3GX2h}Cr@ z=E|oz_ULLTgt0o>>2XZj&a-xr2Wwp~N-J?Zq-A+`FE zSK=IgC)O_fJ*3~=OMhg_WU$eofZjARk>s`0TTG=GuMs<-+!VV8#@hTG{{Vw?{{UjI z4E!ejov*)bFNqqQHn*NKl~x6uOK!Hh2AOd2rNch|04iVY3(+Ga5<@te7~*yE`Fj1i z@dwABgF5b!sA;#0Cy4y@k|-tCQp;Grj7Z9%BDIxOrL$lI?DJXw0EGZ-0r>v_73U8T zicn7RkLLG2_awl&G*f%Cg7Iv+Z-Vvq@YU{@s%Y2iW#`HN00{)qSlirN$mL^ka`Gj- z?2pL-+0}??)D}@%ULrpfyd$LP+Qz%#8(D2Nw0K`vi&u&5R!db9q?aR4iKU8H<&n0! zNck+=iVCr<{B6;u_-8+Yyb0o~eI{rmw9`T@y~Wm#1f=a-HMg@mebuD%^#I8rQZbzH zdcA5th+h(PJI{!EAB68F*R)7p*6QO;w~>viqX^zOBuOnbBnJjT=iA+?HsUh8*M})~ zXQ3%4zK5UuK=EJg0c+vg&l%hJYr}W;Fg21&_TDMIHx~wJlq~Q-G~07CCPTADDdd)33r^?zMYORuE^Ae93PuZDW$zw@llFvaBv&ka7qUC&B*!6ZnA!y+4a2N4_6u zWs37%w?Jh1h*(47B^lud3OMahqt;AnPh_rt|Iqw>{hz)*d<^&%cd2|p)^2B;##WjZ zqZW;5ut_9F@<^NatG{eTo0Nt^aHl1X+=~24{giagOT=FiZ@h1Dd~L>=ZqrXYVcni{ zf>^QY*kaAjNj|mrCcmwGIQSFruKxhQmRh{mI=o&VmN~U6xHpj@EhIK^$hnM6S-=UF z+NUgf1NpJ=m*JO=e0H*F7E(5yqusJK&A|+!R7`-YkG?wKACw+RuNMo1r%sIdU9Z)j zPm9M+rD$n;bUo+e=frytj(#xGz7O~U1950I2*OWnrM&x_qObuP0h6_y0h5+vfIHXv zFZ?P00D>ca$6gBfZ=h+Kw}`w=cV}s7Z(?HeAk)?c3K5akI9_2Qw;N0D&u>rj$^DQ1 z8+<Dp1d_|<%&32GWa`U_w0^(^FJ8ZkFn1W?66>}#Fq}TdlUU*AG zz3~pAb9i7;JmAdUU)m*>L@UD?6iCgLE4Z^TBRnw8bK@}*q03H-wueO@-Zwe#f9V<; zC;Sta{t0F9yG@P@tpeXtNXQ$tjb{5(EbquBBRc|5piB{4RAL3O*g^ zmh(dFtSmH0Ba?a@uoB&>@<_ltuunr?5o@5YhcA{*E?HK z0KvF_1av)CV=a@P4cc2;#T-bJ8!O$&B#`cNv0@}+Xf2HA1Xr8?0Kr;+W`B&o7k_TH z_>Kby_Cxr7TJGA){=`JmtV-7~X|@xeIzKNiP~89v^SGQ6cL81<2sQ0FwoRX@Vynii zu#J38seQ{=_x`8i7wq|}_+#ToitWA~c%I%{7$9c4xU-T_W0lCp$4HOP0sa>GdLGsH z@5lcD62FSS;D#R(Z7tjV3qtXhm2GUhHYkBLXcOe_>Ad51HsoLviuuP(y-(SX_I{hi zch=(1#$GRpbp19jI0++;KsP&u$v`q$j(+H`&(DZowa20@OI;p4pkxg~Re|ILpZR8t z(tQ+;E9Gk!WOVX?RvkV1@t=DBmVe;e4}g9b_@DcDTl_`w#Az;(@cTu(*1mUj#pI!F zEWj!A40*Q%fO1F!pf&wd#5m8ZVW(MiNAi|QgZ9|S#qGKM#p!+!@b`keB)%2XZv~#6 zq1eZ#S;+uPkPVb#HQI=vxFtYfO9KA@FBl~1=$1OYwYvWRW@*}$=AO!uOCX8XdEM7( zzyVcu0HCfiao2%flMjoc(+82@-6kc40!A7LWDm4@Jcz*Ha5BT|S_|S273%^QF{Dbv z<_=3p4;jb{{ut;h`J0wa-D-Zu+KSZVrO|v*M=4SOB?(&Z!cmkebI8rsjFXM`KZZ8bt)vRQX9NyQ_F2?khX1{U|(E z)5$%hz2Io0ocVwV?hp{naKX7F1pKe5>pEV$Byq;37YsKMf>e-COylwWE8K;&JgL)H zvod@+7uft&rB52;W!<;W#&Bg&-s6IEUugdT!9lzcae44-z)UPrn%vN z_$0@SzBqV;;#HT9Zu~1_YvOx2JU^yKZKtF#2-;+td2ViIk~yISm}U)18Q3IjqU&@vgaWKHW9k`h63~ z1XJv@i_QZqghUkvdUCxp_53QM_{&_1KlO0M+>sbK9rDMK>0ic9pDv94g;Gsj85Z6S z@dU_^4Wkfu5L@Q!(EdHfewBQ9BJwmRSD7tbkOtx~PjVNp=UF%YE|J+{)o;bV<{0y( zL_$YpRw0Q4oDBXI%J`4Pz7p`swU=3#!ycd;$dWiGkVwn-u*ju;nLPH+4JNI*aE#@$ zJuK+H7P4&IU-^b3hgf4RwEqAhpT?j4qu~2ce#d7e!><6!#0+|nIThvF*N86krnI)x z^jqysuN8=R3=ia-0tBdLS5m<7fXA-wyl0q?#G03gJV3r4@C^DsvvC-7zJOiKlUmIg z!B&sSaCY;*ESMvW+08?!k)$xQdf55`&&CtQbsUjeBoRhK5(jMj$3FG`X}`AD!x{b( zd~^7-@ZV68;kfYkjJ0Cm50tK)CIN7TMS;PGF_&-@pE;MS|Ae${{QwPdr7(@OZ0;kYM(p}aTI zNqeD8ZLa3~M#WO{M>VJJ`hced;=b?0+kivIwCbLM!*E~zXNl3QwjGnY_& zrhNR)*#Mr{;---P>!X z_S?mg=)w!FR?;N7)2(c+yq_`;nU&1fP8A4HONIGx184Z*{{Vt#csobX^j{qOXV;<) zeWz)Qr|CK6;h#`pZYBe1{p)X3jNp-$z~;ZE@7Oc`3N7$wNxQlIpME|3QPuR>{41#2 zYu0)cI!w~6_UOzav|E@|yo>grvotY)&BS@wSr5ecm6tE5LaevGo|^vvk@ub+VyE?* zlwhx>zPdj%^iINxO@C3e)%9znSk0xZtD)V(Uvl}aZ+RrL24)zI4m**!3{@Rl z!gmE>@Ls)pb|-|HwJG)cE16=23TF`NR=;GnMM2E67G(#Il#YAv9(bzXMDZI-sOvg* zwc>q3D`%SDS+{E;1ebC7vcm+nmaB;j4XfN61Ii z^(bCFTTW>9w)FEgDQ374N03~kO3Zj0x>xfY{v*Jx@c>V-AN)y1mnN3$~F4vEFgN7SRacp;FNJD}FtVPutt(u)vq4+;l0gD-(RD|0Qs!+jF|&rj66;QEra zz2=E_CMvBMOT8hj;dtXxcM-ZWPu?3ytema5`CCVErTB`KZ-UA_X!@EtGcA*}Jr(sB*hV3*1#t_)B~@rD``C?zMGmapL_%6K4jw zrojcoaTZa%uXK1PkqAcH8DpLS_pZu)Gg-ILE&dz9ZK`S?+9?>inpcm+T6v6;4Wz58 z%#*3c8Oe*#WRP)5G=(Mk`T`!rwJH?Ob8t= z9t`+xe`{-|=|TS0HjjOEqRz|?E)3dw+K3o}p<*U)m>${YEn88yxVoR>XT?o6-EFkk z{H-l~Q>I01@wB8EZKTyL=DcmJ;4?pzdE|<>6h+QX&*ATeJOOE^_*=qr>6#yiuB49Q zds|7zioA0Vgyiu?mo~9H+MJHIzh#v8qffWV-0aa5w3o{~ z#<*-KJRU2P)6>E}0ulHtz-PnyM~H5A&3>9i^7wZ49oUjIYbloRayM2Dv=NV}maVC2 zwsgJ`_=T$KE2wzqL620D!o(zY)_NYHZ?0OSGBAonw`ibY8mJEfYaOF+D#B~hA5?CVg^nzoKb16wbVW@)3mu-&14qVI)0m}+D@fY zfht3&+uJ%Q8(LIJ@=3tlK&hu9m9KNPxqlF9J|Kg^DgCK;XJWhVWRbMJcG5{#=8@3d z$tCKvV<>SF`wa6~8dr++ShX!j;w;kHTH9bPlrlE0;rDl8veqYC_ln6R9q}2@rFkva z!wU}x=rQ@1pMZQV zscKQ_vABE5qmsi`wVmAx%A0TJd(>F44>1?r(CztaVjVlTV{NW{L3!YrjnucA^m^0z zEM~T~@fEj}pEQd<_FCEyAMS=SKRs&^HCqc!V$0#Lfp2unJNO&SyRx;_u6#LovrY!Z zYa*8hRr2sbX#vkl=PW<7wXYc+U33jL<4n=vjt{rInvR!zL~n;lE#^x*b}Z*~V{rtr z0CCS>r+h=z^y^(~#a<}5@fExdCl-sYT3KJ|(McgHimj!m+FlsJ2boK3>T}4J(gjT0 zytI8!#CIMr@wJ`KpKYNd6uI#RiDz{jF+eet)P8B&DG2ic#_^tZo+~fIG2LqZ5x?=T zhTuB3qdNpyd?%~Te(OB!Oq0cLs)(&XRVB6%KpX%_0b!ez9;YXVJQ3oU&Zly*o3eaIpdG*lY;o=!5zYlA7WN!gw=P*HgLO|H>c0^6 zFYHvCPTpg=yVGxUyTGs#kw+!v`8L@PfR3sc2jv@itLdxww%_69+_xuMk~`L0P?tn_ zzTHH*X_i}UKX&MfWdVK$;hYW)Z9l|%l-~-Zk3o|D`|X!7LRP}~c-kQXp<-oAvmfr< zjii8B@JY{3U&WgL0EIP3^lt&!Nuz&kSQnP!)_eQE3E!*C0>&e@)4+yRXvxaFp+^Lq z=N98)lv~{u{72$H33v-d(L58Q&*9Gx&8V|pn`>=ePX%8Mghv)E#6CQ+02x8%zVOzwr`}BtoeV-Lb)60!V%^!7%xPfNCNasAxTAuE{G$L4 zvx{#U9c{HeZVwW8OGnZzfp)xyNWZXkwhjB#mHVWW*_F~Uos3(i-gpKQW3s-6*Tb)hJ{9pVk9Aw` z3TyUyi>u1XZ>0E^@?uF=NRf2QrCU39J2A6(j4(XsCb_+8X#5l4d*#<{e$92RvRbVA z-Gmw%yn%zvUi-tl&- zFXI0Ij%4fB8a4gKhN*0~8n1_M(@{mW`J~3HD@AK0dx>OaB|~Q$j&ZePuXxi^yYUVD zmmVb3?1i=L*7pg0rt5Yn7&nzHHn6MQFzCq}ZpYo^^>)x)-H}EZb-(yRd_#NjDt%Qg zQqpG6j4d?zt>Y3Mp?=e(7r1D~{ky{#3)GX7ReUX?#-9tBei;j^p!cvU{{X@(s4lPN z`B^MPJ|j^Nlm`qGE)?QC9&4QVabe-D6XCY4ci=cQ*e=wtOj`B5y{oC^n{DTZ zCz0cA$_SCZT>D^FZ-)LhXudRETj0O^Bvx;&G;`Z|8om9dyy+~B_kPc5dv835fbGYZ z(~R}5Q$g0Rz8UJ4zYlKwA$#Fyjr>O2SI})6P1A236_lmGwx3Wj!6E8yL?LoH9jQ^1 zc1Y)`iiXl>rNcGniF_-f+IW9Znh0Iya~-_4dgOCC1BOj;1;E+{L&nRK*CQ3s>o$>S z-w~kK9&1}zBZkCJD{8u~ktX4ixwEznr%RBh8%W6;PZ+P6G@pnbHTaw2csv_(Kg0h3 z9@%NSl+j&7_N2Piw6t(fn6X`3%96+!2;L+x&BWNjdcG24tUx>A858( zo~qDV={n_&o#u~r0!YYG^2RwvnF&?QZ8gkUA!FECv0YTKP3b0cOE8Vwj8Cbpelhq1 zNBEhoz4iRJG08o|G0i5Er%M?V0u@tJ(Bc2x3oLsN+9eN#lo^+jTV^t~-RQ~Uc>Hgoe`kNfai&`vtCgQk@jlzF zD*1f3z-yFNHi^dg_FM(vAY77%r|L1brz^(0=zSgHO&i4e&Yx?d_*VAY!xq+FdnUQy z-9Xw}B!)lvd9=2YOr68!h2xYnjBZjza~>2LKg4;*g}h^{X+AvEP;wvj_`-^+)p9D?i>NWuo#>qXp z`J(N?J8?6_Rla3iz*l!;;_JEmW8wb*1^A0k@!^M8yq?y^%f(R(yD7t=*vWqYklHIT zJATzTEscb7E20pmv4u(0ZQY)S40?{MZK>%W5_Id>?Pk;2qL0Sr(_GTv`!UNqNu|oB zR%O8QXJA+;WEl)iFN8-;*1S8eS^O#1?6kP=Y?65A_ z`2qR{uRr*C4}^4&9Q~W&diV;$<4>MTYsK*0*`9lN*>Sq=#nh|Hj+j4z)Wk z0r;rf=+>HMlc&2}MIvcFB$r&%t(Yl0r_-m;l^`kth!mA`^I>vF8{uCN=o-Jp2_e_K zM?KBm{k6`N^A8VA;oU}6KRU$)>_cS8N`@n7U`F*Iie%YzDQlk$K0l{|^^f>Z?ffO< zeNxK&U06-y8)wl(7Ho*SUTbx^wfk1!#pj1;Im1SzbFHBG^Iq`J!nVBC?0i|_xO7{V zRbL9Vw}@eyDRU#dQZ#bQHIgwZBWU+0L{K)0?mRK!DEvO(+1?@W?bWWC92Wq-rGIbY z3ng{K8Kg;`8<^ac`A!UNk{gnwS?~;A2EWxWA<}i94#%l#)^7~AULUp^4z02;of1oE z%(&V~Eg^Jd&T)*@a|zwP!S9Nm6HQ!4;lC7W+D+`1w=(PPec_4hh1pplgxdg; zTig^NgRw)NKwN=dr(=1k_-Dg*S^(AcjUL{_8%gYJVUNVJURxmnZYQ{%8heQF+XJ89 zZ8&8lf;zPLC*dtp;(asXMv<#(X>(~LLN668I@OYF0Er&%B3evf;wdB<=U^O~<~|Me1n#LAJqzLCrmJgx;2s#AVnOeAbYK z8L#gP{tH$6WO!@fAN&)4#vUEibwjN9x8h{0;eQXVtEK&iRjo0ql z8I7=DK5XP3xg7TWYnkP+z+fEUnz4Iv70mm>kfVSz+qF~*$^)s-^sm#Fi1}(Tv{EGF zDgp0|i1nY_w_Wh9$B1-a3cUUv(c`xW z&UHJ8qif5pue7vw_Kq#pqdUCWr(ArbSB=CN{{X9DWj~SWW!QCAvWwXM!{+^(JWu-z z>Q>hu1H2t+<4=cHIwh5X(S^6$FSLntyJfb9duvnk&1*6%Gb{n7jz%R_(T>g_{@Zf= zV$!vL3-}-Y5$oMAZ`o(j1--VPsx8I?EK6^6@+7Vsl~T~UGK`Foj-SK76ufot`^|?! z@f+xpTgm1vp0(!LK=BgXTfpZ2H`;NJFU!FDSZqSX{DoS7idUX6El6=j{?tX&ev~rEw{=cnHjWn%y<4=Y>Z>K^v zdE(T}5(z}`hPjRlg~@{E!%}GEk|}($KhZ8-H;gI9ao2cu;_}DDH$EM)(d?UCxd;6k z(%yOA+5{i#4V}Ens#*zR7$Pw;E3e8noS#XX_ICIqtVs@o;BORoH+2Bed}~ zl##52qO>;(_rZ5A<=Y}K#z9kE1-HWA9cZ^wXnr2}h2cG3OP4YxovmAHdNOCyIOaD~ z!~+b04ZCW(|3 zkjbPy93&`xQsLK518cvK`lby$k z^%-qb;jh7QZ+8B2=^A&J=D>bjq(rerJcS3j=Q$jXzcXD+*uph8XWY`2HClYTE&WfR zKj47B4zK<#f5BUHe;4R>){@$IOzN7`MQm zYS&X;S@?NdN_eJYbip^Pz*G?+@}K6BK>)BPp(Gxa{5}5w!KV6khPnG@c+hI|Z1a35 z(r+*c^67F5Zrli5?&bI+BLov)-tM8`&xG3Q#U19Yd#U}BK{Rkfe|HjwiUn*Jv*Oz=0t2sH+edK-@>Mo0X(X(2XEI8X%7Ck3 zRXwwwwa{w64!lpQU(2dnSj;eb4be61@-RqjPF+G8{lEqTlC1{y<8>TOa)HLCOk1uMVQGh`zf0obqE?189 z2K}ZyCGeZVx@&!#;jV((jQ(yTj!XSfnW9*cPRP8tp(TJBV=A2dwf%zI{670oOHEel z=IsU)=0gx8p~+GhjCJR~<^Fyj@Ku)41&%{ zMYj8Edl@7kPGhi$F=&HF9v^!N0aMS4%PUuK!_kf0nn%>|jxP;LwDCBZq@f)z-S+RL z&&0ou9~d=H68QPNN%1;WyMs?;kL)NmCdkQQY=D!*qYEcYsciMHio9#%ZxDP!@h+j` z9}!58eW^@mn8u+|Nx)P=(NPctRqU!e=BDwMxp}UXTefD~EQJA5RYM${zCLE>lbro4 zglEQaU(viJ!j$sbU*Bs}@?I|Dxn`BQqofnjey2sE$ma1$%Bp3FUbz1N3jX=O;MH$~ z_Bs#j@$pOJCaAU{`2LAjX@h5}7;Fvxf)$S15zle3sZ%~M_05!5*wXL(~IAHL?u0FNN z$M~*HB8~TJ{LedvDc-c(=#Sa;YnV>bZS?)TgTHeEKf0qi$9_1f+Kv6a<|Nd$8>{AC zeoP{_;&9l-OCN`^){72zWfj%;S0eDx!`nA@N1R7Kp zQ)(A7S7&V^v3zM&VF^?4HabqYY0!yK<>$v(&$0FWK;Z(q9*CJR5Zm z%@z;d>i0pg$!rGG3~SWOah-}s-CfU|s0aNMemefne-Az^{21{!fm+)K#2O5?PO+9) z*u9vI6_zKAIN%Mhyb;w+ekgvx-|$L1PuXiodA=n0&dU1sc1dm{wQ$h}3-@DH`LGEf zZjhS&n(?2H{vUif(vwxb@g=^KZtwYmp6`&Fx(d6%3Y^^vg2>*-!|_GOb*i|+YNAO#Sz zDGl}|_$+tq>!^GN{kc44Zx{)t_^_vnH0O>Z5;S^|FDz=LurD0pyzDXg z+%X`@f%yvIn&1>6gD7FoaC6?jsxo|Rad>FgkJ?A#xs3-y7g7tO8>whB=sM|!Sj=$9 z!Nlgk~l&g&i-?*=RByAT$9s)2vN`hiu$kiD)^gs;SY;;D2f^xo97Xdtim!e z&pa+k!OnVjuYV)2$tcbAXOo>$bz$!x>G$Ek7wTI70K%O+#Pi&0o*J~ax3_DJM$S7e zLK#6&25rHb>30$XD2y`WIoq1iz18phNovq))(NXvdx3wzhp-B5f zi;0IkGwd>Rj1gT8`|JJ~x%*_XYWilasi|vJvlqJFm2AqYptLLG#Tt$Z#quc`BL=+c z$HZDDoZn#Z=ZfyFTEXN^Qg0R7-CZgoXYO53_E_TpjwZ~4On{_=nz5iu;wu)^HLnEO z%55xpn^n@UlTni8NZ!j9ogtA@NjVD$U~n^<`Dr(E-sRI{tkZNWDeo>J@Jw1>udnKU zW8Z6D9-h-qx`I$ZkTiR)9J%C3?94~qBRtjSnS5(!CxpJ+_RSKyJDMWU_=Ssu>qU}PWOSn$k9EDu`zgZ}`6-gx`NUk-m^PZ`N{`d+!?KNer;ejC0w zR;z2MT3G5gh?2e{cg%tv!X^X=OK)CA+V;49w4sirC~wTa9S=J-qNRqZ2|kbVKZ8FV zSgxgcHlu301p7${bAy0#J$d6lg;?;efu-3(HBD5iqedot9EFU0!*d*Cr#*#I)h^?P z`Gg1}ibiv_SxStOaHUQ#Po;gg`x1O|{gHke>aC~#)1MBP#1j3bWw6(YlHyAP)k4{^ zF)59JWtkD2lZM(96a9mdQH^SmgyfQa&*ePLx?IggyLR04kAXfl@dl4!pnM|MJUx4B zr+i59HjdV}(y4(O-$s}1MMA#UWuA3~GIpaQGX@|G{-^%UU+_`Ozu0^J3IpOT7ve98 zp3A`g9>21%n?vIEQ%Fn%ItLc&lssZB`S2L?xnNxN#F5j*?3J> zIGU5G6>BD>p6Rx~M1M&C0JGQaKl^Ha-@g&>{B`h;M$vSCgi8eWeofjRv|HcqW=puQ zBVxCc^b=ZyiZxa_pLiR(iT)eM;M=9}cZc*HJ-k78JEh&%hc6@0uB}~JTt(#h)442i z&UfIB3gG6vGyVzn{{RIi_z{_2hmLH1Bzz;Z@bPFSHe8?tbr=;-wm{jH77VPxvR#cTxCn#6jsi9~Ws;T_L!Qw)v}h zc&PzgoI|v>;&Pp9Wo(9%H2yNWvemRr2Smg$!J%IbM^AFEh}2kEL}dAk2b2XE1eWQV z%GcrXo88Ny>9R+oXmRe7S8Kf?H7gn8MPHp~mhH(z) z*X%C!>m{|5a{Zs}N#t7=a6yon7=_`+4r-;xi(s2nlf;_Nr)8+TXl2oKe-r6LMP`hg zMvHwsa?E2x)z%_evA7)OtXNriuflq*@0X_Q`aZp`ER(@^X>)zzO*Yx*f1Bi(i*z{U zW5jBRep~jD;AC+!YOCT z+;y@8V-Oq*xC52K$gn%IU9Vh?Hx>y-p;@h>jrVu90N`8&W~f_ z9cxhdb>VBT4Qmq5AUAiO8@ycuP=ewK3r!TUSS+%Tj56fSh9Ik^I_PS)7K)eJZ;17K z>t6_IYSPaSjWn(B7mV!PCZl&R%p_t05gTj=Bhs8xhI@aK(-OFuYxr*LBqvgzTmg9ol;AnmnxU=x|+J}TA({+7!U!Be5+B%;N zYH=A5?~or35)HubW)ZI#O{&Cb=yHPy?cc#BE!CA@2Gt4wvIFh?T& zr^Z%d+aEl03Cv5jQNU(CxjidV9V1S>)o%Pln8*ZdNmp%JkGtkO+((R?YWT>`ne zk~^DUwM=K)Mn37!IZ^?wKNe}9@S0d@ei^sC(|k=N5_#7)de@ql@mv(xo>N(vHn37d zhK+-k1Qi?`R$A57*NUzDF$CTz@MVsS62+}&wutFdY+_ByODq!;=4XstXB-t)$*mmq z6Na}Z)wDfU*TGg^6|=I}uM<@(b8#s0{8y&U5Xg!rlU0=+K**DFpyUjb-WhH@4Xu1g z(>@mXetjZE(p1{X{hO^{XxCBB*lVPpO|gw_nZN;*v*ZkqlpbjKcHcwsmxwj5irT-2 zJU^?=ZwZ|>PYY@iY8r*iaDujMEtUwummLth0De_1T2JCVPS?N}dXI{=#@8mfG3mDU zzBz~f5w*B^LGwYLJ8=GQlp7l$f}s#qpJghCEAusOp~Vov8?P5>b0vC`$s zvBK)wpNM=f;pA@@YPvzY@iJUnJ%+yw9v`}z75mQ(#+;L}mn05fAY_8AjC4@=S5xra z=9g#TS^QV0+RX$~Tpd5l*L1s|GmswVQfcl1i>SblhU7PH-fcTtPaa3%-wogRpTin< znHpNXs_BtWr5`A>$vk(mqsTWFVELDgydt(q<1{Y`*?d9xqp5iJ`wzo9rInxW?XPql zYHOWB_6X1z8cRTl637^B0i|FM-EMeNSJ#Geo>;nHFgT(c@f*B0={y*Dt7 z5tkY2?>;a|l75i&|A_CZzHCXofAHr58ECCqCNe=zR zWY&z_@Gpn7*ghX>dj9~!jT*x8d$}XG7d`;I`#r)3VDnE2)ZREiK-12RxUn*@VMGP`-V8}}|5s|kiBM;)2h-~K4JT;|w z{OFp6)GKp$6kab!vDH<}1X(R&i^&7!C0$np0vA1vpBnsApTPbu((El$#eN;o;J3Al z7Bgu>N|H8R=4FQEWxh>-;gkhzV*s9ONXF(8*d^d zdin_T_wyTY64~84xNtGhwn4x-JaSf=e}=W6jrSLRBfIeat$l5Bu!(QHJ3Y3i9B?K? zV)==Ql^Z1`83Q=vf@_A+elL6v_=WL)=fs{KveNuZuC#_&=CRYREwyj5LKn+3`QBWg zyXR|&nLx(fy(>>f_@nV}MDd({8rJmvS4*Wy=81IE^ZMs_mxE`E#)DS2(KQVcIWOaBHA|gi_ich$z1@?%aF}M!*>*zh=zG@;pBHsI zUmjcdI@jVg)#sabV3$m~(oMOMBFhML>zLXAWRDp_(9eJl7-w}qO}z1jh2n@IJ_%hQ zUTWfYy@OeZMYw{?HluClTeHN#;nZ7AhHm)XT@ix&CoJko?#`QC)ULcmJDCRjliGff-JnkFNcwfbqKM<_7&1>PNmtz&p*Oc0mYX1OiaS%T=m(VoOI;j}m zs~y6QyGR)Yq(W2m3dYi_a`Fp$_ZvA|_{2p(IiN=WIP=DAH? zTTcmVLtVc3esuLZjBNUyuDa=P`?-0qH2(lP)un*}+bnAw?Sp`a85b1R<+&?+r<&=$ zD)?=#{A9G!yiMXi3V4HA({)euD`e51N!Bcx<9L7KUvSP%)WHdinB!2%yCmd;x1xMC z(EKaodkt84f5w^@+BH4m;^SNz-j{M_3^v^)v$e&%S;PFXMzOF8TNwsTf8yZrH;b&S zwC@~Cs0npg65mR@(o7a?g_TTtb>yvX?IzrWkz<(dz##KpfqC%WUkmGB+uk1V+vw80 z__%A^U2;Jq=1sepbiCzT<_cJ9SgY-g%C3r`Vg7ydM{g5$*A71kxw?#PNL zG?Aog7OTuMtP;GkNi3sw!GsFnouVdaZk(Dw1W zfy`t={{S`tV_I6Z#OhYAqq@1ggHlC~KpZOzjW$UlC0Km2Y>k*H`{S_I zp<{uA!~QVw2aNQ)&ll?+2iK+0^$#-2(COySRnvU3NqA(;q2-A|BOX|OrGTwHS4NKe z$1A8yf8xCgSuO%B+N|>G8cpt8KbCeyb8l-DgMg_tOgD4{0akov1T7}4-DuYF}#LFGs}N(dWj1v zfLVXDQ@Gpp5c2QePBK&>E1EZ^3 z^6*7o@lEx()FV2Ycs?uJSdwc!GV)!zlzZ^V$#i3!i8pe;D()j|9C9uA!Z+4$=hp2a z@tx4})X8~k8qZ@H3%};GoKEhlMsXZq1IqfCP28oa#_2G4FGuk^UTAuLjSbzcz@G7R z+xeoq+W!D9>zhHo34m7I=_$cHs_w;hy2pqhhhGwFzAD!*Em_rsclt)_Lky?OW`^?8 zFkryQR(37U8Avrl#P>RWqh^{gi^6wyx3=juotCR1O-WIgB*MifXwFF>vc}5C3`JG8 zy75kdtB)61Y8uw7sC}4hvXa8$^Fdio0W&R={iHG~vhIyh7`EnD6=;TT>7E79{3ouP zpA2a~*$riN=DOJHGM1T0BmFB)xRxjxcCjTk6WwxgT;Gg5S0{x$RSo8|;|qJMT|QvU zFZ5@#jwqyUlM8K9J+Z3@+ZbrDc@9A7)%;w!(|#nyrs|q6fczO1-1cGOxV*W0ElxlI z1ME(*D=RM{e8L9eeR!Ng=SY*Zf7T=ywv?`FC1Qo2DkQW`;5sb>X#+XJXkO zGP0`ylY^SMD{MvE-5R={g?->HT2B*rU&8v#8fsi4v0Yp3aoi`|#QJT{*?hLn@APm^ zK4Nj4%W!9h;$#zzW5JKArP9f16qb#qUdI&R#OB`kf|HXOU>9HvcxGOsE^iO`6XJf2 zrFhfBI$nk2DGHlgolehGz0&9VL@Z;JS)vWGIxr!YN!8JY2_R;?-}c)5qr7YT4t~a7 zF8!o*onBoEZwKiH(!)l-h7CaKTAj`<^!TT{futd&xO5FOOUceqo~|nGBuz~f{{XMl z{uaOBzJInAuk9oId-#jOz7EvyH4g@SH=5(adNpKQY~z7(dmoe(WmPSZ zp(@y~zAA25Fqi7ssj+x_{{mUi5tUU<9BjB;som%pD(5ULU`HxJ0pZ@?| zD35p{Po_Im#xv!KbDSQ%xT#=L(u}BV9I)q++;LvS@6hK`J+t6{!9N##WAWyX@b|$s zPpRu#?d%$En{5KKneQ%0SdoGP<~1jL#YR>oX6;|<(fcm`$=(e8k$wbtJK+BS!5s=G z(X>k#eA~v<<+N>H@nZ(&{&*duwYrw#ZS3WTIT*kLAIGo!8zcK7%kek%h}QoAYRzuj zSdU2fJ*Gz{n)*BoCDUvAzncUGAu#e}Yg-?*OC)R`BH|V)keUAG^qZX<##$z$qgz|W zuIpCVxIgflX|JceR?K+~h~q+X3G*OK9OZ{h{$It}gPm_<(H~WkVcco+%c)z%cRm@? ztj({9H9Kauyg@2ymzr7BEu#Ex|eS_)o%~Fw#=*!d?(MzLM%* z9YasMg4%t+>~+R%pka}yHVY$ewo*mV^QKV@1x^MQRUL4cyCXRb&(QX%8^Vzv* zU~7c)HYvyModLlcK2cL?dWVI6B?){z;#P-zNg1LIg)F#u zRSF&(Jx9emNgnl@*`90TT_eFh4zbfmi9BI>;rnZaW`j?<&Hb&s+k;6R$=wiyw&H;e z%MG~;z{%k){0Z?&?@YXX5^XcY+G8+tC6l$ilyR;Eoc{9GR$Z#08h@1Mg2dO*x(18l zT^moI!M-21J{BHfo;^{n{{XhpHU(CS7P<3gRx8MjlrJQ!16fh&ULWy>oi2mojay6c zTHF&ojlHyILBEpooz4%K&LnZqnGVLtB!Vf(wKjb3s_6O(+P4Kt^gV61z{+@hOVz;|I+?Te{VmBdQZT=_$bGVd>O84iF*!@ zuSWOR%vR&gnn>chxs{AShbBo*Fa|gT0;argfTK84H!nrV>Uw=!`(6Ia`Zk~YZ2Um*SBb9@)*lb*b6DzEF)~LB>M}`?(IzsH z$-HG-4qq*kyEXH_?L*=HdqnWYm8IG=Q0Q7?$847IsY23}KfE|NV0Q7w;BrZ>bgBDD zr>jeI=rJ^^FKK-h+56xB00x`ssi=O~I{Ha>_H#Ccem4|HQermDqhpLQ4BT|bJlFTl zd8Pbv@cShG8(E#6x&HtLjQ+@a?}>Gf7Jkccj|_i_zAo_V{{Y2J;9XivRSbUDrnEQle7>lX3Gm&Bi1Uy@>r@-|wf&y- z$>Ox}KZ&C7%d zzgh8A9vN*~^5yj_D~oGnSzt?L@?La~ONk^5cM=ZO2Q8KW{VxgOD%8hAO%uKS4>vZP zVTY8N{Cw-T{{U+-kk7lz2j<)lI3cibGJQpJfU|tO0gia-?_D2^mL zsa`!TF7T{DEG{ zlit79bKwuff7vtk4gHq>9DdEdI{1s>%@uqV;k`FW32~{x_K75d-3*Q7#lAw!`-#Hy zU*w_i{UVdZmMtb_U$okwjYr>kCPl_M=Yc zz#gBwk>LkDiDUTJ=RA9XjY-qIx@>(n4N2N|T~F=1;*b0mtM(@FcbgA~H9dR9S2pF8 zMWkt_-AUsBPkPqTkTcX|SLaW}ulO&;l3GiB@b}@1NgPUdNv>%DCRINyE_lz^1}ur_keZ!&*4w}5flCi z_3*pIPp$aZP}KZcCECqBj)CzW-%qr)2&G7jN94r!9#JZ-Bbhqn0!?y98$y$A2D8Jt zXxaQPcsoM)-{Q?yK3u?p4UjUy47WjH{EIDw2osIKh&4am!k!9iySp$HXAG7WaBPEdA}sCz@HUFylU%p(UZgxTrEAunCDu%k ziFAENXtc|y;FKMXV)A64b7TnB(RcDoXRke6@#>=z{{XYq(~10Td?^0_f?R*WX8c1v zqj-;4)%-VOGNjVOs`#E(o62FdI{lCB$+)M?7wvJA{98#Tzj8li{{Yx)Mff{lmi|BS zrkCRH9v}{}N-hLmVP(KJbKV){2j<%%TrN5t!!`QL;tz)!pNF+Zhr?bN)HI!MQIRg? zys|^9*~KROy1$(wM2^_`$C4B&$ZYi2ywE&T<9#|03wUQ!(EL|zc?5$}m1cVvMQx+^ zA7Dw8$Aa6*bzpI{fnH5~MOwS)dQnc+Kb4>OI4|KupAb9=;XjF^N&6I3MP@;|n;h z;)})*r;{F_t0bs+^5$`uL%_it{{X zk()lNrr!8k(!$}^PbC^Cp=nz!<-h#$3%4bDzDsoCzS{U3;r%~Mx$tkq4-fcW=Hl8$ zm1VxPx0Xd%!j&sySB6t4RoXEk6(kZ4 zF<+7&Yw>|Q{uCToRYd3jaXzDk_)pTO_#(E!% zyhyTXe-ZRM9dp7SA7#9cP+QBpy;f5Ud1=j+tVn**mSsp{cF4v#I1~6)FZ@GuF0JtY z09@5H{Sp*WYouOj>tZ);++$g;q8C6&+{ndUkeu`wugK5Q7mQmXjiY4&Zjx`VTt8FJ% zylq-2W1aA%S;Z`CweZUnam++`7~la_HOVxc9{<7MI4q4-F4fzVcv(T`GM;Rzoyl zwt0(gf+dLnAW0-4Sl}_q!M$(duZg-RhL^^gPOf}&CG5UZ+1^P8jw6hc9%GIBrQ_v7 zv6jww;*_M)C2Vfrcu&Rt9@j(>cv&^QYf@7z`sSEytX|!)Bif?o;qFAiK4Bv63&0p6 zzGVHWeiwK<_SX0bu71iIXN%KRiso2uCYtHLw5}}gbjjA@7a~b*7i=@f`@$TV22^&g z`$X|4g?u$*ap4aH%`bwyZ97ELS~RbC#wnX}#~9RZg1X898=?bmKtZoU@jjRFm&AS_ zxcGhX2T6leic=AJ%RG{_s=JK!@AxVI z0Ex8!00run`n89KBGI%@w_2+zNhP(pYm%!ZlAOTzCME^j=0l#~`cJ?=h(0v&w~u1h zrSaCg@gv5Uo*g&y=vqy>9SccRgt>=Mc6oltkQm%D!xG#yxZH&p3jMz<;@nfE2~_I5 z&&;uWJwi3*823MEzu<|Vu(!p(23_6!Qt@MHo)U`cEhr=2-F>;`!(B-nDlHo67FlEh zG*y?)WoKQSE`Ok(i$4YY6Y*YCM=>Q?F_AG{}rhG^y7?iEx4`yue}#dg-83pEcD_}1?F*IcvlES3g43u~*Ox3^Y8N1AImkpnEv z7o1hlC!4#K-CrrjnazA`BC1Q^y}oZ#>M|UnmL5`_HSbE-`?Nj?*FRxDh}x#8o*U40 z{{RhY^Mat;TwTd&a|^KC5?fOg_>d_Dg8J9!Cx&!8{{RknHpAe3$A~o*O*YxBby;-~Pkt=N_OI^4-3LAw|-I?Ba< z?1LQF=HA{JtwQ=rG#px_SFU>Yf_a<+gjEV+Gz@D;=!t#1R}h zlxG_VCl#S@uUh!eSGb!008z5K(k)dU$4>Dsi7l1Dc0;xybAIb1yRapn$lRl@Le*=Z z3tauS%fk9bi~f&mw&AsZ3u+eHLQOVVBr@AV!Ja75hSff5vFC=tsH`N=ek=Hpcy_~B z()H~M$7vdGh%{T4)FW$lB35gQ$rU%d6R`Oi!2lOp%i6Akab#9bbUSDGh2zU@S4}r^ zBzEa%H2U6{vxxRD`VFt08bG^pM%0X)1GSE8p3uA%uXs~Le;4>``!@Ss)ILRRFFqWW z);V`?xm!!yi@3bkZUk)lf-OoJY!F7}VpO)X z6DpYtWy<+rj$at#6GgeT@vQe>85>LR*Tl=)Ki%3siKpA!-<3jqr8gJXEg$ZcvyU_a zK?{uF4ydi%kHlJ!iS#`>FNacFKy#= z@I;eYYF;VSJR{;=O?>+sPaIw??WtCOnSxr%o4cy0oR!Ep$6kuxz`B=)tfq%f@MISg z_<|>rZxDDTZ8T(x7Iv4@n^&J39p)CxFI@iRL^2b(LeE3hh@WkxGk{u8T@4hHexjAz&MEi^huZ>C7LTCJoq z6bTp!r@=MKz&7RF%Y`}10Bg&&p9|?;HrDOF5NZB3vcH z7v328ah%iV=vsC=T^da%!@4~FGlm@!ePde_M{PERDA6zXNyW|716jCuivqxv1A=mQ zcV7|Q_{YQk2ZO>orOu_PH;F^7=lmraMBXE`mdpCK-AhfM6(QKScg-=-Hp}WX@3^J9oqQ1ZxQ(Y z^`D2*TQALLsOlyiJ&m{us|A^NR#Lbuu&+Fp9qQEj3zef{)4)3SgtSdJ!ru(^C^WdD zk*2zv!rGO#le%V55iV?QA&`{_+$nA9aly?y#@a@K#!S5iB>0pFc6zfsEaM3!-J-~`IesIkL*5wzKQnsUX2OPMv~?2stj zpU1aS8rZTWQ7q9IL^HqsCrcad6nfQrWdd28Y?=MgNu)Ddc&0pC_E#$>xOrivgpQ3k zFUECsjb|g_8Sev|(@2-8E)h)EDN@)QV==gI7@NOwl%o~#hGwXMd%0GOBRpl*CBcdJ z8=qN(e)2E72hGZ4sHVYhA7<0LP~`qI>RVpPrT&fYY?f%Z7xOxh}EUM zPw37<70q8N7xmX@7N_I5%d33J9@YGe;fvz)54^!Y&YeK)tP_((HZ_OPn4FBO@PBL@ z92kVH+siOE1p*tPqs@-CQ>BSmNHgAj4by`aOMqlUFZ-u+nR@vZ+Q;Ay41S z`Rqw<-<8Rsw??bN%J5uVV&NyeL^$4idDk`$Qc@TnJq z&Jua)j12vhzH$>lT4C{>oV!TTp*7`JYlq6VV+UClT&ToD{rdm#0tB@^Z+EV73=W&~ zt5(^%H}i^o)d0W}M5ZrmW(ff(J2)2et2_Q$lB&w(V{4!6@(9P?VIG2t4b(;7GoX)`&>dOwgf+swdX^xW zD|x?jakKc57tF$ajd(S*3HyR-j0CIhI=Zj*>Z$GNls;ocL{9qy4f2Za@(}~HOl{MO zYLY^d#!lY3x}U$l67?M%{I%J_{-#Z%`>u04Pq*I`l)2cPBAhBxR{Fg(v0s63KDKU8 zd{_eu4kLCu{TFy z@_}v@My=-tXN~T7RRXeO`jFu0H%An^8LsRq!DqF;Ls;_b$xXTo1-4`x%yACk59@X?2-=rk7#P< zu0GSTkQ<;1QNG_6zhX89#W88U+BJ}Sp}Odt=xF0~y9otrZnnnM1H#+@zPu@;7R30i zp-IX4;kkQ*ou!TCUuz*uWG#O{i@aXhTs0Ig_Pw)txX&)$+PO2Q@;m3V`VI9$j4Orm zO208A$fP^fyg94ixO8{5qL4&HRhO@0$w<6~FoiNeS{+2L$bH#5srY&Wz&;dyvB7{@ zd7Zz0K3F{X(Z=Wq7OnppH%c?tN)@W4Ur&s}oz=5dS|1^qzBWea7Fm35Icf=2o!NXH z23zYYc5nZB)EfSD{gTB+q@bPpaaeUSG_eA<^7oGMdT*8=Em~SJxnr{e@w?V=gdQbr zAwN$G__K|D)W18jo{##!%L_p?7|+9d)QZNs(1!5T#JjK3XuvAWR6f(> zaN^0x2)RYMM`Dtk5@m7k!a}GlTDCP}pnO1l$=cAPZ1lHK;PLL)+CRJ*j4sr}<-&?F zzhADlNHrBgIK`+ud@4}|m*A$R9}}0^b)jqTrqjaL-?*xuQ-A6)XZS$y?Jd*_o<1)5%;!#Zjbj~+mhi6zwCpd?M85k|LQt3LTI zT|p5GZA!TQ%szkL9L@Q?iyIVUF67pr2?q^-P2&~0EL5#ZXOkXv@?0^rrG1OaE4^>6 zU{REF0=LNK~Djg z?(FYsGzKeEzEY^F87!&=-Ed=a`{cW_kuFJu)U_FIyG&N~U;B}cRm={t8}P1mS+qTs z)!ISZ!2J)RNVZ>I4(w?j5|GTT6XBd^zn4yUAo&0{@F)wEca6Kjm9ecb6dm@)w3ufUlQRRyd;v%bc`2Ts9$Dp;3C9xXs`_!wRNGo;dGMI~N$;&Zi(cxXG8>(r9;QuuS+|`99dBf)z+|tr z`aOpIJ%woS1!tMWKX0u;CVvbR0hLw4sCD0)Zyrdf(J|AXk@6KbWpmJnEQ6g_lE1pu zDR)!U`q{_G$ZNp2Tv4%IX~zQQO&xRxwSQRRbvg3-$$3-I&B%c$gV{l$C5QQcDy(IB zm3_5Km@eY!b(U%qG8e5K{qG9ev~G#!rf1E&CLdElqc8<7N(g8`nsB3cogejMm1T>1 zZB6Mi{g@RlBhfUz0F7jM_9^n~=ceRG3P}>SI-IPSrY-GYpvxv$XlNQzE9>l%^*U*4 zdA+^D$9n0Nx7R*cC~xBJM!RX$q|^ig!#b&p-py<%Aa+D`T(N#~n>%VUR%*;<=nkn| z>c~8O`hIzGX$N7TMmOB;g7&!WUl*K}Jbr=HjK*4lofJCI2j-$#nFYa5o1oTOufMLa zf1cOA{+ddx(^wj;h+2_9LM1 zgT}R?ue_?pO@Gn3qwVZpZ<{vHZ{NC_*j&%9=t|NO=a!5z+K&A)!38N|%8~TryxgS;RP|vFFCNE}m<^juzvK zj5$H}CnL!?TsEypQ)E^xtyfoKZ|)YFbN!Xmi{FMMo1q1U*p!Hj=M|-pVXOdmIv3L; zTqc68jE)cGR9DDmlE4O5n?3Q|ujgPE16RK_+r6xG+wAc;C4S)%a8rKia-!E0BrHgq$Q70WYyAr^a2HRGUF_+iTHUGRz(uMI;#)CR$*;oS}uT{wJU z%h|p)++pRDFFd8pW)CUak2Tr0nz?v-RC2C*tU$WD%$P_?;H+~xNj4-kt{bOW1iBZU z7w$zYLE80ZL|kDp+U3^njULj>EC1n%?8XdmUbz(3>Ao86P+BK^ZYc!wUAv0tsM%}#6uN$P9 zEks3;z$#3TNhJ_(#E}UX8On1-xGxy0M~YV1eB2 zpX#)#;53Q?YG}Q#XZ5omI7BCH&>0)c3+tv^Bi8GKG+Z~Z=skP~ETfjZay4!5`b+`}hMAM~7Ar zozbzu`el~O>O@hSyjqO{@la2%t7wJd%q79EdE)br9UedUd{GSA>NH>JVYNXFHU*Xg zmPZ#B7C7mlUsRW91B7La%SV!P$8KN9lMAL&f-q$mnOoa6%`H#}H}b@C9bwF0#_PYP zG~N^A#Luw9Y!hw;mo_v4{VT@-q8M>XOg^AMq!%Yxqu0u8nft5AoF1MHVp2GORF++1 z*h%2FocE@wK?iD><=Yb#W{B_^0j}QXeE^H%_LcceX$JEmJm`M!Q5!o*<}~>7^;p-_ z=o6kqD+MaCy+fU=y?3s-@JI5@WnROFmWZ{7`jv(v!Q*nyw?cZUs^eN?k}l%O86UA+ zD4BawsawI&X~wJO{3mG2#*Y*3sY6b6z+@*@+8t;>5ZU@t^Hw+ufh=AsCFCY+lWSn| zQ#YgAd>KZh3TQOoNRd7p21?Y`kNSYXsPT;vmk^JdQVX^Wa|jK66I9dYK_pu0x&k2) zeFX$v&;QI$Y?I|0#*h(gfPZIsU#ig#l1a|8>_Vou<;Gy;f>b=bRJ+OdlzQmeTom9J zZc>D6y$kScK+oqJPL%^?j(t6I~*0HBK}S>#+1; z_oP6i;gSGSIT^LGO&LKIqxcAoZ#d_djWsL*yvrkwQzvac-0W!PgYGv$pIB{qOXap- zr<9}la05b(yH>w~mFUYa%^3JrC*gUYxp^1I^gqeQ&TjcH%d^E_N7r3-zK==uCqRW- zUYlJvjK^NEi)Fm|VJ9VWl5Y-9-YG$@cT!E>y@{{T;-PAP`s()^x>Q5Gk9Q<>qS=Ny zL*I6`IZ)n{3vH$IiW5EQ(Xxy$U3h=$s~meO@z2zCYtIn zeGF-a7Td1^)Eyfc`!#^j6H(>2Jqhl=h(Fa4#k@$ang)AbDh1b(R?ROK-74JThxXf( zcW>qukcfo26+{*+DGLFt_&A@S`AQm=|5{`P2=itJRXABW-XW@ZpI>Lk@v7V4Y{Ra5 zr@GDay!D;!XueCR=kY!O#5nT7V^|Yq)Azv7v`3+Yhy6Lrw!onLem{qyp1$M8)>V|E zQ`@?V;Hg0ig|E+1cFHf@4d=AC*7SPxy?k7$3%V|nC4*^RRu!0tAcn9}~+kP@hTPo79PivI%pN$Oe zId)o6$~6^XOjKyyFBek|5MD_qZ^hd7oGmS%{z?!?+kg60K05QcAx~14aB_m8SV=DM z0*K|!@vAlhCtX@4aRh! zu4^mVxg*3GBI0K9k&u^W?xCz(LWj`}WjSSiHKYxoWtr`OGqucs6pXgNCQM7}vKZ%W zPda^GO!D<~nZg3W+*(#TAE?$Q+eBrN89PGJv;qHpKTbnK`V}y3GYx8gZ1-|RjPlX5H;anFWgsOuGr>!O3hno|G6 zvuus}W3BTfRo(b9Vqdl7I=9^D^$?p?I3=K9cjGdDTC&DsbM1F4kPfy0e{X|VBB4o# zEeU@yG%hz;$8cvG*5l#!|46H#P0?($<89yUpM1&;QNLtz8_&2y{}z}p@8XLiUSS^z zid^rrhXHu{Vf-769Zqgr1C$~p*k+^(%Cz8xQ|3RLk0)XJO?QPAePYM#!%X{zeib)H13LG;)Iw z>syWqxXi|UM&{~EgI>4cYRFuj`e{FV3SHsStfQKcAGT{+X?X2F&14=vd?EcNGz$9= zw~9Bx$@`=)?pVkA|HC7yJ5m-H`>43shNho6yzyH&q%>y>A~_bx=;5Kw@X=)MVg7BE zPg(G*8^6Zo0v4D%eX(3-Oy)LMo!mdJe68b;Z2Itk8P@z zc>S)~YOIK8rCS9sYq{+iW3iL36L0f5{dfgjd~kxr{oab6-jQ-aaq;FoRvyELir;cA z2-&oc+8{j+#s|@J86?nwZv(suQCTp9O14=A4p{)eYfhmrHLX`1ie zEWOz)FNBmpeyGV4MMveTtmv=S1HOt| z{t$PRW#7hzX58EAhXp#g85%ih1v z6pD^hkj;6KWM|*j`j``h%2ye8Q4B83hfduoL_Uv`wGZ+?f$)uVw4sXaA^W|Eve?oC z^;*8PzSPwfD|JbLf1|nr!oVUwr7!JU^CJ(O002{0XgP-bBdLJ1ZD}!CMCe74?7{^j zfD*R4m)TP(*7kg&d^TJ7?}b+WO$ZIhQ)c)g#i9}EDVDE$b;T%cOQfj@&CJ9}IzE3O z9S!w46KXc}PH@1$jZt=`+idz4+y0T#1xS2QJaNtX&Djgq)+e^dy(_4FZkg5Nxx~&e zm5k*J@lrPHA&RYn7%t*2o*}0G3H2rp&b<>}m^6YCUNyn0OPh6e;)ij9>zY3p;R1%s zaeeYlQeyNX$K;4`PY&4e#c^>7r-pq8W?eR0zndN$+JzzlM;kT7KQXXicwH`nk?wJ!5B3hS+#Q8e>)1Numc zv98IF-q<5f#GIM}A71MTFYF(6KgTySZgw%|$Mc{>)1l_GHLL>KJgwp{)DDo@HqYPE-0<28Ul=M_I{d!S zhc|JI9YrXViBg~Z(v%Txc=t8fsl1w`0DkvfD9>!Pu(u);0E0+Ne>_=(LI>WM{G4w~ z307)G-EB2B-Q_A8mm3Ul-M+*%d?4DRgagIoVnf`uIJSzQlC=Jd{v0sw zxYP)XKIlL{>d2HggY9sHn+j{_PmNm=#Jy_g_3Seriu<4>daBpjfXM?^{P`%FAvo!a zWSDcTZ}#TG!o@(4#_D%~htDjy)`@EDr@zmC0I_9af~pT#CnKLk>a0{71_knLi|i^H-WF<#A1(+Uv{1bBwQD+ zjJ+S?S87*vb_gmsSHykN2rNGg^W@eEYl`8L&h;%h(63#!2QFl#T${cEh2mtFCn)Kd zb!{FaylW%M`TNYPd>#xi2nk|9uOb%vaH*VMSAYRb(w9c7gGn-A_?I#I)MxKrRIWk} z#q9^(u?Gh9eB|zY;q0J$F zwa)33rx@1Qm1>BZyXKTG+z(GrU63yrI}H!zP>H#Ja&S-5-LA`AcT!xRdZ@8No1N>S z@SNai^`$KKX8Rx6<#*g8xncwE5+x_-wq>GDo}i;6%}kKnZ)Thv?(!tTt=~K1$6Lht z<*cq?uiXN&u=d|_Ci>yV*W~_R2* z|7^q2C&zz*QwE`5(i=aYOIBZ>luO-_MfUe01f&LyLE zitkB)Xh915%YgX~HDPF?cDb$AnCGur{H>QC@LWjGG!@!O@2i8QY9aebi|CqQvNFVh z9nSt#UbeD&NKrL%xe~$K{g)PCdDL&G=ghzQQz2IIqHde)o&?8JfAnO&O99tKqrwJ7 z*!mFkuN>2EO%34y*&QIS^VzPdKjUxWo}Lj|T&(Zne|Rvl<44t4wMAuiNDgj<#DzL3 zcptt#+Gorf^1X*OS=1F|;gi|$zbkHvGQVwr3np6B~c?=9-t7qcKy5Jxy5 zneLTTy7V?AH4q7VgP*#R5sOEu&A#0krX2R;=dFo~W2Vmy&#gS}Rf}cVzbDytL57%9 zP$3CBfuMKE(gkIZa4r*|OXg4P56!U=nd*s%<7pMvG*&$}P%u#DZ8OsKiGy5#B( znL0a6(l)41CQ@X_4>;{lQeUC#=GFOJIxyv!mjYAwiwftZ?2*P*so%Svp#}@KS8!z7 zgix05SWgi#Pu5;{0g>1qn@^w4j`X;bqaG&j-Y~xmAax<+6%Q}VI z^qd`Po-NuKYZ^Oeo zUesKBvxqjiED)wq#3ZWW(ozm{T3d58h{S(NNCpMG*P6`12I2m0GOS1Lo(?OM`zwqD z3kpN(*67UW25u(5?p&0_Y-P&C3{aq@Q-plTwCTn^5kKB*m8(3}$@Q1F5X}l(+`i`8 zA3+x&0}}5gv8s#rPlL7QO6Rd^H5;pk7VyPK!GGg_WRr~u9hn(fZj%<7;StAGFlZk9 zhj&zjR1k~#C>p!T;q?Yh-|WxYZ-RiB{t|7LUueroi07SMUDxKH|LhDca(lqYXJY&L z483m;b31Y=g0u~a#cnh-r7!x=_FTqyGA@{Y93i&H+|1gz)PiLn6syi?rbvse);+q5 zFJIWl3^PY26?}{)lE*0;}Lg_Jsx0h8IukvMxnr=(%&fF;$w7n+suZBqTL2>eA z)YIB+pgFIsn#YfpdQqEpi=%=g-|r)6nR}86Aaat00IgA8>A{WhQ07Y)JQVsa>EA3f z-6jqd`Q5)FQS9PlW@bwtsubF*!Wqhj5xDg4!g!$+Hof|~_qeiUM2uqP#yF zQ#XmXGS5v~l5P&qjUpMfJq@$K9Ib0)!whY=>9JCDCaqMIVH!V}7L#tG>DRYp$C<3bftm z4eEZhG=_dqlxwh1SGu0njFw4o`rOMt`uehY5!(Cd0ZW{BQ}ZU6!0ePH_$6k(2ok!X zi!%SN2=prx0-ku4e6U=n^;%)vg(sj5 z%2%T8A8o#|;43AV#Jm+yc?$mIcWj?GXOi&ErA;~W+qP__`q~SozQj`Z!LTTVTciMr zE7k?FeA2<=#?Y4T)IMh3@)=redf!U^?!|TfAgIQ0h}IC*uwjq8fBHj6cg;??7sRO7 zh+;?HqpT6ybi6KRDFq@KQSXTvehGvh^Pw7-4Uw%EojP0AAQ6Ef^UDPqn4YTQcx5^7 zQIX}Bx6%7#De^h2I32?z4aUAV6CwHhaKZF%!X!0n)zkwKtP`na#h1EHR-u=8E|!k2 zo$?)g!KkKRldL`McP!9n{qH|KRP-7oV(T*5v4N~dVJ7ufBy=?FaDcr~C7@(NS}+RkDh8cW8rRL&GyX3H%P|pbgeR6MZR^FZ(_adG)1M zR7(2x6`OPGvar#UR0{%Uayw>W-o2PAvE- z7Wu-WpKxL5vWg0u6LjXNk}GuVPB$u8w*tUR^V}Kk8$uBC`4>wZ8Mf0&5?tx@zSWxG~tnH)2-;t{0J$DI^r)s>}6otnkoUDX-SVmd9wi zv~6bhYl`xeXt)7&1v5E2`3Jj;eHE&EB?#BxO-E{@JI-!;EHbrDEEzB>#hmlDS1wPK z74Su)2QkE>L^3E?9;Il8cIri>0TqL|?E9x-#d=7vid|>MQkSB0H7G{T;hBj~AYoIm#g;zze|XjrT4=*?@P}gw zgmcvTAp#UM;%N0=FSSQ(nGYtuuW+n5{va*ELbazPW85WsLWu^k5&Q}j`eU2Sive@h z7vSGM+DG%VbIzp&2lLpovI;9sV?SShrcBgj9J@%AXFlTJl0m-Ft>DQS(ptKAX_ix7z zAIPKaZ{czTX*sRilN5IL!#^SLo~wIdOX7CmcPQi${HGWk!YhjDSffk-(~UJ^(%<*hf% z-O2;WF`*syV{AR6$-C{KS08vhI%h zU`TcNqCkQI=q1Xno2h=z;sy$QEto`KnY2f#N!)*2an$h&X-K0hhgytLR6=qH0)v0# z2`acki@tQ9zRY1+L1A2ui^E@hi1N>?gda+5QzMsvgWQ*~>{^I8uy`^r&kLvfM-$0co*cs{a5eQ`qk|k=ZaHj9dDb&C zcZ`u!za{k<1H|(!q(`8z@*kk3v>{p(Ra#n#U18MpqJEk+si81CuIo%jvOLA6N@YR_ zik`)_Sr+|v!qMueu%xE;5LK)2=12w@b*yUhfqpgJwM-u=vzJA>=&O~4xW?^*DAtUj zI0`HmnWRRXn^w_6S%=vIIK83n+bv=N(i_e(PNxUiVq*iDsW(;V=quzW`L)wIQARQ( z6}zzi_AE6d2H1q8lEB5i#E-TOFCxgVn;nLSYcp>((VUSkpb8)&FEZ;g>dGFYg$w`q z1MtMt>WDJIp>di+h7{6i?DV&(65(xPahl-(q!lZYhq+qh-FDRSn(4{ld1pIrLim-- z`_GGC9Af<>UReDSx(bZOzVT0+^fW+9ZbF86`YVd{DPnw~hL_7bxS98xqgdmRIGF~B zJJ7@u7aF*(9pe2gf3a7vk{{D!XFUxT<*>qeaU7L=7Ya)Hf{Gy$e~nNahq~U1)wn@? zO`ACE5yAe-E9##AkHj9hCMMX>Eb~+;R>k4U6Y{q;dtXXoC zLOykWr%eB~13FFn8vno%dSNuLl7ogm?kRZ5Fw?0$+Q2Dn?OWNi;tn-rj^!-VMNg!7 zg6?VB*NF7H7qQ+Y`VJNLHn10(n?>(A|IzE|@xuXfioM;>*66fh_$}d&oFJrq=B=|$ zV|CH+0MyRdeykf&FW{rkhwT+hq-+sgZ z97e-m;WPeI`hZm{w9~JHWKIs=*bdYKB%|i~c)B!9Kv1bUTFuATmes~*a>n`3WVpdj ziKwhET_Sdvn?(yWLsrB;(1>H(tC~L=&*~U>S)Qw8dlB50n~ey)KTLSwGzd!eB6L?~ zL$5=>TY^|~4;NDn9A|^I8n3M@X|h%xt$L3sq~n76Y;{O|R3egMXTt9lu9PB(wa3@MQw-=JJwKAa*6j%%&SM zPnz(!7phW?njOzU^R59}OIjufri^o7mgbSEn43yy>_vREo|xG}f1vLmHX(?W2- z_O)lNpDmX$>Zo}bDU`@RmENL3@UTG%7VuQ(A;9VT^C5C7DO#VkM@{v;kZ2ura!hS4 z>`r;Mm#NLWg~H3Ns{it;J`%CQXhnM)6med3wtb`_zd$t*60jBe895ez8pFEYUc52M z7N2M?oiD=rhU^SC;$v%nI{p^BH5L3TZ4fke|0I|OgYV{D3Hi-I)PEygy))V+f~rGw zv=ZBPIpe~OuH(KD0xH7U--dncJT+;&tK6FlgW=#h5ZCoyh_Q$yXiTLcb_;UD6nDJ$ zQ0@F7+ClraCgF|8e2>)&r~XiOocOlJLf@Vof@4q;MhAGrhfsnCVY+2xFBfY>(hvVe z)DN_aJWxi%DkQj8xe*t3W@ZT@KAJ z7gS8{_7svpWpYfC7}dc9Z{~VHY)y8p=$saA={3=Lc_Ejaz7BCLXW@zexC)z3SlUJ} zCf?Hlb8t*$$x|wYZ$j2VT(BX=glZxr->Y+5Z&(Mp4}$QtMXt? z2q__6@)BdFQlV zZv`VS3Y!UT6|Lz&sX0Wb)|QAap|nw42^*w-hH(XrOi7`%hqy~6lm!vU2uCEvSH^?l z$)cLm#&FdLcpz^AOGDl2CHU<7J!9%~?Yf_S6O6}AhI9vE)K8`**Ho+{d0XZcpPoLt zEJ0xl_1VP{6ZA19*Ir$A67)NZ9H_KGPT~f%qkBGE)&SEWlI1NNbrp{Se00foFU=$t zcw3pW+>)d84GPWa^b^aiviN%!*2lGj#nmu+~O9GB*7-*Vn=#`E(+?qZQ`jj z2j0uSJeRBpDu)RQ#4#bY{ZcC|Vq=Pdqe+EyOYBtlO<|U}uG|zyG&|`=ZZqaqBM=9n z&tTypxGbYDE?tvKjdX3lt&{^<85C$8#)@cGpUOS{507VE0jIrW+7gQeY@H4$F@o_e zn-gS~Js+O<{o;H_`^0@e8XD{9YUiKI&cNlC>5A?*?SN&OZu!1T^())waJ*FR$E4i? zAU_OHghi<6&6PeU7gAyb_SB*!<9x3qhR~4)RlSj;uaaU6xNAr#l?(%YxO*V{m^XtnJ~t!Var3rHf$4lu!=7D zcoT1pTI?fOnC<0LG~`<~vesv;DenPS_L`i`kVORWg#&69mmOa)q^QZIInpxMlZI4z zhucipTw`>;XSE##kD)cUA6+GPaoOMoQ5;w`WM_cI%-b)RK+g~HxyQ=AzRzTY6s;rs zC!4j1vRGFl6+R1VJ;v#gx&9@>pFcw!v|K#q^p5yf*MhTQo&LJ#PSa|SQvkKQq`Kq} zdP$H~c-1V^o{)IQ)||YB=F1UEP^o`#v1h{}s{!gLgG4lwzOH3HQX7`1Wu?vmQ{9_4 z(Q4&I$zqk5c@n;5uP(SC87FlFTlPAA(ioEtiqe|9VSMbzJh?v+ zRuPS*7V1-0#Js{W{j*Sy(SAd{IMdu=K3`PO*WSj1k8A%Lt+YynzrW7GYe+05^?L9V zgt5R4>qPGR%){XD!+G_Q{MUnkpuwM*pExX}K~E_7sphgO*uF3}#Rl@L4JDpiwEFhXJwVEgU6t81T$U(}s1r zAp$8C3G3pV#z9@M%CixYrhotBRGUM433qfe0|SIE&epDWcBu#mrfL-}fJjq@TL{Rq z4JPp`tZ`pjW?GG3arbV0J*?KBzYcZhAiMDNxv&T7PoU-4YzMg*2AOjEF zv0x^bq012X^h$OCwU-rEG@6AYXw@97J6Nh7{=RDVAnr{?kLn0s@{qP(-y55Eg#9?U z$uDxM1I=Pokwkv2KO@Y)^aQf;cqOTSBa8p^$3k#jF9?v3U;e90^nQ-(ks^Dsd4U<4 zz&df}`U+I)@i6Fhii4I4;-NRZMfx&^_)x(dHv5q~Lu`_jqJ@t!aZprrfjLE^8xUut*>nAuB0#e zBmA{~3!(A1?XiqFAu|JNtspX(E3j<*9SSmbDmSA?Q}Vc{>$~C z7E=^ZawvX?`-K_Ae+NWuFm&!#>5$>@ri!E82(!=l1~0w^|M-5A4jC7Ji#o@Vu;1Tx{6YEwwQU~89ag`(Pr)%$s1j6A z23|>A)QW6d$uVIJigNDxRP&D(zW-bmid}I24=-7+-+hcFkmf}@k^5uLq3M4Gm}gl> z2V0ZL=fSKk9gS1L3=ZYVM-wJbOM`;qGiipQLthl=2gkXIDYfK=7|{P3<$snksIt}= z&I)Zk8?(|py@#tzL|h6?tGF9}Qb8F%J-z1~s(^%9V!+-Nhj8f?^`^C^rnLKembXx! z1M&BXbkD448Ymeo!%FLpptr7?w|2pdICs_6X3()fzdXTvp^ZSknaj}+CZ3N&q?tWA zmYl9ME zKB)M*_SS01I9Fbv>F?XMEp)~=d#5Aif@CjNZm`M!3{miiU>0PahzOPjspR*kq2xz^ zQ=OX9ap$3Pmciq; zK@A~A+(MK(KbY6j6CU=(UFND#KyW`$^s-FkMRu^gs18nv-qzfpRkLiFnC#-Z4tjzp zvV6LdY6*=vm(W)>tv~&7-)Rt_F(aj`#Arwx{A?cEO{tN)HB$()Oqlt;^V!$$uVA(o zsZ~Ukzbpe5_aEu$_3tKAej2)jyY!QNxB8nqVetk{?X-HF&U23&$)2Adf2*!|qao)g z^NgKl@(QqR>$Qn10Zj+EXu6(Q=@w(sR2x$Fu1uqTpIH{&DzZVh^N}>5_eNkdM^evg zgZJuT`S4?oO_g|+rlDglv?>*CUudsa!=dWl7U%kct;t1Rd0?&ov2zkXOOnJYsyM-- z2D(Q^Wo}?FGH~d<{j+P~8KCOEV{-5mm#IMxD@eFxM1*FQ+Hb54(cv>x&uC3^YxM-w zKYY5#L6&hj5u#Qe)}t(PQj+9F=OI!)@QIv8N3`6GNOqmU!E`A~!s2l8OBbP4un5MN zDgsNw1obKfypHGn!3tQ@0#Lpv5O)>HB424c1IRMlFJE1iv6(G{d1nK4FUx5NKbRa< zyw}})Xt0Ws8x;JDiA-2v4sfZnveh}k0S_pm=^>X``b6U|5<%TnIIH2A7)LW+|!Imp)D;T(D)ab(RdnSd* z@sF{zeNwQ}q2W z&i?E-R9Ea~NWg0tYQJx~HdrB?(5l1_c0Vp7A;`5RzlNi`P-+v?LlaJZRVQWsJA1Ck z=7eep$>I9N@e-n7{6)cbgfefp1e)Ca05Jbha^0thwDI{(+|h;FwhHyq8is91{=vMW z92pXXD-eWx_as-ILTixA+*+mDmc=K8B>eCX69WBLH+-kj9o{7e`GTkf{SK*~T%+!G z%a{IC&YAgd8h?Zv^Z3Y9KVZ=9%V$Mhw)?wbyh17=y04pz5=&~nSO?Q;m89OAt);Zk zK7TA{-uvk}kvCnj^*665MLD8aJ#j3@A}+|n-Q;zHp_6a9Bv9HXAhYwK_cN0R=1wqj zFp=ab8^ZqZM%a9V;aF6%j_S`<)kr1Kx@mOe@pW;M8{idt=}N#P{ing~-=RjJ`i1>? z)9LCavE+N&M#i7xjB`uZH~U$Nac?BPRu`x8z9k0Q;F0wF1+;}mBZEM4!P=XNK5ny9 zk~yA{0SuiP4Jhfpy=s1>tkq|}WtM9JUdN2Vf3jf@saPNZFAJZ2Swh+-;CxO;$Y;!G0fp-HF``kG7dBRx3wr@=!C_$ z!uJ9gd)=RN;M!eu|9K$rGB9+kM63R%N{3iG1^3p=-i>Zn?~T>6@&<bh@&xIqRCg;NY#1u}3s%lIpai0PdD?R? z`XauXCa8LVDeru^UU=s2YxbKAFTSW5EiNewOf1@~FO;-1dULUVg9Z{{97gkUq2{i# zb_qx~wpT>^E7MCZm|%>q-QP#S!WExRP^#EGXZ{8G8{k~2py^?E9S=5Ll8X?tKVZk5 zx}iQTTAR7HF2|A5A2+Tu$A}BBE^4$to2ecm2(W~`7x94$$y*wsp+i-x21gi!|ogOWAU{d#(;Gkx_{tEwk zR|$_jF+EF7stf`R<{xFr|It>5fQ8upZX^OaI>nq?? zrh_;d(jGF7d4DJcfqju=V{TqhnzFePXER5V2jFeH5>X#e7p}e7!M&*D6va@pYSZ1I z$>o0QKlyaHHJKhx+)a67TwhHSKHYX?R21Rai|iKq56>kmGOUTP25s7} zLLVvw8S1lGv`+VLRQh)lesDmrP?9V}!6J%L+!X6tjk}A*_})5&&fGtk+Qe|76#6W> zc%qT^D<42anVM)`=*qz{eSp`a>aO6oinVOR+$tGxqHTP%fgqZPhCB!Yv`flA&TKxv zuMW-Y!#>eANhs@voCbAf*|vE?3DAZ3*%$3PyK6FXddRg-#cLx}E#h)7llYUIxO zxD&Nr#96*6Mk5kcRIVdhn6qYXwV9#p{QkUgke#LnN_;`sT|9|6y>yzxNp%aT!5wa> z!yy|SC8xGei!p1R0nHzASKw3!SZbHaEn9od%?QVz@LF3Fs?H{5538>0nT!f+WE$j( z=VC0u(3AmzrI@I}mX(D&>)%?grO&C@B`5tM)_!W}r=p$v}<2B0(kN6}gMHT7^|90dghM7l#5AdRH7 zfRvPibWEff1L+#w-5@Ou5|eI_21)6TQPLYZVDNkI`v)vOH_pB1JkR+)6A`?A9LD?C zNfe1;vCxwJApSuYsE?|9^QQ&sY75^E{anop+B%cLr3#r0~W`qB(G95%v_7R1zF$LZ1Fn-mf zhJcqStTz4J)@>anZ1+Z3SVlGUohF2Lw*-wrRy$NH2c>eI<4Q>AmPCYGJUL){Fh^;f zt@(bq85^<7tQfeS*ZP8WwhkWS8Qms$c{9#0YEHH+C8#(LJ0(-xGpuV+G>Q(fg?|ZQ z!U@lQ(-cwm>X-7y9{u6H;KNHG0SjN6811NqQVQs#^Q1L^Y8dcZY7(J+H24k+q>#0e8)#VD$LtKo!&rT?RqE17Qa zIV~(=UO+OO$e<4L{$<5H7c%3d8-wvMRish>mGoQW{pb~iZWEW}q%l-vDxcOOp(ljy zh@Cq`0_!b|Gj9z}pnAvdH`Ds0DZX84^FJ(n&#=3l#}~b}{>&-8TuN>vMTOjg=Q!V# zIPu`ino*2MWzB_P1bxH|aPe8o&bX<6-Ci+1%za`jDOy%k{a}&&!J>oBEwR?(-DqEe zd(F0=Hp~X#PkMD(6OnjE?GgopsTv4L^+Bo(>4+>+zc0FuiOg+NnFF@w%ZVLJn%|*c zH6W~+Q)%P)jR5z zXtV3s2gSJ2S(MqZ+xGKN#Hfy^eW?TUDS=2nQnT~vnx-A*VzO&v*+0(Pz2-g?CAftf z)j1+C$kVYNw3mFO79-f`d8w{aEjfD`#~eUMnrPzKON17Uzw&QHw>fzr;7fe^D2?Qj ziC;ENa(n6>()wHi>njTW46`bN4L|fFKR)Ar#yUG%JmNWl^xswE|C^GTq6fc-Kakms zKYoS0zM4QY9ee$-O}{pT_P2FZ7e6;H*o;d18X)S1Oz$XE#B~u zLl;3YsD`2pKXrKV<9AJxk%@*;9-ExbGfU_Ke(qC9UpAM9@!8>yc?n6O%G<#csij^$ zR9Ib+XzV2#CX5oilkiyF_%yC)@vH0@+YX$RDmYI1X53fxbBA)Snm@-$`sG{R-5-uX~uHNOoPN*43VQmAVzubAtI znhHMnK!nIC>?X$9K!vq6f51t&j-Ls$T7f;dCNRYflVJ?;6Er>iXsc)I0JOvP5cH9G zwsWGt-2lz&K6`%Rp29bGMcF!)psE%kTH1f{`6n241R(f(I`ffuCJid0>aP!aAQMDN zP#e^82yEQ4tyR%1CWu0PgGABE6xg?kTBAhC1Qk05(;PK=Od%G{Ta)kn-ljVzQUm7y zN~*_2Pid9pG`Pgo*%7#blI)I_n~Ttf91XEGPfL_g`O`072%K2QVz3-Flm(1nBi*1N zNE*s<5FEmQ-_SJYZ$l44%Y2?W+O5yit+ic;c5Wst^`}TM$2(UBq{Y)&#DCWV&$7^8 ziZO_2xkoOqqVEto1E~NM6ak#BOCBy=Py4amDCrSm1H2(0oi@_OfNiMKqg)V z53KoK!jCSuE~a^fO75hKOP=}HSLNC){^Bhm^SRT~J(WH&ZvmXm4ss$9T;3kS_);vYo*UC|H%{a=v^ z#TgNoc8pM4AVj4KA!??V>cTu2ps_P3T*)CTvCw-+n)Eld-RVBVyYZHdBIRXqu8M2p)CVBvLvzu9wbeLj@O*5{_@>eQbA8U)jThe ztjEK4j#(^L>;3XFAjzZ6Wx`N0PFT`A zN~YI{nit73(HNC0ndn^H>0{}URs{Wr0Uw_QrX*%M5R^vVhph4`g0Mx-XHV?kr_>An zj`4~>9PQvrc% zcg>v%Z1@`ZmBoIl*yHSiYlU*wLf+vrPXzw6Q%OsExBX+PCHWVz_*sK-7Wkkn$5E|i zQ3+G3?S!T$&KgVk$e+{QVLj)%$bsUY|K_@j8b!o0O~-T|gSHC%cKBpJm%qs=u&2m^ zNFdULF3KpJG;zM3b%tew0~$KE#fyKLJRfG41kV~qDb)#=NxWM}iRd*A470H6Wg2gA;aNNz6j381mdv0U_H#=|pUAyALe;FGxn7hIaOvZ0+sN*6&cbe;b?zn_+2$+vjOZGcyrz9z_ z`p&B63``t|w@Q`i^Q?ODCL;PWL~o&#Ca5q^haG{pt$H7WS^~l7F~s^fArXC<@!6=62fJ z7JgHf-`#P7HOc?ft*Y~py8}M{!4%^6rydiFr@ak9dUkgYcHcZ<084q>U>DO7B-LbD3fi~MBBoW7}>AA0-s2u2(Z^Ml4bRHHA z<4RgN4`>tyvPAY%k9cCsBWF$&lIC0!{p>B;U5FFjUVQ4&P!Nye9lbZ2#qd>Br;>+9 zUd*EeWE(8wEE|@FTr#9brvqbrceMK4u!5s9gvN+A=Q1443J$!Lf?u_=*%Vs%M0?5p zw$v|5H5JBX1d)_!NuPokd+E7)WvHazcosXj#go~sF`@`+D`O!$6Ao<|=mTgo zxd?y8QCS?Fz&%1H9F4>Xn>`5vUnQW4XGIS8{0nb z%9fEO2`i2NdmpMyO*r(Z6RQgi+1#CH{|k4(u+{W6Uc2WqMIF#f59q#{vnlNS@@@-9 z{!L^!HaQwtn4@>Os+@wcn{B*qH!DJu**z+%Z)}LMO^O$860*P6BOMreCewyPoE#?( z%R*^i$=r7~REW+AOUa$~T^&-}QqcZ=T>swe4gF}IRiaTgY2bj8a3dp+mfyvZ%WR*4 z>EH!^$-|-_4RC`*;)XtO!jzG`<>WtoOO`+VN_(GSY$;h>mt#qsuK3pTgwaFp2MwwV z_hG85_M`#DIonQeiO4e1Rq_GkmEDe}X*a6E6}pfX0fWSI4z_0>^U|5_moTJA&7;W8 zQ+%}b14(?$q1%RQhnk8|p2d@QAeMO02;S<%ZO^mrx7gIT@J+VzJ5dbpbhVHpxvA<_ z>}w;UqxIVTX+u)B)9=Q{NiqUuEo>dvo$Cj)Uzl&u$XFu;l`BnCGWj*-wq9UzY=u z6rI2_hd@>K>>)I|OGk*TwJS0UwehDW4aU+T>80pLs~mZHZ^LdrI5w|Ro{VK|TY za?I~k0|_6Lx_0{T3i@=(kxjxY?)7@ICc(R~PJQ!w5S1 zcv>V)zdPqKDRx$E^S26>-E}vo)*T$>nvfRu>+{YT8ygj`lxCT$gwAGMj!i>Iep3cf zS7jY1Zk#6-F~Z5u*P?+@#wB{uylQ_bJQQYP3BgH-%vW6~ZEepLaQ+dJeEeG9p8}xz ztLEl>{x@y7vXi|xd0V>a5hcxu9+*2co;=+{@!b8AHZAuB5x}56`mm^h`=1v+k zLv9D!w%!)R=`T?ew!0{NL-(pXK^l))ws2kjzBq1yzIK{V$!rvX8QXPGQS(~B_5wp9 z=&^qOoGSiuPF(Yw39(e(Jnyq;ka`onY`?Dp;uH_!C!YQ(de9I5?`iY`RblH?YVOG? z?oLePs-cKxN>cU5__iRf?b9cIz_!m&x`wR5e9HE(4b|3|QMf&5zGJgD!X@Q6Lv9vYE_6vs?2C?ATim~dhTh!- zZZpPT@GUBB?e%6Iv&f_22oq*0w((6~<~)^Lypu}Vq4Tl!N##Ksc4_t9Gg&%pJz;h5 zl^?(wqODT>JdXjJ`z7N*O?S7BmW=x|oF}vX@m9`S{Kkp=+=SdXH}%zr*PcLd77xLc zGB$h>E}lRBNmHYP?_+p%_y zH;yaPoT(yWC;j--P_gn-vJ#cpPKw(W3|x*DR7J*zv!LtB(edva<0kPXpcHf9x_!GS z+{Mz|`SD-Nj0(1|Y)Aj;|VR#y&T_BI5I729=^kNr%Y<~|+@^VpeR||nv>h3wb zB`9coAYJ8)fRXX}y^zX_q7lo+zt^)zsZdKQ)eh^fF;|?`eSgL^Qmx44sD_+f4|7#_ zFwvUVKo%z(_j~zJ#7h!6a*w50D@WQ2l>BoHXWE^cus(Hgg2FdKu{EE&z#X ze+tIhq!WdBFnOn3@pMwhbli6BggNx{GZLPm7gRrM#7grQOKQ)$x8f-Ww@f&9AD#E( zMEv=EqHk@LX(L(s=^PsQ;)n!`tro^t3X_?$?wpZi{Rt!iJ{D1mTQysy@4g9=Fc4fR zcrx^5d4XVyt#yn=13gU{%U#8WtV$2g=lW)UsY(%~Z=N{KbCt}Yy6r)|MA*bLpr`;i znH+>63;oEdrV!YxDx^G#a>6e zE;Rq~BG{;rL}Axp(8`LRiAkco=|3!`KiC1)8TLe3<#3$FNofX1CwsD9ExOJF zF?`IVC&|{v-_*MtA=BX!w9G#ck+pP42`87rJvOSBML=ad9TF`>}RC z@{uIL-SiL7b0r1}R9*-V#?+Wo-Lb5akMs{^FkwaL6Cxd+Z!viL18N_VO)pF`Tx)oE z4Jr!Jxc9)!awD;eT9USW;wKDurVwQr$+|TyOc^(*8}sF5;vXkVpX7uh;kHKs@+9go zvkiS1coR;%iGzsa0J>?+>457-_-)@(H|E6r z4Qt5NiQgIHwK<|RL~En8#&@TBFwsy&HaXslyAdrz{E&EdK2i~eR}YP%HwIy2|6#S| zG0JQtQOJ;BbbPk*0y>4;CfyK!heeVR5!E>5lEedP==S$fw~LL6GG0!Ey%%2wbJnxh z>-m1?#jZcEVu`+*Kq^+%{tZvWnRNStgn| z9#YbwcM>r9)=Ap_P6>Xrk*Z@F$E6uaV1*wCv9`dGsp-D-6IV?50^s*XOVgv}Rqm{N ze2w~12b6jrQ!)`gT&iN`Up!hUc#}uoe5BwN*{lA>HQ`pphp_4>H9=4Gr7gw&F@&G; zxhZ8T74Qgag03jC|jPqkfh z2o8AQLuvLAMR=&eKCzYhaK|TW$WViWdBiFkd7+}J zKOt5ETbgMKQTuuUL9gwYKe2 zNtOL6x0mwDpo^G%Kp*JUO3om>Q;HC#0*ngdawSl$$}kB>fD%>2pO`u7y8N3{g`H6b z|1=p~0)7Tmh15I}H^&HZj}Skh#aQ7}*q~7kiOnTWEVEpH^`>%u4*k#~6~TcDZB(_> z^+agbMwWj~Q;6^g(G$;y^FOVbdQs#k%dyVvYf3!P8}~G=flmQi&+7#Q8GH$i2F%Kf z3L1pMC@LQWZy-%m(mWVB0IWkPa<|xs=ZbK2hy-^uc_$}MhvG^p_@wtvpxN?{bx`Yl zx>IQD&G4A&sEzCY_GuC#W?g`SwAy_s1qNpA_)Tv#mPG&EB(^CZ&D5@KzaYbX-y&lYb-O81P% z66!oS=DtnBvj8~E$@dTd*MOybQHd|gqXlz;5df}i&vx1BLvQF#S7d!ggO3roUbR3| z^622^t$>0VtgNQFC+Ls^CphNS2i#UPJIymT9-EMDr<{z1L5)r4wcE|FWQd-WA^ye(I+AOyAJx!1R6O zE?^BeTypXYWnTC@_?$Q{svip-4!-{(V_79{+5vbja@)z>K-VP7=ccyTs!l9|B&L?q zRL&stk)W}#3HB<_20?rI)akpp7%B0Ot>3infV0i`-z)|RAO*l1*P*8c=;WV3M+_*^ z!mrcTluh+5MV(SUD|&)Slw%Xq)Ka%Mf&k1Ha@SY1kC6ag2&l>=_r&1U%qYHy2l_SkFqk^RM50f3bO7(-&a0rY<@t?F=eeT+ z)eEvGF%(jvo~eQo|I)(Ov5y1nG{DSH+X{Wf&?xE#* zzP6z}wreQM-Sl;Ho33!>=~CChR>TSUZNnBkF0ZyJpuuPv807)Eu8!dkOv++YZanT1 z`dgNAe}!j@T%bB)E8lb$y>nCl88Y_sh?>+k{6YcAMKDz|8VKE?%cb-kL9tS zeQINTL#Ibrh0#HiZ#q&QFfbtF`%TDUn%t4}dBLZbInsYtd82-)I`VR<@Irgt7O)#N zH6Sw>dE8JpM+)0H8G0Bg!O|$dP$n$x)@`neK!qg$Rdr^7hv5G6EczP=tIzyt`@Zb@ zr*i{4Sl*onIvAlB=KnP>;#Ma5sQ&8L3`@LPeP`TFQKplJ8{Heh<@+>Nm8Z@_m7j@w z85%G@JO_2=g^gUNT!%rvRdpi-gE**Au$He(s=sU(Qcor(mjXb)Vy4y?@M@NUvC6%+{1>c!iO`hZt5q_+VB+nf;@>)tQW5WT-(zjo>cHAYM zJMbpno_NoF=4!5A6gDpj-(~FW3dxiSvZzHd2B+-)PID!w{ph%B1G9Htrv zA)jh1GK^UDu@)L8Zy``*;S}`~LZ2(u2KS#=heJdvKs7}aR~7aH_*Oup^)u(!`}oA< zdO>qB%ij#^cYNqU#Pw16mKe-ygS=rHdb-_^r#+RQH+1K}U`?sJytMdS0VGh3E_#$H zppT`qB04ofOpz?c!E03F?=XqBR+@9e;}cSs*G4|>hEhXu@FiX^)bM0{xf9Cx;_rT@ zc;{#Q>TP9?_ceeb3-Qp1ze+$>XLyxnOi5S(ghk9dEq|Uywf+0d5~{1TrAro{8Qmh@R=2*TQ$rNv ziL@tbl9C-Q-c^`ytbWut@to-PE=DYe@|B>yuikPCT_lp)yi*8lZHkf~saeD16ylPH zBlM8=*EE)UgFwimK5tR3$JLkVqiVO1kR3yldE!NmrWfv;`sE2R8^>EW@Gjsf6E6J} zjE*-&N$nloNQE`{(%?&(avl2Q$R9ri;44s*52G5Ikp(0@7#!<@A%U zbswb66ITy8p8PfM0LnJiYeH8tanU3w@MrK8o9x@M{kPv=uI1VKnyhK5KMzbrw^?5) znHR(4k<}@k%*d-!H-rVrNz-3A&4^mu>zrEVwMJBHb%dR z^A)-RLHKdaWPC!1M%83VUr@rb$VdOmus+7imO-lTn*6uj11GB2l;xn4Wu_s_a*=Az zO|>GZ<$+8|0E=IoSMZ{`6(+t6r1(+!+iEvV$<%-&WROOui|`GP(8AeHM$f#@=*8-F zjkz^7-SQu-%?e+6Ez>rcRY(*+{xmb7Iw2@;0{JOJx{xN|=rnQZI-;J5;0Fe@7lLa! zEK#@ua;#*eud_|`8lswIqzWWWU>?5du)7cZ^Cvt$sZWhHyJJ~03Ex$f(BuQ59%`na zXwGNO=vl0RN`e0UBhI!|8D zHb91}JH^mJHVv$=j*?8I$3d}mcg=`~8WK*-Z;H6j$=Iit6Tfe*Ltv~v72Gr>5s^+6 zrp)Urz8Sbns?96Y4LxZoHDC2>xcj!x)RlQI=AuMwGN(R6W9*?ZuRZ*|w&nB6g#X(7 zTC;8Ks5f;seaCB_Y`>VRhHL!tE!O$KQ$Z7#LOeO+==_v7BrfW;GSTb!7{`tHXeD)< z1}W{{+n*KstfB2m{bZ2=$UkO^I>UR{Asok3ogHW%q#_ZP?L@_)6Z64snB->m_X5I1 zssY85P}G25>VASxL;{QCq?2CONd86oaA8p<5Dsq#Q# zF5z;T;Qs-WPTQv+b&<+G?XhzE(-Zhpf0cgLQ=~)aT0@+Qvjt+aspdyz`Ua%hA|>e; zC%o?&?&JL-hw#O=USP$${_2kH$2h#?!#@*!wY$u4F#ksu}P1g^6t9aMEr zneDHA-5SY{-9d(O0((^3d|1}G8h)w&q$xXFnvwl&W5en0E7?f_NM9!~`&%(7O#@JB zQT*YH>8oY*Gn9O}0j| z24BTi6=(j4Qkvn(Js}G@)3Xrhel;|Nql`DPMt;|^af*K3bp%JsbW_|ilvWNBmgG@| z0jLGZfokt3VRK?|-n{#L5FRcl^AA~tcAhZTdP|5oUWJ06yL7=khV!;~#*ZI8vPsM1 zey|8@GpJ{8ovCimprRj&Bml|L?)p|fEDBrAJ)FPK-F+DByG~p+j%2Hwyqsy8Hhl$w>aBSMzy#W8l&Kppbe=P0jVsuGO+4cJ5RTs@`9vY}W8rKnR}Gp4 zxVoPr_l{{1Ge}I@Rcn&s=6eW%b3yW=>4u2RniXTHuaJvflxVgFYgpFxS0CQ|B2Rz0 zR-h{j+=xeK9>fu3#)V&vL8E^}EeMuWWEzH9S2HkvGHc6{%4}Fdqy9&9JQ(-&0ijD(Hsb;!j8gqJm zv}CrW!BigTTHNt5dT+BS$~8k$Hh;fyZhkrg1UY`XZNe=eqN-+)A50=hcjDQa_gz;e z%#Q^Isp$YVW4^RPMfbGJZyNQOh1HJFx0HD@3^l^Ps)W6zx*Pn{ZoLH#za3}+&?&v( znM*ujidHVWImD@ICL~n7`{-ouZDG4sKu}(XcHr5TaH-2C;IY^4Hna>6Hsm9^cw`|$V_R-rv zVa+ks*x18;OCsUL2FqZG^kbApahcKT6Cai;v%-i%3;5Ee_P2i{sZV0U-*WzDaQl46 zgSUG!PNaeM`2myiS$_;8ndNDhE-~pDUXuuhQbiQd`WVf>8OZv*O5XX>**o)MjM5lI zC-TwIGupkv+d=R@ERq{@{YtLfvYSfsju^i!ty-GE{|GOu@uw>}qy_L1}^=KsSICNNBtiRLyP%VZT~u?XDRan^)s_1Yr5 z#^y2*Id=2f5ejV__E#h;$DGU5tly0qHAl{&|L}%{o66VMM+)g>2E=*B`Y4}lC~t%x zCqV5(1>Wn88;I|X=t}&;H)pGmiLk1vsh@`7!j+2E9Zp_N!=amLUIEn4k$)0b6#)iw zEz`g|++YJm^?euJ)SHZL!@Zu<@ar`x;I3J9t>^%<2_&LN%1(KY+AnP1?5uccG}aB^ zaVqmQ6{J3Gi-5^*RpA2-_6o#XYAxN%rO`8u+I7noZ-t6lr!;Oa5o4-X?K7XhFVfV? zx{d)ofTmBgjhBb!9lmE`JsP$-TA<#I$>y{tAzuwDO@qtKd_k*}+Me>9lv|ohEvOVh zUhQZQM^g>OS*8*0xz8`t84Qkk5X3UC)Yk|mX=;yQz?(85(p)I@^ze3VON2V}AxU+z zhpm^Btz+tozFyXt37q8s_K}p0V`{a)tjw-H3ZzJK>AI(c$t2EPVv#CEo$Hg;%ox zCQ$@z5vTi@w*koM88!Qco4hN#qY{Ao=a`_<1GX6f9~@Z^=^|okcdbDeSFDU|BdF{h zK+_ukh8{_xpUVM}UaxS~u7CKA)uOY28;tU!^wOTj;BDE6<&;Eh^4e5b*VKX=tr^7TUYCzt+O7 zsh&eD7R33ib>2kV}u zx4oJo2h#E&^q|$-k`#qEn5+j-?LNW52V+_;pXNEQF|NI^P z8iMt)b&dUdQD83<_BS!?U{fa)-vEH)Z`b1)I|i=w!ajkg8Kux7=~+y z+-m}zal7g~H2xyvC)C@I-+j!oubk&pW__{P#^57(=a{@-Lk-(z(+Q6=q=;lKuQz(DG(8k%11iF|AKP#Ne`%{(%k2IN@ zra^T@oep$(^|FG+SgW~{2PsuFKl?q+i7o3#CEO7(Tx~9%kZ`37p2*)fR`LEeu*;V4 z@i36M(l0_Yd-Xk@K4ro_-{gwZ))^GZtJLnLZ#BoT(SdsE3@PSu$R6 zOhe2WHS&MZq*k?kU;Ywjo0%LVD?GXMFtT2!87AymF9QUK;;SopE0L@i8mn50!Jx_x zcu}*%GQDImecR^DorNgPqs<~u$ZzES8VS-ZF5i7hETe&I(e+R(il5+lBuiWJ_%J=W zhL{vYXi9^mpO|N#v}H8@kOKSa0$O(K>N_wm;|1n{UzcK6m-J#sn-cvRZ4k zTIfmc@glTt8EbyHi>Ava{-rGqTXHFyuSc3CTyq+l#wCW4KD4#82|25R?2K2xqzm=2 z`hvk0TiaA;UTfdSnj;VUw|GQ&70=ImLdRk{!YCS8u4dcqrpiS!E23bsMk@dGn|l5zRH5hg}X!J$57=}#k^d8Qibq90D?1+OA`Dt!ev zrIdcRR#lAwDFlw_4-8h1|Cp%o7$6=<>5>-uLQ|a1C5?LB*fVeIHi*Ki2BRE>!8H1Y zXstW7I&L>A^~zx(ZN`kR%}xEi0q?kzlo(#%Z9Ac(af2}i8@r9|sJObaRgXmN`x zfdE;~4&CE+!WQ=Ze%}nFo?*v~dD8j2cK@S0gCOC7AIEkq7F67$f0CWr#8~3IfTZf~ z&Y1mP1WV2B>4ph=&lx(!P8_$-3uaVUA6L6@fNy_IOj7z)-kaVkHuy_wj^)q<*Rr+l zRJHJ!o9hG}Of)Ye|6yWwwd$0Dr;^&BnmE6$@lBGb{mOUTT}xa>iUpWb{Zs>&Rrp{l z$$Gy%s%kbr%C<1@wr3v8YdLZ!SmezFmR^;oc{feNb>#77l@Wq`?1J!a5qe35$PUHk zeASjQJUo6R&@$j22N|qVG&DXb)#5WDoQPjqAHcu2I{g56RNo#6|I5&~^>7b57~x@f z#6y|rMTyHA{o#?=k|NixW`sW$fKpX&Z|s_@Z@c#;!Ir%C_K6-@X3~2xOrTCU!$g;2 zqsri{2s!b8iJtGJnaf~H1*8|kN?7i_NliQDPfV~Wl@CkGG;IcBNz zv;_u4OU|0lKR=To);E25p9v(bn0M=`YygPOo>s_;$jF#&&Q44WDSmYUi5X4y%d+49 zVYzATPDvdIfK+*>L@@}%WoHE(`wY&|fZpiikvs>O%&(j`XX(n=m@ zo8j$t{zHO~aS^c^5HKO2Mi~E%c>|#8>JK|=-{l%CN&o&VuEO>qsELGwif0Va;KTea z;q>eO^crQT;XJ&Up}&*18dKe_aRJBr)g*Y7Z>h9j#Ux*cdjpnc&0zPXDtI79VzbI` z?uOY3T4c8dt<)}fJbHyt5UEVeRh1QF!nv%rizAs@pC2vjIsAUm&Xw;s;_O*4sE;H+ z9uF;RKad{ZIl?(|YL+T0gn)jKzsM8i2^(hXh#z#k`@HyX`Y;#o7ZM0=Z|-t;Rq0#$ z45uZ1fR7sdXzBh<9jU2{b0LET10zL`n>HJKWi|;rLRwv$o(r0R(x!|8oCPCF8&d|e z&8+0JzY)TMAt)Babs)yltIY3jYv@{5nQXR6W!*wJJFQ7Ifwul|%r}snQiE`CY71f{ z^?hn)%bo33+e8q>by{X#6c0?Dt&&NHrX-iIO6xEEQ?usW*_hYNW@l+?STBw&qOPhQ zPi*m4Am`;#@!dS~awz)B1nT4JhPDhdpL#3qnAX@NuP>KRrdAUYUc7}5%>XJ+5*ogl zmTzVa#JjUEO)b8W>aU-7QD@=1-DwNdXWQ_2boI4;o1ae41MWB5MF(@-tm5lilb(iO z=;OD9^s@X2Y|LA$Py5Oi^77m1?>)lcdC`#EYUTMHq4=p)tIH|?Omf&&Ut|~-i)}hg z-rtxkJYLAsDQ&RXJpJAoFJgEV#G3Nh;*+Tgav2f+yNyuEFnQBDLa+Z!F271oCyh3x zHusFOH!n1eB1~rDHc&U~jD6n!G09TFB)`@eDceqX;`&rzqkNioq`ngP| zceloOh2j{T>;Dp?d1nEmMrol-2Sn3vkJ3>B!;t(X1!u(DVSqglof?SeYEoSTMxAkn+g!o<-&HfS4I>3z92=*}O1y^*A_%3#pttf?d&#N1YJH9MCp3{V; zKf`ptE{#tK9!Yx?FT;pB0rCbo9Olbb@4BA~zwlMqY~~^%k?|1j<;};g^Cwm1mw5MV z9C8zi21KeSDs#XN!H?mpf1=*q_1`01@9r(D1xfR@syrl#!QD(>sW{UWKLzQDF-)Vh zAMo8!E(paONC8@D6CA{2Ikg*ySh_uUcDee2EwnFjDApN@DS!F1w(*Al_sbU?|6y6o zb+v}mAdNbR{>dYD^Q)snTTvd}Ly}4bOD6p;wiWxomhHqtn--T{^+@DB=bDy#%Ot7@ zj&s%RVIOCC9NWRqX8btliNI}*ly7XEBq`I*2!G&;h_0z_J zDE@IG}Y9*4B5KBZO)d=xH*@8ODlJ#8DR1-m)F4Grmh4= zM^_wtVtK`Xz(;T)O(v&M~d1XYwOBB)H z_MJ0hIEqy4&k%B-O_iLWLTcb`o9`vg`7=Qc7{4YAdHU5#4O*44+duq)Kd(nz^ZWFB zm_*T%k)$e120e3gbDT|KXyn`XHz{;1ud#@@C;!88gfHi*Ot0|u<(RsrD8ng!{IYu$ zeAPdtN}+tboUKbTe6VpwK?rb@?N5QKQ`CdpG4L?S`mOXi1dSPCZ5b>n}ZS!v_?4B z&IBtWqI6g8Z|-YkAo18M?=vT&+qFv$I+^O7YTElcXTC;z@Plgbbd{&!=|W5v+v1mS zJoFoY(Pq7+Qxv&@Uv|_$(wxg-T)x=0D`5D`(kafl^7EbEd35y&<jXZ z{NcI5oucQ9lQqVu1bQE}Sdpw8;pyW)vX5XxATF{%MUR&UgOVKm9@3(~7q)vU`ih&2 zWK||RkI~6F*?#?z7}R}Y{r>h#l0;m?uw18j^iHruh1oYZ-y4=|pKeB^I#Hlu9)|em z^2Wr`L={kyo++fqxUTx>p-kJXpd`DuU(zTMOOH55$fK5-n&MyRhpdf(va$hB!33J+ zBxiF0(8j|;YHH>#TTj0? z7$2r`UsN3E>K`G(qTPzji?Fb6ojP`5sUxtIgOA{v>StUf)-B za@jF}phi8d45^zEfY38m;#+{9ZbAL-Y*~_78^Z~TQy6$;Y}J&;(4gb)a;M*alQ&}> zBMe^D-r%isaV&u;qe=rC@XP#G+*t44FFh`om>F{8@;R20s9;K^xfzKT(}FT-#k zWUwP838Bc-fn1W=S1ESS%1nwovTHUf^GllJcK@C^S5_VmGaWzfN^``80D7O+dQ5VJ zAMF@IqMUd%R&ulvFG%~k6R0`Xzj4f;C3y8@p4P9U*&H!s_=vlMV8mRW|C1t?@!UQ2 zOx1tY42lxz#7g!%R{=(@S`Lzgdr3ccV)^{Q^ZBz%q;-v5^P~3}Xu#g^g#A8&uimCz z59K(J*%joLgQsH7kX2wutp4q9Z_|jUSxp*Z0Kq}#w=Gql6lyHMz6u`)11_`f-?|U< z$h5Z2_#i{V>sUCugzLR^zio+aJbhb17^%Y77}ZxJl~tCsH4@Z1@63B)?7&UI0Ms#m zFQTvp$Cx>3l9`jOPcPzqt(j&lK-PBh-f7w*w5L>se$C@v@F|%<;1Ik!4U~EA+U<1r z4kji;#uqivwS;ZxS?EQU%A2WCwcY=6@NJZbS=@0ygg}Jk`K3vZRb!ezu4~ezXZ?|kwd3M2Wj7-}Os5PjPD~OSW_72Sa`aXfZ{|mHMC~O9 zVITKX4RO8MxlU2^hqRg>Gc4Iput!bgOyGlSjO}y`1+{5TVI?vZnp=5zGW$TrsFvH1wMdu*Si^Jb))h-Cu2+7~W+>L!V)LM>^{iMM=A(E@oiAJt z1EYF$=F7xWX*))8X;1{Y&u5@8y^SxOH1l7Yy^c@oA*8$VLIJ0)8ON9|{uAg#!1^08 zobCHZWqx!2H^D4f+)5XsTn4<0Q-<*0t-We)(33=Vv%Yxq4-dClQtwYL2Z|_aUQS&c zRT{QNE&3JV*+(My%>c2cG%E3v;~FknpQX~03qHyyY{-U>UL^_-Jd(^ga?W7M`#vvh z_{;n3z)ibC!frfW4&@Q-ihtL^u8~XDHDe8l{yv$(ns_ZJQP(&PDyp#trGd7SG$3=5 zTqtv0B+qtwrdq5MByrC$xZ(So?k_|VC``p}j=5&r!x)9=)tSivEuZ8unKjPdHx^v3 zWo)VQTstBni24t^m-|an=-J~RI@UpFngTYzB@`D2j3k|Q9fl8-3^B_Q&346x6Ql}< z3wN*znebn?ov&$?*qsImus? zKZ9gKd@e!0TpOyIB#Be%!V(J{W=6~trbPr!LLao~Dzw#zzu=C&I=ZjFZ9I*xh= zMEZ;HceX`d>r-`iGZfN(qdY~osmE7y!zX_6d@DWfVj>LuKLABRy1pHC%~Qi#4zF^S zudP^Gi^*&ynE7zrZDf%`q>ML{8AVcBPmnPm2!7VTvoDQ*W8d0K#UHZw#mJ%l-tg@4 zOBSi6S(9z3Y4>VmvA%@5$+0Jl5XjA#V|8&?(+wX?YZhm8=-{v>yz;0zxdgJWDy(>~uRI0+00#E{ z*4`@6oBkJ1jl4bJ`Ji(my`o-gWt5CA$sjsng$Jl$K(Em&W*C3xRz6mq3a8pX5Mj8~ zZEf6K+b!DlFLN?&VbtO@Ku6#+^sUV{FAlWuCx+E6Zsi$}Si+1l2Rn-)I3A?e^mp*f z{tejhkA&|=t?|d>8teA9(grZ-b{}JT3H}Cqn~T{10O7o_q;&?q`S@%900grC0D@)w zExkS=@n^%?{Amj1yuEUHb!z|t`Ab{AAzXox<|Y)3lis|#nYAVG>UBd3{{Wc&I4)(M zL(`{iSz?|D$=@y2!U9wb@{1T89>eQh$HD&q0(h79iPmB8&+OfyYF;VS-AM90F{hi+ zz~sD$hB${ozF84Qc{w%x+xU0(cm0pNE2@oe;E%$u1=z_HNU7v_cEct!h7tX&K_=0_ z1LfKWIpEi%+UvRpi7lT>_?h9YKf`TeJji02*;d^InHylcv}mJ{Hrz2Ptho7=LFbaD zaP+hehf1Bgzw5F5h5SAL0D?V#!AU%CZy?sb9FGoZYbyD(_@ep*aGlI{wu1I~S#zDM z8jNJ;j(<-+3jYAXzrPB!4O-j8{{XZn#X%%?AG{hrg)HOKVY+18%Cg*F2_<|H_o7IP z5Oc>B`@65`n(nvZYkMskMYoCs%x?;b;AP`$G-_APUI)su0!J#@>yN5{6)|*DC(LcR!Sm9o+bPNb$C_ zqG+0pyV_Y_TtjZg;wc--M=Ks!5zgFX;AG~$eDQCJbUh!%9%iy_<&EMBjUkMnJx+1J z{A=c4jQ2KIe-yk!HOAfl0NS_Z$>>A#&Ns&YSQOYk3L)dl_a}dlYx8m$DS0@ty;zg`%CN} zEFc8VImtK#`+;1?!|&NA_MGwMH3>0h8f z3xD8}zqSvEV~bGnZ-;cx1n&We@gA#f70$*u-5tHDx@Y<`CvlwdT~yv8e(WIr=FH`m z;c0&!rM#O<{tw`OZ$16~p!=*5;`1?p31&F%GlBY7y#CA{F7XF~J}BGS%{Hm3Jion7 zGfnd+xRHhcA{fDtuebq6eAn&|fd2sCl%E2A9oxaL{D1Lgt>Ps}_sji_Xx2J99F55> z=4QFd{njC(9OsPSexm#-_&@NM;Kzq&(S8!>3*h@p;Ha}qf@?`7UXn;RTg@Z5;GZyq z#{}1nXi)kbv|kMar;AGc$-4F_)+n*!8aNWo!^&i zmb#d>)fjwA5mBI)oOtH$T~zyyQiDr^hR8a1M7X#~{Ln1%6ZZSw>Yy z2T#~x)}E)t=9#5TWmj6S<)P@>p0(p|0{Cj%Tk)Q?Hl(oJG*55g8!2^I7XyiJ>(JzbonQ1faJ&vn~NL0HL$Th-@YbrLBr!XuU$zurstZV(z{1xrfB~F z95knr&2Dr%`!f`d_ZAV&48@dl3Dait1Gxu{w+^QS5IC&MZCAn`G`oV`Y&FQdMW;pO z=HA}w!ssFmvG1~wEKZ7gGeYP_FmYbSz7Fvc$7q^O<-VS6W68AETkRfWjIxik#>QC4 zVg~6Vn}9NNlTmo%#oB*{wDs`!#m!DpW2i=nB!o#Wo=NA)7bFOV8=eeo4m;qAbstj? zLj%Q);+tE*`ZdOzV$HB7pMNN})T8-wcCmTkY#f}C2?LJ3d3)c8{t?rBNpIn=h_XG^ z_nqgbo_@z0K0LJ+w^2zKnpg}j)^ek&WD4zM_<5$squoW|Ei4vs6GI-b^IcdfsmX9% z#KS6nP$Xpp0!OV~CtdMggP~sQUlAtKv|B4j1(QtMsKV+_R#lD^GUXY06Gyd4By`Ot z`;IRA2BvJ2Nt6B&og&pQjLj@dsy)n+Y&`j9RASt=esdcr>Ot#R*Z%+%JaysE4q9qn zF4XL_2$;(bZX=HF);t}7e%BI&j7gu}<>O`t1CVfN)?W&IeXTEyd=IbcZ1*yxmm^M) zNoR7XOYCcA^F%E9KQv7V>$|2acEo&3vDe!~x$u>>u9>V#B98Lj^(3Asa87UFxm0`} z-d~yqau>Kwue84L-$TK*kA@y7)D(D!!8+a5+FZuV942SJw^;$qZe(rqMu#W*eW*Nq5gS879^PDg>o~I#%u1)OHS}Nh@p?c zUI+0t&WksPmPKf^iD7|l-c~#O>q$^SG8~`32_TYD)YUt`4EWPT&@Ap{(PHsC!hl@f z+uTQEFksn34WS|@+=q}S3|rf$R4Ut6V&QkW_`~Aw+3!}@p<|->>TOoqay-cG!MakI zJGO{fBTdVYps2z65ngfNpVHGUEu{iUgQUpW}3lu7&__BHZ zqSbA_8pR*>rl)Ndm#ynI_>5e|6{evYD;?Mnjr)NBf6GJed9MQ2d|{}3KDfBe{qCb{ z90lNlOLVunGG(y8X#^$;EqdH9BSzdOT#P%);P!@5^*aqBxJXw%7sR+aeGCxIzJQs%vz_4 zEhN@F3uhmPzqO)ejvKJ>e6hI$7IP3$(X;b;Y6$6(*W#ZPf5A3>FvT_X(@PRfGI@q= zGTv)hr#Z+C9{rBpaKV7b8%J9GDDfBVPw=kV+C2y2o|$p3+JJnky+}qQ-rTu`LwU)M z?;W|v2NiYy0Bw&4cxOSk@b|;{q_LLNNb$AGOL~`KzGaNfb0R*`(Lze(00%*e*AF`J z_)(otX5FNITYNeH0D@FMjPWAT3wY||a{SqsFH?*W$PNlJiS-*NDAQRP7rUB=Vq_p4L7JnRQU)d*5veoWn zQj^-f!ALNAuI;XwBZ9I@qnu>c96ueP_-zC05=(2~`zz#@CA7McLPk^X{gesjRFOv^ zkdQKX9V>d3tIJ0tIH)CfqlWS4?DOE5Py8Z2Cl3VPZ|*Lp`$eQ}BbC}=yBi@n`nKY5 zLGSV3#J_-^6Y!UbR?hRqT1*<9v1XDRi?FvKoRAOMqw^84e*381!0XL=uZ}!*;(rue z87%y3sq2>42wE$9xZ}EzF#yIEXo7&Fou9hhj!7r4hj?e;_r#xwUL_h|jWp}q!EhQ> zH&BV88%B36&{9x=&ekj308dMVd3BQ_ZQ97Yb*x{{V+7h;>bm9R!J)Z=$H`lW<3KqR zA1-FkVy;_`MR``C@Q+8=pgtlNw=afxQT4elC7v0$$R2vEgn|dkTON!>euC&<04CD} zTIYwnL8e>8=oK{SlgzW7!j4VMIF>ih8O)%14z=zd670SRY1%cumE(^GUihNG+kK($ zQr=t5KH6qqnOH^sl8(6l09&vqQTK&Itdkm`<)YmD7V$2V@lV6ItsFXmju6>xjWwz; z7{~gR8?%gUEwuLQRJk)7i?9Q1&Sny z0l7=t&%Lpq<5nd}83QLBmx>R9ykp`!k~~|j{{X^Oy<+n%=38jwDe?f8R}19Iorjj7 zF*)ZZysN?%`UG0^fAFt#`_nzEdDpf=@;GE@$sqZ*(?rEf9*?Z$8m8P4V8A1c?Zsb=PH93Bm-DZ7u@NK`_E$3^6s55xA=IJ(PFmYpH44$(~=Kd@H`U@aCs& zdo(^cy*T>>Z3wjw41fk%CMDt-Jm>hh80u@E@TbC02g~9+-wK&HyEUw0`iy_Sa0! zwKo!DOfV$wt}_1s#af?-{5g4T@fr#IS*gJOO5174Fk72^DxsDerwT~HB~DZTHmRnh z_b*E`$}T<+>(^JysA;|`u-7#Cqy@~;hzmyQ!}l`6%^P6jfHREooL7=~gW)%Yulz_Y z{w`TxS}n!Mf;~g*FL50FswQp~o~`beY_>)OK=9r^yPCaEl()SuyhC zCBex#;}xVUTSjxbZtwmW&+*^vbKp;hC3R%+K9OM}Mt0k4Bkc_+&O!4cA;(NBA;3T0 z>t8i|OaB0bKK}qfi%hZ9#;<>H?6;OSw@r5eF|l3d-uv@)*T05YSM7_WgL_*mC+KBuKp7q{T+A;GacI-l9u_`g%y^}!keu1i5w4M>QpHi?JRw&vVc;fR1Z!>V0VCZ)y zM_jix*HUZwNT(@vXItREioP%L?8)J&HJM)O)RQynf9U;k2FRi(VFy2bM;wFRzU1(4 z!!H{PNaeh;{@>H#j#hUOz5BuC_efDzP+P2R6T1PD!A=4E(1XYRFYs2S4yoeL39NA2 z&XUDNd&vHKU>p&${t3=sJyqbg#)h8~zHL z;Tz(&9vATS?zt37!H7R(U^v_jsc2eN2Y>^yJYaHb=|6@40JW~Y;%^N>rg$p-Gz%S0 zR-bO85}ATC?v^wYF}eq~!@RBr1~Fb_@U;_eS{<>PYR^;a{{RJiM6d_Mll(65%>E73 z<#=Mdw7cCDI(&tgB(U0D7J@y^lP8b@$VOZaGoWw&Dp_kVS^QD)zL%`%?-TF2w1Rtk zn2HV28DcFZyqlQqGb2V7dLhOSo9SLH_-xuHrFY|t$aVF%K#vqw-fP|APUQ)Ev3VF9 zWC3zIn&&(bulxX)##Sxx!^1Z^J@JX<78fmb1TwDD+m4+u`xCT42j2+e7$oG@5rc1C zKd&*#Elww0_@VnYcw<(I!$i<^?E_JJ`*um~%=UH{vMZ=XK2&dH$#TuG$MVf8o=7BP zxAE2A?M1I@7jk%WLeTB>i?t|pJNXTS?d6gW`b~(s3^$AqHUv2sCxf3#m&1P=biMbw z#-ZW+`wN*B-qPg0n(;#`Wn>8hPYk9$aEHobFf)@}f5e;r0NH=TelNJxyia$o+un#@ zZoQ?o?wMknFjO_nUUUqgH{G0V1K%~QB&}pAZh6J`fIby?3q-N;uf>f6SJL&Tia+1! zaa@RO!YC}0-`R;GTsCmbp&SgHjN{fk9cSXN5&Sw!zkwIF{xDc>r(f-R=7#1yvK_+Q z+*+(qtD!qXd0X}nWDZC**XduhH;i=82RGTGE9JYJ^0ADjt$6BtLV{ffD$H`koo9wZ*)aj{Om~Q0Kiik~a`3l^6Gix0s(6#cqA1);X+MyVDIkStWSB>ErCaYSJCXndEqJHL zFWL{lcamOd+DDDP*%Y>>IHSK2%Qdv102npAF^jJ?5;^JPNN)B90S%O*ypbm=N}S22L8vN4kMP&#J(oepuI!p%VL(I?%9!u zEw&%rs3ReVl0g~d5!dsdUGdk%&lg>tIq&As)^c(nxjP*(jD4BjRZ;3A2oG?2*DT%! z@xHi;W22j`Qc!mP0I62_yJ2N>oc{o_Mtv)~I4LFZ95Td3t629h+6MRbCD$~&JtyG? zi?3VV!86Hmdt%UAM>>QdBXztp#vx{H{{XmDkQD&FHv{oa=9A$Y`__}eI$hjcQLB1Z4O>dH&VH_ zSTEwVL?iuM2X?~p0aZR;JMiMMf8v_cZB5PW@+emf`?p84uwAg66+LiCt>?7T5XEV% zyv*KWhTEG!>KE>UTjnJ8;~*Y+s@Ara_g7myB+RkO#kQi5r7_;)Yp8BG$@Z~Md7aq>1w*QuA>b8c|KZ7H#?GkP_T`s zAAzWJY;?kRd!BiHaRc8x7q&A!#C;6grEDIAZwvzC(}V7Ku8J)e#y%sF4vlvd&cG;$ zAZJ!%g+)NilHC;K4)_&v>et315?jklM*yA7sJO_$3xenU-hUdcE}fxV+?gj5gjFI( zfC%F`>_d$6>@kkDi=ek=w@X%iFX1oQ>-IC!JS;We+Are9uNJ*;ArKDaZ={+l$lM>E8@naBa~m8FHaHnq{i1L_jlB5b`$qoDUk>z*FW_gx z{Q?U{i|qQP?xA$C8;L+wQyGFcj#iiN5iP=z*-GZTQ^h~GAMIc92f;TU9M?1_wzagJ zNgbVbZH=UQ2Fk8$qRx>*W(?n?~E-i zWWRk1^2wpSx3-G`=9f;?e4Dg}Ex}!+z*aaU8tlJkEd%!d0QlQsDe%vWbj?!U)QfFbeaW{b}Kk*`xjnRdn&*{Cn_=OQt~XuwLkz<+Q>8 zS8EcQtQI#xHtq#rCm?;&a4R_0s|NX+Diosbw)Fo1gZ^jGKMAzW7vTPv4xRf)d_>TO zhk2E{n&#>?KtaOh&S>+swtzgD8wYDQ86)gJ2maYVu^)%8VzuzE#9Q}BrDj_zi{T!f zGKIq;Ez-!_Sd5{HL}S?FgZTyVNB#-H@RLFD4dt(ZyiKnsie`XZ%Jx_4_R{1HjsE}> z&$Tn59ZXQ5Ameau<6rAH3^B@Ga;HaiGySN(SeYl z;<=qD$zPb&$tIEgM(E$PH--KvEEZZsI)1CALo&QEGAe}$hlrUSCy#QK46<&EBxDeJ z?D($#0O5VJ#pC@?RKL=7CZEfiEk1d*2S!nnwM z7@i{WKCiBPF#gJ!R!i&a(Q7TVjuE`1ie{R~+n2yighE&q1pL5z$HV^siXJ}juZ6F5 zzYgE{cgGvij*}j}s6nA$#?d<|dx`Y;bsITE>H{H0MnR3p2_bkVO}%3O08@7;^#1@O z=-nH{N#cJ79W&tet!LtGUTJdAsx(4YRT)&pjDAu&jDp4%3^?1sstu!jXJ#!Y!@Dxa zleS$uZhPZ)o*N#S0Q&P@ZlAR@k2IIwH1H0w;fQqRO4ezx(G zxulU-%O5e=MO+>m_!}@2Cy%cfKg=0CN2-+|_fwOR=yGeCv@Y-Z{{X->g}<$TQ~%fe za)uue*jZX9hpt-o#Yg;m+uI~Ef!C8fz59$0>rc43(e;Fy@5M1LmE{ep(q0X~fz_ji zTw#CS=hCrnG+j>DQI6w6ou+uk^0#+VfH}r?Muk{*=~Y+7Hb@!#LmYD6J8f68v}bkX z05a*5^*G|Y>{u6Hu~@vDf#;plz6m7q1!iixhMnRmZl=_h z?#1I$43BbPQgO-I6o7y`=Nab|%iS|pmi8O%7HRKgWj`d7h(hFl6j3VfARdjMrg^Mc zd`aP3%|v)xP&Sh5va#~5tS!9lLD*#CD^Rk}GjMPb32wr;r1iTxAs4zu3q;m5%R5&jp<0aQ-iw=}voXH^hl~HS8CvuiLSxm&K;GQ_bq@uc=E)rI{wfpFPNZojM zRMD(0zR2*)W}_rQ43|vtk&}W>-gxBh`c-%>Us4wmFuS&e2M@Jvi|4P(!^{{1o)r35 z>?V!>00cw-0D_BY-XyWJ@jr*gixH8Q0Y0UqMHH@}L%?XTiL_#tQf z6b2jJLtOD6g*-2BV*n*Cul4&#CAbHWRtchX9+-JR?ntSOV=kmQCD-vk5H&3h+Dj{I zX(U5v4i-@~@?5YTc1svnR`g)!=~q&J6WvJfqUuv5F$B-sZ4BXuZHyWh2eD_}^fmjH z;Qs*lBlnDaEvDFbxAwaDyWziv8YT3`+RVWWTAi$~xa~yNVY^Etk=1|W+{eFmLYD(I z@6-PP!8U*3k$(rR;@3ZD{{Y%{;_@ru`DUkF{VM!J_`ClA1QYmSqe-E7L*X6Y#hdGCj8FZU zr=2tRU~h@#)HREU1!S9(~~LQ%n587{#L#m{{X>Mbr0F*$oOma4ERyu z-vVhxR5VuLS%IWkm_&`RW-*x7hcoaQKpyOa_eVvoi!QFzn%X85BMkb-Qy{| zV|{lYiTpirr)k%A@koN+Sthg>C!<9j%rh9!4!D*V3USEJMsJC{U2UUknq`-ZucflR zyJ*rhCFQrcb{m36bvN!N&hNUXo=E37{w#05;HW>hCyabrJVl`VP}MXWMr=c*SzB3u zWWt!omV#&|orZpP^JjKqbGVKM*1i7#1xfz^f~4v;QE7Td#f$9%$Z}xQ>~ux8k%7v{ z$u*71^Ed}^Va^Bu*K9H>t7{%@Ok{0z?tf&OSHt}}&hGD8(AxUQNUo`Es$4X>T1xoL zo1B@VBq;z!?58;^fDLL`c&|~=d^xFjqs5P7;oTQWaSL48-@&O^seCw*rie45fx|TE z4%XwRC;a<jmDEB^q&Nk8D6pAV;tU%&X@Yc{c`23OO(J1gsq+aCE3ozf*;UroKRGCbJWk*&qNVrDtop*J1hFHk_> z{8^w}{5-u~C&yYIqv4C2wRtU9SJLkEx14R%53*~#$(OVVzwyg*~S)1wNj zJ)~>_+79C_xH39prZU;|BC>6?uMK#b?R-THamefmTgzmdZ1N`rG-38gi_w9@4#0Nv z-`Vol_K?(m8~9Eii`r(NFNtondwDPJ^!dfRE!sw+NaeMXR;2p7K?ST%4HPc|fN84nS1{zf4wiw;C3W;u{TH z#Sf;~YEl6WwZ+U!1dJU4u`ddrQ@PdlkG;X%@E{tXrKf zW(GTb+0NGcJnFGbm_IP{PESN5r_NgM?5Xf7O@CSO#qFKvf_2?<3q9_KeLXCuMlEkqo z4QluI-vnT>_?M)4EB0#*PhD*@QH(LbKVxJt%BTm*7V04Q3aGWO@UQl*@ZX5F3;zHB z{6Lpa@IQ()%VRbElX(>Sb^cwNJ+atbL@wu#EAmSSk|*Se4lC>rF7oE^Txr+w*uflS zqk`TTrxQMRg=q#bpbQ+i2aIuCZ^S#_19-o|ZQvh{-Wam+XNUDT%3eWvrbVdfI4pLF zlzEcI=K+EFagaguB;BkiD7&V<@9NL+I8A=v!d^YRPY~!@)#Zk?lcr{}Ef!2ca+fxD zCRE@KF%q0{oE|!F6UpO!2KBXFBf@cN_NeM2mFE!Nf3kL5tWX>i)+Xu685O|zQ}$8Q zd=PHDSMU$w7O&xrH&V7plFsTSxYHxHT(YIT*x9L|Qcsr6b3Vbsrq7nRuM2p);_t)z zEqhY<$)M}LC9}{Z0u4geIIb)&m#aCHMsAY}joa;1@G?yilkR5Jc255Qf$ulEhli3| zuL5|hSr(VqY?oJdnoNe;_2%1#ZSaVZvCwYZtT-aDw0#zR8&00zPD#8=c5eRw(T1e) z+R82Q%x`xByiVc4+8fHq;fNfZZBOGz!Ot3LM*GHEWtNwvnIL#B?;UTfXC<4<`z6A* z#(b*-9APAkg3gQb=d{v(Drr|fA$?=v{{V_ENG_@^F6|+n!<-h{=VZ}0pvw-KBQ>K< ztEKl7DJ7@oK7SP1-FZGK_=}*Xg?FUfc@Ryj3v$aTQ?fW+BrJCWfKo;Tj+n11*020Q zrFe$lSke4Pa?s8j?KZMPFAO=)HuK?G2W+}YfyYYqxqdPHMe*N>?tBg7FNfDA#^LS- zo~>@ynx3A}s^t028mE%-CO0F-aB?z0&3G=Q;ctiDJJaT|_+jHKjRVP;$Pb5Zt>-~6 z3PkWe_~ZbMqA<#<$pbvr^r`hM#I2n*Ay9K8xbb zE5q7MdR2`6I=zAi`*qc?k*8dV6~0!2+GS|njx)4M7bBsrpTfQj_+tg7kB&86B0U1* zT}Or^tU|V0sSTa+#t2ef3hrEP;|-3L*$T0>qq&@+WpAF&DP%`qC zw;^r`URVg^0SX)d26LaSVC&xk7f#hC@K2AtOCFo5U)!wG>F;osk}|mkKh@y3`=DjI z;Eunfx<7|}GvMDA2sKY1-)kBoLnAeYt*4JZIl#E4+ePK!SX7ci*>n)F70JZbBnmQgMdEqWk18|Tf-AJ>~qTx9*4nN&+NzGF9syS zJzgD3?qHH6cRBI+AQyJ)<@6FD9&-`~GOd*Z8%KYad|RUGnl+r0 z>F~a%smNVsM2y-Mexym~uIzFN0GuAF4Cg|73sGB0I{!*d`+lY+sS!-XK5>8 zp}A)AmyyB^s2iT4n0;5RbV({0-{RTTz;eQMGZeI+BxM`N+#t5UF zu0Ce_J<7_&XZS*!xAQtN0m9;5r*!- z=M~_-8u&l1_)El6d@b>K7P{;b`EgomrGsP**|xbyaDHEzS~*VxAdz1}-QW0|!=fEG z!`2bR*2p6!_7=W3E&%zXnm1-coS&K_o}`Z46{4Q8I_bG5X&3wrFNJ(huAA*QO107T zT{dvnAVD-YGlxm$kcpwYZrwQKZ%AdKF3p5iby_dq#zuOvc6jE~7aYI>R7| z(T+kHW66z(!4I+e432o?d^zC{jCz#+0Pu@y)&|lUqGwt3+3mG0K2sP6W5aMaiAei^ z#!GfJ%j^FD4(|Rg-^f?R_p7Vh?TFjQDYu>>ouEsl$^cxE_m(vE=NYXcr1ZF)_P_iC z)pZ{N$Dnu??(TGjSw6=wOPg!(;zb`cGT4J6z;p7mg(?R-d-zv{bgzM)8rCD$ zJQ=HMw^nVHYj~a;t9Np^5-YG14WqX506#E13bW%6gC81v32ASwczViBVi==Tmoxp2 z`bjbaJTW;Dy0QKp%EW&UL0A^|8cl|a3HXC}o)5Y45E~4lEjv?ph!Q|x6ts%GHxguH zkyXnTbj)O)=;$?n9sEkT(NBi8FAHhcIxeb86`aL)Zv365Xm0$uRP+i~6OMr66~Jo$ z02Tfx{7Ufzw)$s?e$REKqT0;{k98S_5Tk{b$;Xmde-S%d801%=_$yV={0*yWqs9Is z)O=NBmPmAqY4oPI)I9y{Ssv~eNaYeKRY!3euF_bX@$(nNe~sD&{{W5bH2(k?{3u-y zU59PmmEFRhv?z^)gEw%P)PuK>?!beCoZ!_uS9Z3gG}i3&KNEP%;@+2{iFN+~4_x@B z&@N~4ZQ_+MxCPoQJGssqY8-+5T`FAz#r`eVygO&8*xK6J=#v7r@X9VNql=PuM+BRk z6(N9A&~(b+*Ta9bAMHcq)t>50PZ|v^+z3{|R8G+xa(v|rIB}A96VDg}7(bs{$Hl*i z`o4#0Z>#ICujz0|(|w)fd+T^zi4MUhW%5e_hVqqL-n%K_9OC5kIp>Jw(Efm|_5EQM zJIzB+w6nN|M-OA z`BfT8ri_deg=O4E1}pO`;U9s1GW=BV?E2TkSYWk+`7vuf<@A?!(U`$k*&7YO0thG$ z0pW?oXjuNvAG9Ubxp6nfo2$EdG`I|r>biC9yIQ;|a>%H{JAhSwZH6`-kY|HS2TeW9 zrHGr-$J5>={j>Zf7o5Hq@U5Pq49%U%Z0PvsZ<`*XBL&YNbh33huN3j`?WN(pUj0{J zw!hG4R9KQNI^`XhDE>QtDXnr(`==ZRAXkv+-?O*EpNN;n9S_E`T;0GGZ*4B65^8aY zIYBM7(A!)<>z)H69x8%c5x$YParq7tWTWZx{tFLn&qS%7S_NdS8J&ZTn07MzPdx z?lj#i!B!Im6W=wwcG7Tz%V9g0jmZSrJ{%e(h`K zWeeWS^qmj>3F+};#oB`Sr{X`0lf=4Z#3kVSSD75_0UHc534E*$M+&TS)DA1qJU{y# z{3bfhzK`)U!TQdTb1mA#4A*v(8%d;Q9$%NJTS;*b&T)m3ME6l#U&6omDL=x$hu$F; zejV^-)|af^F2f97V(u>adxnkBTO_J^4zUtJE0Dc@*RlP%uNkfNeRIXyJXX4hQE_YH zNh7zri61{YC8>C3Dsl3W4_uR6wVokcOC7LK*_~uRvUi63J*-LM9VYumwuK+&v(+t= zP>v?!1BixNWul&uUXPDF5kt)7ae1qB>Pb^=H`L z68J0NPab$yX+9Zh?QF5k=Hcabx`xtd!jo>$T{2A(3UeeS)TucD6I%9`@9_Tn?EW+H zMx0~Q*Uq(`#_C|rDv%6D&R@6~0#u~(A^W6c=D#L9Y5QJ$b@7n6(aqh4ryRpFE!5)l zNHOw}ZZ^xgPds6UYvo-_#Iov>jW@+t*Y>Ib-2+VV#zubZQC<0x3}YZ}+5q;g>EI^y zZpIVDtLT2r`0w_`_(9+p?k{aL{YJx3xo40y)^x^yFy|!}Hi`H-&fYsVdi>k2f5Akx z{eN1PPlpz;X>6(Fmw0BlG6orK_wcEO1~Ii{jDyJCo-5)%40wL`P>x+Y!NNP{XuQD# zy9pU`2|irkxNP;zdwzzx)8Rc@t>{)uRX%xjU4vpxw=Q<*$T1 zS@AmW?OI3n-l9MX!U>G=D;}i3%F!-IAf7laiu(uQ#D5F4d#4FwcdlNkD{d~V{JVac z%QQ(g+~D$1a0gDc^dkPyTJMA)y3;Lw9mfh>qO_w++p(pL9kThYv>^bIxq{(OOxCa6 z_#^rp&S}c>XXiqE4EV1bZr1dRyXNyZtf;?iBaV_hhs{uX5XDDRT%Gs9{{R(7BfYhy z&8*!DBmJU4T%Eh3B&b}2{9_-jewbQ*)zDnpBkJ-?XB?6#dv$p4CfkhT&9@4R$;jTt zz0X?mkBL9DM4lwGlStCMJesY%hzn1BAf58Ojkd}}ih3R-U!gU`F`MmfmdYron>q~wfFU&MbGemr=$!dHF^@h-JJ&XIi>H*ifZ^FGorlSV|G zFI}q2yx{s*h^y%~@hmgTEtI(|-cdHO$hgK2nYjlY)Pw1Y_0O{SNp_=M+}p~E#$<*; zY6v8bW0NkUwpElL%CU573#GT7*F&~9_Q=9PD#&DZ$m}j%m4L@PWaHDC(lJSBijFz7pUZl!0bO?tN;A(QN`GU~4);{;&g5E$}7 z$Ojp(KzZ&Z)eWrZx_#c@Pn&R_ZV1oI3apV5c*ribYVyxlNaFi^^In#9M7Btl0Q=Y~ zsN!d5Vn7NVcN2vi;+@2iu2#mENunavN_ToOk^P{hYsJuZSKy_=_&7t@t)g2{hY? z zev@D_Aa;$dq>POI@nMe~VNQ7R@p+UW`42*l7IxJBTYe9C1HpGUk$98E7MiZFr{2P1 zdp$nRTay*eL5%&X&vb1T11kdPa@j3`gZ5{`pMYNnJaMQ7__^@!RMGVNRU}Dcr^|U7 zD>gSw)-biwDBu>8g$Fx`$sf|+gnwpF**C-9G_}@8gKxYOs99VKm6ql1tu1age69AA z13sHFGL!O5GtWRrA6odM;2(`r)%8ydU0U7gEW%i{yOBPgfE*AbxVLFkj^MSoJ%|W=L4ETeocnU32-dl@Z9`jCJMT33K*A{ot+g`>u zDF!&M1id~`PAli1i+}J^8=1T$zYV?>_;w$O-Y9|{z61WtH#XKjEcm_f)(uW!9Pam4_AjX$ZLX{H-Py$R#z|f2Ce_+U z@U?s$r})$0?}Bu#Lr49cFYTp(X5td^Cb@HL@u?;$CA()3+mMHHN&_g}v}Bs+g=^~d z9ZGAWKNuqL=f=N?S_Sr*;)vn+eQ9Ag+Z#@>(rz@EqkEYf1~BYkW1qfLYh>;!p!4-V z?ECu${?hH@D11?Os$E{_tr3-;7B$M5E&kbg4ba+Sid#Eolrp{_E}8Pk`I9G)7V!uC6_4PIelL#u;a|fW ze+%l1bgsINuOzxdPEceC;?wf7p$G0=N86E*E3INGOWlGdG&Q?N+W!CvJ`n4A4epiV z&xt<{?#=%Ihop{s9b!8Ob?G90KzOw)Mw0FfcV>l73GKSR;BSUnw}dq_1N=dT()0$B zD;<6~b*GJFC5F(g1==A$bgb?G;0`fgkA5iltN#E6L-^NmYpr+>#THs6rlO*IdqHh1 z(KHgS_}&?9LfmX+AgjovFg$Q;xbTPU{{Z_}>l#(IweiwBuM`QGrNyt272H~@`(kUKrOY=GA6$e?gQ zz^~p2u)pwc!kt{(YTAn%mxEIKRn^s)j4t(CO!JMm_o*_T*V^0P>8`~^@zf``cWiD|QqxQ3hUD^*sLx9EPYQm+`sR_T4-x!o*1Suj z&1~`gqHQlqH;_QXBt6aUme*>KjC{uC6bsj;O=bSjzBtri@r}-v;dw4RB#^j~b%l=i zPL1}CH(c2ydT~Z1@o*bGdW_fFpRK$9*@5X8Gbo;^PSz1h1<;Aqf zB#ol}$`-`NMei(VqWitXA#r`zfm78-sS z@d{d4HN0L*jsi-u4WMo$@_u8t@c#hW??1HeU}-KJ1*Isho-u`TG8BLZ3_E%X+v>Q9 z+VV^K^FEzpsl{DCxqqGi03-TY@J5mU00!ay%eR8V#M;h-@!Z@^wk=yz(e=*}%8FG% z^4iiFbjacaHd+*nAL1pHll1q9{{Y~wKk!Iz+0#gW0shS23}<~l%1G{^)czbTzLSX; zBq5qfEN2Y4K5S=neB&+7Kaq&^TTMh8tzr}Bxdew|>T&EzD~xlBvf6g1W<=~_w6~P; zAXYJi^fHhK(;~B3JX^fv%;lIlEt~oO0F{ybPk+Qu8rGFPy{Cab5o>o^Oq+$vYg+Vc zsYD1N8WxTV+h%@;5>2;C}3OMFwUO*ezQL!fwTOOcslL7_o4_m{Zs zGu*t*6tfjLW=nXGk&u|;zb5#E7ut+x3{Zxi&M9IT zVYHTD6!3Ykv2-8!CHL*+t=%C%3$A=Wrtetf*F0Bo7MG;J?7?=;J2dma+~l#g+Hr!w z)k`nK(|4ERXYD4|kw!kh{{RIy{{VuC>-Trse~f-M>o?X{`-IwjmReDYByikY2}-Mac&TcA+i9E&=MjGJQ>c^n5@6 z0E2!%Xg?IgZ>f0e;}?VU!!%ir9}nrb{vwRXz?BPjC)1;J=yDV|BIVx$w5BE`t@Mn}W%^Pg&aXXpVRMw@>6jsf)tW_&)%I zDmA(MscH5AN

0R(=}OKeO~WZ8U2!DU)TKeq2smV{wf!4&Vki?A*fuoB>#RPlq| zK4eRa+4W19^;I#pe3@?z;l2tPerU^K@yM^fejEPE-Vyi)bSv3y zV@S5-FpB)f_c9?Ez?hS`f2_w&d1ka%y%cNgowbwnKhD|@hyMT)E%kdnF4w}^<)zlG zv2CH$G~3GyiP?@oiYvHbc@zRZWjGv?KpC&mABKPMNI&=}=f&+N+fmT`AEfE_*CgcY z`qiuq5S|?t#dQ;~4tXle{oLda?bAo`oELgB-gw5!Zw+dZ&TQTtT!eOFyS7`qw2N^Y zDaw>)Q_e=BL--v!1zFj|C z)GW@QF0Y{}mgt~vh2mFk!?Pu@~RedXW}4|w~-jd`GaT-L7qSh7hKuAe=% znp)jTqvn!nA$cWo6|t6G&(wei73hEPi)#;Lbp0nmgHY1sEGK)3$z7!3$!8&BY;ll- zZU!^RtR;=Dcg-T5B;PYWH~1s}00f`?gT4jb+g^M#@QsDU(YgCm_=Y9%BM`*pW4Y7} zmv+Qpl~i_&bupaR($C;eg_hEt3&)=fbj>c(;u!#zJu=4QOqvMtNPCqZKCjc zmPw1~FhK-IXxo!$qsu73jFbw^*Ky{&KgT}-bgzh>AAKj{j5l8r?VfXUdM)RY7^8_C zI>7ceq8FCfa3qZRo?s-;cKhNw+pEwCKbkfwvGlqQn5VA zwR!2DhO(}8pC0&gO1sxSDzp~D#e`z#O`1a;)FD)xqNt^z>Ng{0h6hnmPI8NFI|=Er zouT|8_+g>JZQ>sa_(#LK72JC+XP(~0bla6#IgAlxe$t_oWmZtRJy4#V9ksO9n(fY{ z{wvk>OBi8c_6yvp^MU^WmRIB^R|g-yP)0BtlZ?L;>Hh!;{wdm7c$-$YZCdffO>L@a z^BAv!BZg43+YQc}!6h7#$pm22HT$25_BWR8;hQKfwArT$@!QTI)NLZ%0GOn#zHCd9 zUNA#+U>F=xI@0b>BxHwaJ}A1sS+r01N3XQ&y`E*%9wLHkxC}vxag2wR!XBM5O7G$;OKN=G@ps@4?1kf$OV5q}019*;7Fylfq`Ho| zYopp;S$T@2NM?~s-YTMu0?~jJaC2W1_~-rz&HE1c-tOmJ(rq>057}HymrzHi>LDx^ zCso7{MRbnI8xNT6a7IoD9Y0od?R#AClrO6MLVH_}Ft%O;6vntjY|jmo9qPc}6niQ4CBJoC+Bc*Ej1 zj5G_Gb!`VvgHy33B97n|8jNK~8=eqJj(&BK9k+6M z1~kd!9@TePe-3E+co%3bmRO3L$>J_wD!F~x+NHt5<$+#@k?6hjqfI!ANah%s>qJGxj5qvRlG+SF6Mz+)!YeVNW{&+unHiB0y%op$c zfB<GCr;hj})5(vXbX*`9M+#rpAdWkEb82OJm&R7g) zxT}v5+Uqf2_-ErUg(tJM5H!;*^ftG;TShvwPNc`Tq08)H-D~4-5B||UE!Xtrj^^rp z0V9Zh<4uhxx56k~JWm{Q07ewD&yQhU>-=ldv`rwv;yY`vvs^|Z7ZZgRGBI4?8Xqta zMpw#_gMD#Uu=dkqIo5YQV*6gwZ7eO@;cm4Si*F19D|>SaU0BJ_%b442#{-2i%y0+Y z=~geia3j;2U0yjgI5tN2O{c(SwpVgN4+h& za^U^0&?$sCAM(;EcdMV_EQ%NoGn)DO^7F?w>2W`bHCd;QA{CoXmF{8~rE5pCAuVH|{vleve`*X5R>2gA=8yX%^D%3R8ac)N~A zOR=xC~-^nQ~NQhTLOjyL_Gwx`4St{NRD!2j0}U?yvFPH&HbRf zc|4ZC4CmCe1!UPHM6je&0r%VO5Te~Fa(6JvUQg1#Ub65WwW(^$;cZP~dE_WQ%XKJ? zh;fh=0X%=Ycs2G1!~X!;_u_|ytVQmhJ4F~x$!_xg;5S^SJFvuc$sU5X`n@=OD8||$bpa(-=Hc{{RaHZ65wyB3#33zSw!-P9hEuH_^i4h?Ky{BoZ|x3#d- z6Z>W>R33H2GOOM;0418=O>JX!Dq#@;7yg`Pgsp`TKT-WwZDY8fq-X;}02IVamJ zo4Tt;eqsRt*O6#H8T4-nXqtpx7uW85Wq8wuvxicWD{`@I3h|qbl5ZoDH>$HA;{v@i z!auZ)?y80#5omGg@M+GIn6(HJXkm~XtEAI@?;H+OZzn%dE}xi6YS;XYo3DhPAkw@S zVc{=^DQ%-Yl=8DSw=9~XNf=_w8$Rgfi4>eE8}J4O0Oq6`LU@0{Q+U7PR+(?12=5xw zYWrK4TyYj58clF8(WgHq(4jCygvVAl+zpw^s&Rn+4mg=~IKj zU$fjd?gFnxf?%P!3E1>LG5xyy58;m!O{jQ3NsnC9?ZOZ38#$tqc#z{OC5&KC5YI#N z1JvWT(2Ym%kmnoQ{v74}75gy!9r5;~Y2%ND@JBtQ6K=V?(eI&(34;ukngwGU9D4>m zG1jHJ{g(V~qxha3M&nG;wapsh;UZP@e!-|(EQ|s>Lp-KA)dBgLC01Y#*vS|_HF($l z3Pa(xgICe~KcbB;4PDl2tKB`K(JzdMKvxql&QBYDbZyJQC#86{o%>b8ad&$(9wpQ@ zTXP6ywY`Ek?;ZM>{P%dExIAH9V;;D!rB1H){Y;yK>VA}KUkX2EEmy>?@TbOikT$(! zT&0ppHH^ySf2_B5^2~%BW#eWYiRQZ)JPCW@Ux!k7bKy+_mYUpO+CEz@Hnd8p79MauRvkTg ztDh0SX@82|Exc`O!X7Py!j?%KP=96W2_3zRk|@C+X^joYF5P#G4g!&bn2j2Pa*g%; zPth6TyzsAuC-HB_yC^&#Yoj_DrPMAhzQTe30M(mm)q)jfIL6g-#~X8BC~7)a!ux*^ zMdD9~+V-Vq6~w~k;wUE<_BVNK%l?mcM1vsbbTSS|;=WArhsTc({5HCV_r+SR*tfii zB)q=UV3$zS(T^K0CC`xAIUhPO0~x_J#_1ol&%qr(`$@hLX))Ve>rkpK-ke$dsTV)L zODS9IyRyy3)gGXbD@bAF^)hlvT>S*J*Su3}aXZD~i@j?|X&1`7(V1FE;%2}%HV&5a zsOO(A&Owf~r5}Sn9LwV?pABffB8S5I$Gr3C?q0 zn+M}J?IC_v9ZOKM)ih*DWLu|(TN~vmj4Rlok|_Zh<|FeSi(fVPxAwO1cZ{^@bYF=H zW#S9lUCRZwxob6z{i);kkX*Lg*c+dfmB&$q0=pxGn~0TY-2R3%o2%~#X}4B)kXZP~ zQF)ZuwbkF)VKEK>kO-uIHE>TkQ_+a(dT$2!`&Iaxb9g6IB~^~1 z;z=VrKvpdgCpj;kYwan2XixY{?>P^Wu|K8Z8q{7nVVaIi3ck`+124&caJAB zh}R4ixvqK8o$UVr%xa?9b^f+LLwr%;pB-6i@_6Ur&&7)?v3YbJ=S$VBKeX(`Wq@U~ zxwa6?3+>2I!O#(sdXwURwIA&h@HgS{x$y7AKLct$B=EGB%fDN{)6M3Kxw@;gA)Xz8 zXUvr&0R7fDCcaDXC&XXai{PJ(?7lmEGoIqg!f!F>)3n`hIwjO{#8gPSV>H&#uyedL zYJ+*{$*-S0asL1Xbolu`tEk8D>gPe$bOAF9cr>UkEbb$Kk`_4<^hhSaz-NhqvB?2? zRZ_xKZ6c?Lh4t6)XV`iNhCE3B00*B+{heUb{4=F%5+b| z0-;!dr{z!$e60_|{{Y%2&(|+LHeOpnW2D<}x4Y1!l4rY7yUY>ToVv-hDOHXZcP)-_ zlkmsH&)aLqKNz)*GS}jgPpAI?WTx9#tTkJUJF9%)<~M+>1=k!0jm|z&+iPb{{kU%a z6?k~Ozi;tor5xbdH;Cr9w}~J)7-x&jX&EqZ5kYS3GhU4>B~`LHCt4CZuY|v`$H4yp z+P}rp*?cwet%a_wY(&?ZLM*ZC>aB+@b!T+3OCABs0}$(+^P2kB^Zp6@`$zl*n)Af} z01rG1Y2*DaILZ4xmaev%$pCd&F2r`Sk_QXtARP$rUld#a0Kr)PHR^sa(R_5eH-|>K zrL30L7gA^`a>5D5*NHA(U$h9_Ko6Nh@IG4WuRm-}e(zD2##f&i^$0KSuHu^aN4v70 z#8xhbMKeclC88wOOh)BGa}<1hhZs7}jdrn13b7%))wW`|z+Uz8SHy4lDF?%i1#Y|}qxdoYAZV9X z%CpO>M!#d5dgYPqZdImrTm9hFVwJq`8hy z3`g@>G_B@I3^8A=&7z~R`GRa^NW5>iTwJj4i~LI-QT2U?fV&BbdP}rtC&QY<#QxL}_{t zhV`rcU%;OqB93o0{{Z4&fA)P^#tn^!jbpWFW{iv!`^Ra)Dngw8L-@zzpY0Fv4^_6* z^G^Q6DLw3$lvR*u<~GTp#nq;S~>6~j+!uIiUl&*MZojm5?>aPdcK=Qtp= zuqx-C2JD_WHRw{yB^_HiYU1Ml6n*F7Z;O5}v%N8VIq^O9`mAWqY-NuUXCSB9*omaa zJo#kw+H2;o5Kc9Fi1p7H$qe5+?GCcW{{VZSie@A*Am`=C^sHSE;clmVt1Ye2L|0KB z_E#J8#$+lsfP3x$cRlOquL^t(&@5q##Xl0RV}>h&7wrjg46+GOa&rgTWGFfr7^vq2 z*Hj!Kc5&3Sta(+hi|n4wq0ww4mK6mGzF~+1kU;qXZg@Bt$6AgpH^CRylUv2%&$G^q z4lLA&e6ztNB@NDT%8`@L74}z$yfOPYcz@3?;>MEt<}#@hXJ+!*O|vvi4*YFm6yO|y zafXk$fKYV%z%_7G#?M7s~7vxX? z{yl_|fVse~*FyOH@v2DfWB7mYwRG!ktU^aH7Z&WNnAJ{M9e}uHdcRsD7 z**d9jvEBoB9!7ZGAC)TSxFNvw;=JeL7l;1UAU79PIZV*OpZ0z36*b-08ex@a2G0ya3>TAJU$JV;W zvki`l&u6NRv)==#lGzv_ugs+OA~HWJ(ll!&cJwi+S`zqHr_&nm$Ik`cUeBTUbHi<9 zLY78^%#(si1w5HzYz{M*+P}EB?Mr^&w0*72(F?fm?e>#9T+0-;gnx9TaT$Qe1w>$uIqzIO(|j_y3#HE2Hdl7V zR{69@!`l#kY%1O5lmLI_nw`9X&raGl1O7w7JcP@OE5sCxZ88JLTJ~LM!Dmmo@@;j?iC{nkWl?RI0-=F7 z3+66Ha!RP@j006Q3vEkMwT9os+DuxSLAhYQn%G^tF~grQ@^WwSb% zjXXOYrm1fXzaWX0@zO^mDuXE}C!r`v?~pT&cTn)=v#IJbjZoZa(<>0`aVlIXJ4gUq z?tP$tvdQ?@v};!SmEDZ3q!R_dm-8g*llg`=;FTFMDCeB7(AU;~v+wLh;-86{N_fBH zHmi5w&jrK-L#d{kCbhUs5(-=fm_Uk73reaRw@g$^5gKajaG@zXBlBa!{{RcTckv^{ zw%!TQpr2Bj?G!W`ZK5y=hlh|%(#f@(8;bBfMSnqmu=ng6`zZV^HXc2{zPQ#rd0`|z zWyYgt7_!o*!DosK9ZElw0OQSKBmn%5fY;EU2Y=w1nof`Kz6kytc>DW*N!3H%CaHTR zsC%S#!i18^7AKA{Jh=Y=a9F6rF>JBqbwBteO^1l|>%aI&wf%h|hG)y(+#7|$ld?-` znpS0qlk%eZdq*BvuQL;k=ekbF^r+B`cRyRa8?1O+R@831H{%iEt!BxsUA+A%-_MnS z0!Zetm`NBoJb8B#?mPq;&0qL+C+yLpc#(9!h7J9Le6ojqC|u8_iC2-ew9BQ92q&({ zf$zG%7(6-u00g-B)1~TGx4r=QlXVV|LT!TbSgz3a-5a1s10CI;1Fe0H;Lm~o02}-( z;Zvi0TGO>35!>mm3^VBs2CWn=3vQ0!K`c>9vJIev63D>vtU8|-aCgz%?wViBevEjV z#(x%kD${BjWuBXTXCZx%L3RC|dJ(x)3v};qGoSByu>S-{4J?YLcJkyrpi#h=O>J{NWfgO{*Z%-(Pl6g}ha}Mc8EUWMeP&iIYk8`O?wO-;kfH2j zbzq$BWruO@#EjR(e-WYhKk<*m$4dR6JPsz8R*miMHK?zJ-LwrF?m=>}Y`=@A?$Pb= zSbz$#$^3r!^W)!+XT^HHjo{A$>AnR`b|5at*}l#viV)6>v)ep&KPN5$LxM@##DIS| zf8i?ewVtT!b*sJ1ai|g-hWkVTOMe7W>~KBGei^S?hFs;X{{T~(ScS4j?dOYrE&jrP z19eHfDe!~h4yaz7*ANukJKo zh2;2h+i<&0K~@JF*bTBZ!pj~*C|%!a49)y2(!NO4e`~+l#(fSi4(tB_53D>|01oF; zvP-u@pkR-*Ixryo%F)Cz>&0^#&+SX_Csv0|(=@#*T`u27REi5W3wS~XEU~4X+%qYR z5^=Pl{{U}*2I~hqYMCyqU60g%g1;ZXWbcT2Wv`31S@k~#TWMurj_oh*Y*q!30(^;J zm5~uoDpFKj_C3E{J`;Y>dXIqgyX{xN`s@5bw0O+geEPJLBf$A8hFheLSf?dV1?9*c z8z#RZe`O!rN8qQ0?QQ1xmoAk)i*;+fM{T0dVR(~ae7IzV*6P*b!EMhO4Z8p?2_vo5 z{{Y~qKd?`Lyk%{9;ZKBDz7o`I?iPRTDD5VeMBB zB+jZ_ty%v7nfk@A{@H&R;nD1Mzly#By=xd`SdO8j&ODfeM7|#cvjJyr|Uq$;uYfa*B zhW;jk+TKDGuB~J*tzc8hM3T_0+-lqn%w+4FV!ofWPXTzpz?-~(;QL)uRJf7SShQ!B z?6U8~G5}+OLmu2SMnW(rZsr3u&Z8*#Sn5e>Z4apQ9U9u!%TBiVOK+!L-)z}5x`F|7 zb07c-4A${W99bNlv1|tbU{-W4U3`F=^aMjLQJ*l@$9+>x7`y_n>Z(&g_9 zT>L!MCbzQihmWtkRTZ25{{Tuimd+TR%DleY=#G6^w{7X%TDb9rloPxW*=VY$#X_;T zb^sh;w>yui@79yzFT_uTo*&dC@V~{chxh4aBdXf#THc=}#gn!-f2>7m;*m3*z$oCA zd;)9bxxe7Azpy`tC2tQ8+7IFVxUeSQ$`izPd0!^qnnHeDW0Fl{8@*yegL@G5AUU z00!ylI;V#rzxa>w!shPcakivE>68RJ`P$SpMw z5-i$gx2Rfq?Wf#Ku-%46!h0yyw~U?H@+HE*BTq zP|vE|w2^dMi-&h-Wh3P+CeT1VMSahAuXvi@#%SF# zz~=<*lPY-m^Tm3WkE81s3*^rho8;^tq)G-?)19!NPE=b^4DnWcHFa~g7T zyWY>p^Sv8LwD6Uy=w1QTFI^*!G_=s~B}+&^URpS=<3e`r`?1J(ard!W7Fy-7)ZbZ` zO`l5CuAS|kE7XoRV1ON}42-jkHw_te+s-q^Wq8NpZinz+P13CXH@(DS;e4jnEiCRL z)80i23r@}+3qs`KQb`K+8~SIf)OD>F#2Pu%-qO=rv2{rxhH|qM;1u$m1|wnA6yw{y za>ci*cP3G`==1%1!`E6py^fvX&2ldbTFdNqBBp22xIP zL00WfoG+vTZ7t7o)P5%X93CT!#r`6fPmcE9CV*U|h;=yCmjwio?^$-|q>SJZ$2hMu zZ;ZN3T7B-Rp=y_&IEpY9DD5rdv(%Ld#@F*6IHXk>$&Uo!5QNvxP_MxqGgQ2T!B#f< zE~7kB3s{3Q=?`(8!Q@B=>v&uQK`w(!-uTi@uGejHeF`|mv#(9CjK zqjp%z3IW=-K=v}sS3lHXB+YorD;QGtqY zDN;Z|=Zq1>a~jvi4-qbzXQO;H@dcNM(bh2RY@!z-=a-b=68 z*~Z^&YfmX|lpL8G?9I1v!TGWC9(IG$w~ie)x)}R!GwpvE{@8yA{t#*};jbENlG|{R z+qCm}cCzp{7;fchBw?PQjmIA#Dr#(RBI5ccUX1Sv^I^B!}Zu0nbQJXdsZQkID~=4UKKW3l?p z;qTh-;*E}*Bm6{dIw&E^NqeI-t8a7(+6;D4$29RV`^=_C*faB-W~yo*wvT|U^!T-n ze^sAR`wJGfm+d0c$vOFud5+#=>Ns-|ZKorr%aAxMafIJ`ULFpJubPw7Y0<)6aQ3BS|S@GY!53Wd?JRF~H{oyRH8K;H?&3 zOf!9_MTWvRD3Q!TDe6(-SCTdZI4VFrEBT|-&&3ZFPc@&1E@8NhS1zfr&IlhcWh$Ux zkfm>VBMP zf45$p;>&w&BjOIZZ{VwjxqF3O4PsZ8cMN5D?5$OnU{Bt|4t)#p1JF5$s2bS1jouIR+Do<7! z1Fb~z+CTE`*p-+|?P`BOJ{tYJbblMAr23wZYdwvWiFIc^;M?U)pC~INteYfIPdiB* zf)7KvYx`UHF3VGn>rnWQY%K2NZ`^AFCzdwJ%G+MsEH?6Fl|@-p9+_^r{(CO|AnE#S z$4~f$sl($LtwKnKNZQWo-+*!FN4iE_{{VV0TeP$gx*}@683udnN z#1Kx_&Lm8pjL6H57O!lZ^fmtwfCz{eu zq~zn|DmlUR`FG+U#vKn-v!6%PE^XFS%Evsmtu(Ny0lryfloG&l2FJYnL08O!!vTSd8^8IB~n^3v8|s&E_39YGNG`ke>SJ>{S*W%W6xS_Pw?=LR`NDFF=A&js;C}fPs4&$Gf%1i(V zDl4PYekfb`c2fqqX$8g7sUymP5-h0slt{-TJm7)}Jvpy4KL>R^KJHyg8D+bKKbz!* zDKK0O%-ix@^c^}?`+peR=_s0Lu%A(8EE?3WpR)&O+`CYwa4=M_9r0K7Xv;{m2~DkT zeL^{Y-MbeBB)G)ttFst8wzC2_$zpjuab4ZV?M?Arqm1a#$!`n(^Drz* zyN@G!vcBWTLJuFO#dh8;u+-A(0So98g)JO6Dvu8F@__NL-h zq+F{gBoaBUCf_e(+Z0)AJxKpI6MI-LqjKU(h;37v7u`QlX4;{`1Fej6TDx{Id3Ek;~-`udj z_jbV9Tg`H;g||$^NxL3|EPXo`SOHroko)x#4NCsiM zxAT=Z$=!ge>ZK?Ct^&}JR zLc-H-2Kn0J;3Y-JfueL$0ULG5fD)Q`l_37~^iY|~}>S9p9e2pVQ zNnRL;&r{bJJf6b2JDcqW;G0WJi0&>eo690eGawxb5=%JbcLWZe)#Lso*R;JF$L*Ha zQ%JkEh-8??RhOJ>FVp>?GCJ3qTl``1eVxpal!Iz7b0%Z523lHmwEfSyox+DIgvarjmJOXB^G zlNwvJGRw1-*cBo!PH;1pDaqOnGwDLZ>#8Z$Z)S7XUM|)3+i<#-vaAaK05!xbBr4eX zNp+2YeCKdd3IcP3+ng}zrq!YFye?!c305f_<#Uc^QW1x!`8dZMVyXC-;s&*-NMSlO zKRFHNTSFw$BWIB4NkVbl5spSHht{urOZ}X(8KHQG?^Dk$9DslU`4%!sAZLOA>&13M z4wBHB)xPG-YThXqk=kok7MAwb{Dp|-Y>t390yg!?&(^%pTEFmiuc#d!<5ar5x|&1& zhjlnHlEnV$5dc;T#y2@{YQVkmJ-zg1Eq?6W+Lp|g?4e_J9N+~c`)=bOjd7EBlG{s? zX*7E)dE!x$u&ae#-L}FoF`m1L&o#7YIGH)N^gWV4i+>dS1qsuA7 z=&;oF3w=H{V5K94?)HK+ktC%+9dZzsW!~`|o^c8m8DZ46 z>{mo-CiNrQ$sKlq<6Ex@>Sx6Ie6wB)c5x%y&Scy_yDG+ZFl_zSTywy$f42CoapJuy z*F_p-_K6*f36@Pd>2&**c|P;BQL72g4@r;!JC3#UhLx)LyHB!f{X0Xqg56&^nPJ;3Q*W{G2=c)|YwwlD59=`UL==V?AHS!C0-8#bNBI!PR|DgD=6l{qco0nr!dVJyT7Wow<$}g3SRRbZsnyt_iH~8-CI{--aEe z@YL<8NhC^I=E;&cK|`I4tXglB<8r)cq;?}UHj|@%&)yK!pr6C~ma7$mqnYe2=YrzV zjH7N6E41?z#lcpHs`n^s=Ud$#(&xoD9wL)Mit(W=(OubHz1u||aGqo?ury~MeS~{S z&$e=nEAE1I7Uz&^{{R*(b=^wx1=BQ}Yt2ZThFu=PqPb^61Io+0%MMEo@3j((mSgs0y*|jYAxL?=Y)u8RN0zJ@Hp&0VB*XTPy1@>DXig& z&PZVjHbUihuTR~L>WOVDu;A0im zLEz6F+grzBu11Lqsc$akXPf0_IXh!w0dA!31&33`ex+Id$i6bt?V#4J<p&Bv;s<0l(mz z-xO}VOKXG(yO*I6dr!*Ae#Mx_$T2OKMf9@@HOl- zIdutzOCzug^c>TiYGXUBU#5M1bg749q{w1U;Ht@T|y zP}$}(ADK<0tsC2wQ@E0iz`z;iysG(?NjohMLKGmQeb37nKj4vZ__oSTH$c{7)NgIw za!5$yU$@VIsX7=KZ!aW#ovSFyMs|kg|xC?+RX*syt1H{GWvbQNENm)b;^*+LCLP$ z;TP;_wmuV@!7bYu=xK(BwN`ssnV!s&Pn8+gsbph*tv(F7mt(M+cwq{{ZbLu6Sn7 z-q%9c?>swuB&`L+7Rx@brj6Oj`PChMi__GKudCPS{@{3s>upclpNV=u>@9KP8N46y zF2+40Qnq%1FQbm{O)WBw#KC{%#{|T5A1MbM5-ZAnC4a#}z6Sgy(zJg9{5kPewyk%j zFWD~NQ8zboOd%g3d2CWOSsSanhs$&r$*;-F{{V{r03S7vJ5sWNFWV5**bW>?E6CAO&c%3Ea=lj?M_8G5+ejxtTFXO2S zc;#iW+UI;W1*LaZ8U5nVw`&7~h3nTC;=Hdx@OF;AE1vtq-Y2vWPi&J)vPj8&y#3gj zFvW_2*LTaFdgiWM{7JU(z0tSvUxwSnSCHJRTFI$N8W^^H!iFS~q}|kRaB;!z2I_<( zvcKTSb(`GwZwx+;WEzKsZmri&yM)N{UtY`hr-@I_W3iA$9E=YbUCd9*&1UKsHoA;) zTIo72r>AKa*^RXum$uOS(s?eBO8o%oxkDdp*OhpK#u}H!?N9qtOVnl4?wU1;;GQ_8 zhvg*gis~_xIVS*>9R@vo*8c#+-YvDbmqVFg)h%t@pgO$j(G97{h`#xkla1-g$I1vf z7aCG)Q#~2!cAgjTjrWY^*R-D;#FE;gEN0jkCwVjXe$gwXvD^kZBph-(S4Xbu`t;FS zTX^5X66&$WU7^%9Vz*Pt8UEz9N`aM!I5`A%ZY#j1eIc#i{T|{ANl*zRg^bV=yyclf zTr#I5e7uec_2~R9&{}GjKiSgVG?I}U{i1a$*bO9RKP>M9AS{RF7(AQ~l4-RK zPAh38*Pixm=gdS;?DW7q;1icVgtr5yYJ2WuYZPugUEp+s`$e?-8#|LA3FnzE3E9RL z+1M(R>J;=mdi8oUc&zEkt@v`*OVKEml`Z8<0!!ejFwM6q!N%0bZPm+bJ}c1lYu0ws z%U!Yt1y&gA=tmAC4=RbM*6bnk~7&mLSBCOA*C?IsPN@e~!FKANIxF@$n7eC7SC}o?$FN94UBbFCw=) zK*!S+#oBnJ-(G!^>S#^CEW7QPTejyIBpEj;<0GIstfQB!B#q~Zi&nAw1Mz?U3k&gv z$5k)jUkYhDKiXe5nvSNUQIX=^^1N`vsL9264e$IF>-LoKex!8I5na!$ zA;<1?t621?Rfrk=aac~(|Yq^r}h2tB*b{IYQ z1JruIhPxB(+B?#5?GJNM8;= zYR`jH=$AUj#B2L2Cy#>+8r(XVakP^o#L9^h^Npd2-;vN)k$i9X`S6>=HaEIQfc$6S zzX8pAZjOJn^-zWsjU0&~g5prIs{#q~*+Y;zkSp**Sn$t|b)N*a-}dyN-FlWvV7_^)df-}##rTv_ouW`>1 zebcD?8u*Xn{Vn`sul!@uyiubj*?@;xO*3k#KRH%kn`pry6)|OEz-X*+v#mG=I zx+PyQw*+BIk4)B0rKg8AyP0BHr;_1Vln&<7CQ`!~+qj=Vr)dMGYt^TRYj^h?RVR0# z-7fRNFlwK^w3I>)_~V{n>&I0j7|vL6&Tvn9`VYf@6?`?SY0~Kl@aD=}epsW+=KBU^ zWL>*^5L?!~GI+dmZE+R4>DKo0GZ@=#zHp=NpEx$=EIJ&Fck5EONEK8eP=|#|Q zBPbf;BGPaKthfrf{v!DwP;*so^&%}Nsq7va*X%BJR2~t$PYtrLCr*ayNpiyn9#zlC zs6hF=wd>C8*JKYTeS?v8tcIIX>EPk~+>i8b$rUMj!SHFy*Ea9dk0 zq*y;bUBolM-d0E%`PG2vK(2eOj>t++W23S7;|GNF$ZvnM9vKxxaIL`d@ zWIb|5))?oJUvYdp{{Vu4e$(C*w6xc}DQTo?7Itv?QbBuhKC^CR1$?=_(fg^FXa#n( zvJac@17Dx^pAfz|_>)Cw?ffTV@*Pwm-*hOUbwK)~(^@aL+Zqi*))ewAncG zUe*~|%#1i<3~G9WQ=I&tB)$juTX^yKD^t>SsTh0H-} zb**#71sD4o!X^v7Ww}0lyUD@X7TDVZIXQ5BD2@^fp)r#qFIr+JvVosy$ntop`TEPuRcqst5#s3+Hvm3~jB{k?K{hgADo*2<|IY zMLmIHAKG?5|I+@pJWZr{2gafZ{vcVU#EMkGaR%sSv~@U1A$5N-Kik0@v4fGtb1-fFZt$*?!a*?0o>g-3umy_`yI%}=U-rH5ABHtOR_nk% zJ8u=}$qe@cM7xD->d)(G)&`d*itY7b$!09IsC)X ze`rf>cff{qvi>BnD?eldJK{i2}pCXM4y*{j59Z>L2yy}i$b z{6QExNwI`1avd&vWJzsjh!E(L8DnBLouGzaTe#Eo{{S4`r~DQr&{=ABlCo+L zxI2t;$)7H0{Z>Fh8R~f*9p<;GXj+Cf*e-9hDZHp6ztR>N&zRUvw~#D;Y#^Vy;hHna zW?FXsVM2EJb@e_R)IKoyTgJZ^@4gIcH@BV~@kA}=+ga*x8!L&{HaU{m1DUczV5V6> z%K!)_$B+IUCy3*UA0O*iGF)2*5qWA1iIIXw$mcjcNey3nc+0|{2mDYikBqDzP&U&U zZKsk@w-)CcTgsaum5=cu1xUy_ug|S_;YWx*0sXSmz{{!Y+6CRc_1sX$cGLMcK4+SW z;(6{ap^s^0A7)8o2N;eez^=t5Z+Pc&wy@Co&s2)X!H{AZE;ZYGDDxci$kE(q0}JK3 z#^Oib8vx^uroI>Wk^6XD{5{nhMz+$gTV_&ADN9Qt=jLyb)wbs#ZXYXtwf0wwJ|*}! zR<@e+QPFN;zk@9ynZ)+)+zs~gtGYPJ!Da}{(Soodz6{rVOQ3k)#P^yUIz@{_fd2a4 z&LU%>JgM2?`kERf23W+DnEE?A<^cFmOogTn?G=(^1l6d%XoFYerUY zJ2I7J2kw!Lo4amKeNSrlu=A${?jy>n&fVGZ&6c;X>vr(?b5^-q^&spsNQpQkWI2%V zPCGFi^fjbD53ao#o55)_T@aIfmi2PXySAAN%Z0#iT=8F9Tj-i$hABK1WNfZofEI{? zSXxqXkmwPEHyJEKj^tN2Y2ZJI3$7-gScD2Ngyy1R#K&Ls#ZU=6)*&2y! z9SrsAc?XBI$-ESr_ciE4!Cwz`?ME7uC578< z*s;mw$1Ar#G=QC+oMWChafJy%WjCcz$JFY`nMhHDb7jVAT#vOhdSqiwHB7kNRG}$0inDO z-eRF7i5B76j!$!reKB69HEj-m4YGVlp3he;^CQ`bMc6IIWN?thpHgsob+13uymP5) zx5I1NC5-8pf=MQtG8g!RDBPv7{4y2C09U7IR(}cg+xv|_M4nrlxmY{tv*O$^i~xSm zIaVwQAQdd9j+q&*c{jX(<`?1CgQaK|z98_d6Y9wvs`nPsT%lhpZ`~plO{H^_mMlLS z(D0wd{{Rw22iW{SqMOYe0+Q+RLiZNU-!e48ut^7Y7<9oMbJbdF@8IoDMACdSCYdQN ziy+?MKLv{!SR|PM4CG_3PkSq^*!bEBE_`ERG}Z$#UTLxy-dGg@?3;980!HZ~l@D>~ zQg>FKp{GNoxV^vlxUuQcUyljQ>Zj~C*9&cC+lWviYhZrwP_iQA6N=$Aj}LgK;V!)J z!#T3pz)sdhQZ@`Wk1)yRB!J`P%ws3-ZK`^&fp6^Z-&zp0+$jS^aV?BqNLC}}-mYbt zTPnXZvJ3;tgNp7vEqANwx>Q#1=tlcZn{)k>Rso?5SO#QiN)60LSITD@=dCq$@allT zNjw|l>$E!r&E_*L&Ad$_Z)^u?B350*?ZaSjKIpE~{uS4VbUS9$r;ktj9B;Zi$VoJw|0@eIp3PP~>uKp|Mj1eS> zCzb-ASp#k=J7H4LGpxpD@39c?wg>R6+ z9Z^>}JDIuZo(^4?+RzV9H~ts!1Q6Qm7V_y=(-I|Dc^S65e4&`hi#w)ne zJa_O)*8Uso`R;7pL63cy^I$wN#gMo_LurW zkVdgY5Xhe|%-(9g9{>3^mo?(*a*36Uzjf7upVYdXdGrC0@;0!QV zW&?wo=Ct36T2G4(ZZwOV$5r0<7V8{x$ellVA2>%GK(;KC4wuB%M7@n)}Ya(pt-+yvU~4Z`_y#;QJ6f17yPyzuOYsjZDCUA$01kM6L~ zAPS}7{W=R4ZDYeQuBEUOIhI2cMUq(m0Hde{GIBuNPrrKUHLnNwXT)-?x{FIo-Twe8 z-0wE$9Ds~N6<~T78+!~_CF1)}7*Od6j$6pV$jhc^yE$)0%yaMx?*_$Kpd66A}DB}SG9svfbI2A2+WWx~8 zyL?6QU&lH;^4RN_cS&oxM%_TI%y$6e43G>*#yBCA_swt@o-ei5G-A4xQ|W-SH`<}Z zaJO7~*b`=y6^`BO~&p?phyZmpnct7B~z*u)YxARvq#zG(nA`$_n7lT{`0&bh03 z%Kk-#%P*G9b!S83~xNEz-pJrKcZsS~Zt&#EqD*7TX1QPXtq-a?R%AmWjY^e`1V$L~GH*NN;Jv+qG>l)))&u6k* zqA;?^Ws}e%k+nGFoSb@Bx9A@N{7)5)&bz7Fjbcc4`Dv%xxw#7caR&oF;re9azLEG> z@KfOjk8H~Lq)0B|$Dd~!2yU(2PB2Qk%^A)KB#&&@HA=9Kh}~i2)z2L8Uw~fMN}pS` z)l7F7LM`WG1bdD_4J2zMiaFqg<2>EQ z01Rtax;4~+Pzzg#?jw|S#Fnyou?+XxLB~8A`8VQ^jM8SeVv=0Ik%0FadPfmi2mmtXLa*xC8a%vVl@gdw}ecgXqqyRI?pIIlbKuYt6^7XCS> z(=Gf&75-_h?%CyyLBT#$gaNemDo@hA96cK=z0N9@WH;L1g)eQ)cCfCXk@7>p?>Im1 zw<!GII>4q1D zu*z>tZ69DqluMF-x<*Iy(Syd8ng)Xg0NWjFdSwZR7YkO9`_kyk9NSsJrMieTb-Ou;d1ptx#@-y6l zR^j-+Z>GsK(?f8Rr~*N91fwU84q7+kuw44$x$Ey8c&k~x`&E#;(=DVZWoaZ4#ULkc z1+gWE9Wk^M*P7_H-lz`a!q(QhW|D5LrMx$2?yUi|`!GNbG9+g!Ajct-c5}#Hm6xk) zUKSUAefEQAE9}cjv65T?(V7zwZls)s9=WeIyiIP&g~7j+!y|5yt!>gdQzty5GvJlu z*#au-lZ$ z{{WU_amG37K{Zu-Ez~JV*++Aregz1ke*!pAOSL{3R=F7l$^88%zY~K&<{5dmQ z-Az8&VU-H2{h|xDWVbfb6Y>^2ut?gcf$y3Z{1XTEYi)MN#@-6?9qyfGvqLxWbTHhpw8`aSzOB-1XfGA|Jo#bR<$0Tn~e&Xcfyo!;nDP24I z9hBt+uDAQo)~#Q_C*a?R?w3vQME)Z1eVXh|N5eNUF}GGxlP;SI!{x`<r2vskDMyR16#``3%yZE;b||nUQ%I1J}hoed9Oyd8W%J ziu5a+J%cjHV_T?+CkwZ7l{<;raG^oRt#3^x(tnAJ)|($SFYQn98^>N7wD@%T9p;T} zk`nWHdRc51H9cNP(UdUc;RBWGO4Z5qABq-}Z~K=3a+FG2}#}YcU6s z7#^H=u0ve-<>B8D-9;yltYxuFnAQ}O%9h}wj4so6)9orp1uO*+yKMuTeuEU|}$NM7wMcuN}2DpP5;iR}9ypJc6 zOM&Z-m5HtVNbx6xEmOi?IML*~vF0#!Gxn1qZm@|Fyv7*dU{C zxrN4dlF-IvZrH?%8-hp6f_F7<5BPh;+EuhZEE9dICl5XTl2#ux`@o}?`AFpYbpThJ zXf8ip(m&4Zpu=I|--y%LUHFpj2ro3i z(XGI?aF-G}+RoxR7A=d6?c@@A05Mv--uE<&dbCBOct=9D)b1wuM{x~n7*MJ zHg<_OF*s}-908HQ;;^i~A$XTlj9Kg2GHbSlu^d`~TV_x`@e8q}k%wX-&PONHvHt*s zR!4*v&kM5?ly$em0)L1%9`)JScymtiCERxYEYhsB=pPnHHX#?Uhao2Nw@4g}B-? z3T+=wYFc};{=Q^7dg#u5ZoV1Z8E?MFcV^7wM3*v)yEO=>cJxCGq-T+raHEcTRxZ8Z z9~OA2+-C-v+dWC9vtwk)Dqra zTM1Ppm4uSKmuvF^K}b%`gpzk{Svq6hyF0yCS@8Y*E8_J2-_zpr72~^&+A!GZ4b-E| z_Y4_I;9z4FjjZ_J!I~?@2nR&JEF|BlV*6trO5$Q2LuV?=gn@&OYE)$OPeG=os|ybi zYgQ0Nq4;RNr5BdpEv?Ed5hz^Zb&#)=pmYpB-5#}VHOYKK1;>Y{w2Mr;8C~{U?q$X} zkH>`DMADI6#i_S$wvJJO%~UbVpUj82~~GoTxqW{6XUX z01WuL+IcnNn!Uq?WA->6IZyx>i;0X!JPZ{G;LM!juoM2#%I&OT+oMV5CmjBaV&^PYc$I47lJU3kaFa+$Q>?6G)|BAFWQ@e5^2H`a6v-69X{?ruYk>&Yp6`MHfa5*QXa8me+ z;>FBfG?P)(VRR72*HI^vv3mTr$}x=KGXi4On9m*jntlB8Z&lMs@}r0k zxwhkbWam4AjC2CIi?0nypm?4e+sjLhPIfqoNp%O7akzZDVUdhcF=Rgo0hB7wmMIYN?i+#t;>^e)Y-vog<0xq@ky z39)G|@R6zsQ0zIuNrP?AP8UA?>xS2L?++%X_Zo(iX>BAcv?&gxo(AQM0Wv!y4ZK#K zsixfCUYESTXrTk=h6ut)PdNrw$U)HKXeZQHIjQQh!=S*Hu!$hDkh{xvjeOo;Oa_0I zNbGlGHQ5N?SE!{=OOW5(>mDPuYs8vj%1DFlC@teqqjuC~#Cx&)InFwolGYtgSRJnB z4G`P=mI*jR)R8Kb+6eV6j=ihfbT5Keo;~n1emL=Fr(xmU6U;3O-@XeS+fL}W{_mhU zEHjP+3|Akkd@}fX;k1tL!oDuF)vj1SpZ%U$P#(#NRxm>h9Q@?<7!{JQ4Ps*$H?i*D z4SpX>@sq=Lrfpeu2e&V7Y;_m8Fz|7JV+J9B0U3$Ab6=)@75@N&ciZ^y!^++(hCMpf zqDdi^D|dMDj0E%ARO$%8*%{7zSLOcz?AhYIF3ZKsq0i;%t#r|#xZWYSExXHS1&N7@ z2FOyUDi2;O^;bstm+`m3lYeITKgAkXimc#>x>{<+ODT_bQGITbKrlG~+YfVICN8AY zv`SgAI(u7Rh&Hz{HPk_c z-<3OCsVDB6h)^Pt+mK1>0L6T<;=hPGUCh&XyTN`W@a5bvLXs14JcXBWARxF6u~@b_ zU^051;;2@e(Thu;HYbGs5W_q(cr!-v3>uc5wpWTM-x({YcpE-eIm(*EiU2RkP-(RHMo%3@e5G~! zO?b}w=1bene@-(@#oU`xa?EmhVoB$@B>I}i)U|=8NV=8GMt6{TjV0Z~LKy!5c~&*p z0VAB86V!oL6XR{cKzx`JkGteR*!It#&OK^nLJ8`KjaShB)BMA;@Sns>!!y8|m)fml z3?h{zEAoySKP5oO>{pI4n)dr!{{W6K7_O!8#;Y9iBDvO9bEioO0AZRbWoYIYA9a)# z1J=Ij_@nzq{5jX5m&3C7ay>@IS9v1HmOn2C1f;taquNG8DJMNE{FQ#f&=HdkhFC;~o zz+u50C)bmXwe0$T{1lJk+StpXc$&u3#TsSG2<5fZu_Qx~L5WnUiGjfkqi`c5-{jdf z-wA!a@2_vJubr6AVT#^q_uzGJnPcyQI3V?^4X!Qrw*C8fK3&4Y<#I049v%3s^#;m)D1N2z#z*HY8(W_g}=(_;qC;fN&3a{p0usl|ES-|f-y z2U;sN&YIDyOx{y{t(iw)!wC`EM?b=SM|%A0wu?;GH2J6Sm6~1MuMzEOEre>M?k^m; z!u7|>Po-~KYf{{#QQNJZz0@F?fiqkzs?CA6H}d0C#yBi9>6)luDaS~m&l=w6(ifi@ zd>^j(g5u-EF*n*J!(7_IZq03TAUqqWBvAy6N{^o+PzFPHti1;7;1;u~P2x`s>G0_u zA&+yzrOPG67VX>2jhiaLfzBf@jNpO69yzbKhi~m5(dE<^PP-Wm2inEAnArJ-LbJv? z5EP7kDQx^P;UTC*pV*#l@gL~-6UQyWAp;<_)V^YZaq^y9rUnLUp<$!bMr5%L+Md~O z@&5o;@FbsO@c#gWEo9Zpt4wU-m?&PLGmb$dbyfQD*6yeA{{TwyZK`SBH`A9wXeKQk z#ht*GEbClIO~-^m3I+eYCb1xn{8ALBT!gM zH1WtxML{I1E-;>l2aMLrlu?wMdW^Kwq=}jI>rtoN#Bd|K#3h4~oT`F*^QfpY;gKXAOMq#)F zXDM`Dfx!L~^KsA|3hA^j0eml-JADsZmh)Y=-Im<)KbAUu^jAM!*Z}ZA0=I=ky4;?J z%X)8)z7}6ze`omWL#I!ZvUtln7g7K+Ll(k8>T!YXUeRy%p9|V0^p<*jnr_cD-lIit zYa~sbubg?YzDL|D+@7bceY5bp;LpQ78(WSq25UMtqZme%*xbBOY7nyEjp+(H&c7}; zsw-!xH9LHE_+@8*bEVtKQ>}0nbh=il!xB-HFhtcRXjo z`YxxdX%?{ft8a)j%jDd!J||e>CxDaW6f687sBSVq&TF!VP4J$8k?DU7B3b2mU+o%p zwRz=BBIkv)crc}x_(a$w<0N9W4yW+@#F1SugJ9D1+xxK`lFtg6;2Ghbdsvwv4nV<~ zLHR}qIN}*T1^Idj*EM(3ZM5rn!%cr{F!DE&HnW9lJ@-^L9&9l}Y z!B0M^Hk#T^)Q-kWi^*b&Spx7DKQJki7#p@30}Qu(Rt~N46T`kXMz+`VsO~ixG8*#0 zh+L3xrIE~NTQT)UEsx@ZKF77_YCb^*bLYa>C5hM(fl(S1%@1o_R!2M*wlp z33cF~g}VKc+V}>0STzY5m2I!Co>!R&0ahtQh!wNIGDh5ba%lFE>Wo!4(VkIt;KjM$ zsQ5DI_=?3teWUl@b0eMrn%)M6NV&?8<$>=N1n(R+aRm}NyvnpyD*Mz)(2mDQL;?gw+ zkom4d%WAS7n3d5E<^Yu?LaF0ExEhV5ejh7;YjZBGHNTV=TWJzPrywbqq>5E~45%Rg z0C-n_s=?q(Z9LrQ)4z!<+lHN8r9h=QS4o2<4Gsy>xaQe*ic!x~2n#)VL z(PUiij6yD>Wh8^Q?BxSzKOiivob|2;;r{@KwQnDIg*-9g4SP}2wFG&wZ>lW%OShIX z=H4k0Jix#cjqQcYFyfqGlHSAQd!x`SWYe@clnHhut`$|~dxwy*`FA16=lzA+qP^Qn z_+4}2wp;AQ8Iv0t{e&qeP0JIIp+(yMGB7{F ztVRbICl!rWw@lJB^ttOAR+aGM;a7=t{{Rbk{`MUgRJk5?qsa3xW@C_JyJi;q^Ty9P zn`jtN0LRmFcuV2VuX`tmHMrJ$&86DmC3F)zjv7@!sUc!A3Be>}Fza6lXcyiax6!1B zMe&t}yQ5qjU1_tzl39izos!zj?IJKE{+YIV=DBT8$Coo$X`USTZ{jF?Ef$>%Lmi4( z7O6h2cRpvLpXpPG8h~|n1$m4Ktm9fD9gO8j1ajg75@txai zo-KP#M)%~DT-z44yI9Eqb8xIfZ{s2TdBt|np!a4qWS#Ay;NCm%Cx@(z@%Up(9D z^ZBrx9BuO(5&{oUS+S04uYvp_;^<5Q?gJA>17H=Dia| zpFmF%NvZrk(scg-63T$)+TsQpoD0+yxRPmOp9eVG<@yCV=j%TV{?I=aXT2W}E$pme zkYF{8tsj}PKsl1sETpL)FDa9ZWY?yrD61olo>SrP*ze-s#XAPqwRO|9_*f{4@>_!? z!BS2Fi(u&40GvF8e=6F${fvBj;f+QM{{Zb3@lBj8tu>{k=+o{&2c&UaDT&DGwWVFd zC3_0@czz#4;d|%SJUge`_*(6ijMnxSQpcw|za`$x??iBUCOwBd70)BeptSMIYv$j_ zHuGvW*K=);Qql;$flKg;@xJSrd-BLv{jpr~ttas4UW-I~@59NyAH$=F{G01PEv16u z4N*SEEu~Tsqw?h28|^tJKtiy?DI=Qo$UJrXId~rCE6pETlT5ldJEi+2yWi>Tcvlj~ zDG01b8{v=~k<<$L3smse!wqZ;4;5WQez5W)zP>U^8)pY#Mv_Dz4oVo$7~>VqKZkrJ zq*%|U{65s?zh!;a*0<8IwL-v#RJll{kw-g}$b_ElgIrUfr7OvrDcO6c-(EWXx;`5C zO6B12b(EL($O5eJNgD4eIrAZ1vn~!sHO&THH&T+_+s~2O>NyIKsL{%$mC$YgD8cE) zcGsQ@_-%BG-Z8yvGieYXXlITcvKGfN7C9Ypl?OdWD&dBvz9n(0?QCZ0x{t^265PqE zF1>MW4*vk&EzE*xqQepx%(1rf^JC`PPjD-}(R52aZsYzDkBR;^y|B8HZKlTV*eYEq z+mjPc%?!*z=gZE~&fM3``t8=A9h}y>4~SO$MJWMokv#W}hT!s+$jSSmLLX6#;<)&J zENZam(%4&S_m>v(w#jXE2T4mhblV%DSeOot8ivn8E34UQ?*2tOm(ic2n(yrY0PqGo zF{StmS+>$O7>GOMh+U=zcNSJFh5rC`m#+tk_;NecwQdY@`KxQY8^6{t3OK`z@k#5NM)J`jncr)x=^(yFOfK#lhS$2vuRv0H3aV`E+69xNSefQd(czsKno6Pa zcC33iM!*;(HhpW*q|(ynXFq1XBhs-BpQBtw6B0jqvD-w%jzMWyWOn&gy(_HnN5OxH z{v}D5O1FmDC2T^G%5I}Z2b|>-X~D<~kUQ~P-VpH?pQ-C{_*+=DvPG4IK`va}$PPF! z5}Yp?97w*k`lI_bd@t~9dL7I@1e#biIY9DLRFYY2hSGO@v6FQY$=Oxh1A;Sx4>l(k z%dLy03(Cjki~i5Q61*XG2BGl-N7bUUTo1PGGuwi4KtEa-Ts@UBw zr=!TRf=L&v`}mjz4p_Rq;epY1T4Gh2fgvGUYcI z!^In{GCFgO+>UWxIi?LdAsvs246=>DZn|G2bivz&*~dKYBh=UL$L%rUjcGm*&bF_0 z;!SGOH(BmvvbDC<=Z`018DUQ-xJ+OH8nMG2xy^nBj}mxy!@Aq)TD12zvq^FL=d0D<2>!2=VM? zCyagduTio16RG(^>%tn$zDl>2w&(X(K&Z)CC6ox%l1?_~BL^b8e}tNyhluppo^3Ny zvxK8|?DuLSbE(Ry9N57OL;05g zukf9?JxCpE2~>;H);6Y@b`1ws@GhBcu1s1v1b%M)1~Dv00|8o7`<6W5Dm z;kS=<`#n>}8V$6U5T+u#ic#l|;ocSsj#ZcC`Il+P$2G%Cq(vORY-IKi*Gq zV=QT=WO7Z!XtPLEj^;GT=Wrsl{4eoiOTN^u^$4_4V}E?Z&%V&%^Cv~e0}`x+ApmaK zwD&ybvy>g2mZQ5;KSKT+`~&gT-m7FZ8?Wu%9`v824!*^VgPY;fr zckK`HtM)bcEemLWv=VAQCe~RR2;tK$rIzI+w*xVqLNJiWFlHYt*C!sF8vLm7Z-zWu ztX}E2$*La{+FZF?ai}%Ho)_W5vhYx^-pb>rBfWGN9y{?jfn4}|;s&wy>vUa@X@3N! z0%Z9>Gs?l(4Bho@J7yc@~)ilFx@y^cHSfXMD z&9NVE5&$=RiSlk>c^i=TBNfGXZ{hEUd}F4QZ*gIJsxwI%*j&slXOh^GSBTa>F~H;? zn*@$X&TG-N87%c3GFd)1TUuN;{l0FOGRb~sJOc5e-%1c|Dm)k~oijVDo|!RcZy3v) zAPk(G_oB~z%S|tIoSzeG*6Dj?b#-qElMy}ZwUkVwqbm)Iw0f!>bfW&xUAvB0yj6E& zs4JI7nPFij183yAR)i72IV$<%HOzlyMX6d%t!WZZWi`$gNcF)YmRIL`PdRxu@^S#c z9l5I-u9>WOVUjrnx2I?lExARuVyB>aw>K(zDwy;lsl7BCPUkT-mx1)#ud&_QE~e`3 z=02rx!qPQPfJTiRP6zj#IS1Oh5BN#7ttD->4+z-3$$j2Hv$Tz^He(D`<8vHf;~S0+ zF_J4P9~1aO&UZx8(@maMX%^)jB$hTEN`0kWLmt}@R?YSIi7vFeFZe|)WRlSlzSVs9 z1sZo!#5*j?zEHsdk|65DZUA(mT3X{$Oyn(_!&kSn-*}rzO(J<=$G0p z!wKin?&d0{8{)U9Msg)B0W#%AaIcK@1XrEwUOm*K)Beq?eX#kV0$Qvj+cZNzIJst9 zbpz!A3EXl>$OgD;jdJq(5{M&$IFNuMl`e7zIml23J7BQIIAPZ26P4NO^Z16_Nq8-^ z*)DGF9JcBl3UO-!2T2|MUB%e1kXF0_`8-S&`t!RIjt z0G;kki=M=OwRtC=OF5>xvJ>GKSC6s&I;z1bD5--dN- zmxD)mE$4vkDp5qk1Cg=7WGn1J9-Psnj|=G2K9!)}MRBOP4J0t-r;wA9U85LL&vFNB z=9{I(Yd@U_iZvN8Uh#i<9PmX9YJF2`1|9ilX6N#@SHs#=M%Ti(%L}MfOKV7l$4uZb zjT;2#sbSojMID!!ENM2McMM`B$C$G@RXG{P&<^gNxIVRKN7D3p?g6q5YV0yr&R2k_ z$2t^y*I-5X{c&(*vTyTMHdq+R}y6WpER4D!QdQz z2B>^WhfmZdwz(G3+8CE?zG+lkZNS^KxqP0S41tn!T~+RdsOU(qq`V?`;cX>H^R7WS z0u+WekhgLRcjCT~@Zap|@jK!dhn?oo=kWFF4>~CwK1Ub_cHAtB^8Ww<8;L(IdRB9* z81BuZNxNu%cxzrB@N5PcZlu1tYzC5BTtgcSAMZ0TQVs|Ylk1A7a})__zv1)Q*?{1#iN7NY&>1!;*)E-c%B$`#3j^& zTe9Palg)N2g?$w{?TV{Ef_@fTUdINntlZqhS9r8jx+au@r#Cn;D8Rr03->1&t$kBW z@%-uK-rQVIZd1%zC3A6Z;kXPQM<5RSRE%~Q#ay@3d_APf*EhEe*MkV>PG69R`^R*< zK|k(gv(mb$$?S8+o%Cq0pzHTu7S-Upv(TZNPKrSCBr==UApjT->+%vgAyo9^HT1WR zJV)WL3~RsI{{XZ90F8V}sp?a@jh|+D?4mpe@~*CcRti8+u&2xXL|1`DK8K@Ro8J!k zS66Pq)u)lt;4TQk5@Abw1Jj!K?JrpHFN{1lr?#zUtlfC7(lH;GrbOaJ3IeN%<&kZW zxX$i=Va6*Ov)5E~Q>Sgt-rO2 zcx-AKECwBrHR@W=bsQi976~LPh8P>SBd8e1IIk46@P3`)4;)-c`%YT;r7kaI`L1M0 z@9nNmzSt19Nnr$`Lc0tG z`7k}Xt{uKD{5pw@e{Hx;ryEgW+A-O1ae>5 zDQTsnSO`vXe|IGFF;Tz+a4I_-*AZvpKL#X*PYd{RZCc_I16&B@aHNh`%$=2TPj%q_ zb?<&P_~+nD8+jw}M}hpeAh67KvHZ{3nopSL__8zfubH1x)^4>~ucWa}Hc3Xpw)YP> zVmbMNL|&Y6l09qaX~VeXJGY_U>E0l_ySaOtdlR8Eg5}9ajmqF+L_x4I>`BiU6%-o1 z{lx2Ou4)r1%;qrhq)?#^&UcXQj@T>>d8`Qfh4R_>iS9?6jCqW3Ut&;!29#!#vxE9EO+W=eIcY=|rHS^d{$br+A;hI(CmPgU+`XJFV3u zzva0gv}#o|+_MZQ++C;iQ%sEw(z_T!U}1U!mHA1#!lCCm5`KEsd2M_)khPzV z{6nTie*iAdBzU8FQ}||v7DD}bP(W^_am7Wh+4y2j9Qb?0TGUs%ilJEq8|9qt`=VsE zR6~vk$lAw@XBC^G>R%9i8KbP%8Wp^j*5t6gSWcZ24nnkpV39VH{5cFp4PhyoI3})+ z>{)n2!a4(MF!**G=8;25b!hhRTf9u!4sK`0)F0hDiS3@-79K0{C!6+vCrH!na2+0d z=ax*jPb{o)C}K0**DIrVOXFvWt)uY{qps;z)|ZUS{)`RU%m~bk>c$`dAKl5^4>;?Q z;{N~u_;$&yt(RW4ot%e;(8~64%6{$_WQ663z*5Hts5#=2)7}={PN!JC@XoV8*)O%* zh$Y9EmT1wVT@c?eTc%duCOA>Xdy&?m(=N1a3FWoZbX%*N%zyx9^B@hKtkTJZ+@r7~ zaqW&q7l%9{;q4OYDKES((Ot+HNSWaU*mAy9uPM)dtO5Ghy6L(uqp!4f+U0~JOjlPi zPj(jJN32Z>N%JW`cX-cC; zJu%2U8tJWm4o3tN>pyL{)vPCsys+4`e`!TO!dK?=0PljQr_1TpUVJLlW(h5o)y}DI z>K03DCyE&Y1M;Yi_q4w49@D|&BD~<6PYPO3sd!UI`z7F#RGM=$iPLEdws)ASBL4so zQy>HkuocYT@(A;-0&f&SXFb<~BD)Yvg>5|BH1SHizJ1F0OmoQu9mB6P{G#iXS@<41x65{bVVTK zD`9(rE6My(@k_wo8}S%|Jx=EIN00RDc&*;k0GzMN&z8mu1}o2~uBgSngIcSex2yOn z9U^#S@u9dPSY>NjEhI=RXpsOj4LcMxlO z6JB|Db20wU)K$F23BU^+Yq69*(}heE#tuoTv`>vc5i}VNqv7uo{gUJlnk!p)ts`QD zi4~BKo=o6jhDqu;uadl1@w4J~z8czl+4L0L&u;{F6T)8ulNHsm_ODi61%dA6tV zb4*Cu!^c;vcE}Yz)bT|X%Mf}>p;ye?oTzmJzd5bpf^BLZ*Im!h4R7O~l{#Fvh=!Tr z-CjlBw-DRF@y^8LdC`Un8MBijuLq0{{y(n#Oz}b_+Lwrgm!uG5dpTnk+ef6E<3c|S z=drJo?VraQwwf+93r5v#OpJ*nQ2mn)j*hG%zyp@X?Bk4{E1I#J$9lb#noY2`vW7+i zMYnf--IzBBi-LLk$DG!&z`{DjC0gg#Px0@?lIcI#db<6$PggruOPMCPgyWwoUAvUy zw&K67d54L7Q)x8N>K5re>oco7GR}}r0AJu#Wmosw!x{UzHRhN429cxN$!h*(t-}DS z&vPn5Q7Hv3H}w6=k^Ni+)9uEDZYJnlm?2Eikedjbw?)PLbB@GaH+dhWBS z>Nc9X;ze~&w%PfIwZ)W!$tq7RBuIuFNd1GrwqUFjoeRi+sYebDn#-Oe5FX;xPj0Dg!=QE%B2MNH0NjNbAHu&FN8G+ zbu9*Fy0nerd2Y1zNtN9}^CJ5{%_IXHZzQo90IoA$o235$!A*4CHP+)r@SJ+=cI-+l zu>%@o<=!Qb5rfxoVZp%WzdL>%e$OxByT;aJ6KVG`mWp(Ax%*NuAYr`KXCb&LgN7I? zKhks#T4|vW5y1qJ@c!RTEpp5M)kVLlZcI0P(>ZBekg1ss>vv)P6&2gG1?aktS zTVIyz!V~Jy+3Hgr`gPftQ&kqGXN?^|KQbe6^khsE(~AA*_@CjcPlqwfuXrZ@D`^RmRj}3M zgY1JUKX)6bS)wZ32Klj*m0(6bEcl)A!qOWRxYq9cn~5M`%WG`cYDW&pW{fOfcn_G9 z(~<$M>@|D)(MK&>cXmH1BGvWzbk??mR+QS~%X=GGORFoML#u(f_2|3DQ(U?y!{uMI z-@yg7>gCUy8(cyek3$$a1p6rf^#E7Dc-m_p5ziK_{hlO7j1Mwrh6z<0oPOhIZ&Q-Q z`q!2>gtYRP68LH2yp34P<--t!j1J5Q51yN{IL8(2({XEE&NiB7eWrX{*8CG_i>K(; zzCwYC?c6MEqZn3FxChsxS6e6T1>tWJSzE>9e-Bz}-cw>Iqbw$XV?Jb83{qIhIKwF< z5C|3XK!?N@*M`nGt*pc&bFA-{Q03ZRkvMUHxV*8uQ955H0W00_pF6{5+inC;v? z{{TKBVU3CCy9|W!*MpzJrc{&M9J-$GsD98g+gVy!cvjv!eUXcB+LVnf(v<_0k&2_g zW9BAF?rW2{_`l+~FQk^|OS@Ta5GR`)?QFz>!2vjC_5iLrfnGbKO{Qq`&;5;I71Ze8 zEycKLg7c0x<%3}V023Z_$gYW|)%9yJ72ca|2b545X1G(_Cr=P`7+w_zC!2_(uwtWWa&=G&a(lTjN>%^{$k#LF~fF?MT-OSvFz zVH%ZQI%NZ7brof_REF>S9(iwm+YlQgg@INL*fMXFDCvT8^c6{L?=&0JZuWaFOKnu0 zU(Za8E0DXk(;1^=WV4nq35{d(Rx-exZY6N*+$#`s*S&jRhregfj~^B+t?sP6AL0jB z{{T~eC~G8BgN=gWm7Pf-5EP8^Ytv`_jsE~??F&i%(D4o42(f5c~YKWi7Q-QFOYlHD8!#@%0+MISC61LTEu6(_)!xgbGl?}@vJ3;!OV0q^`6*8!# zN>27@(zf`Cu2|g1_FYvDCgiJI(5Q8iHwPoH?-7BKanyXI<3CEJrlx?`X2-#^ zq_9X?S|xy+RO1-)A_HR+&)(xdbmukdHcw%sL1>zVt+Wkt7j&e_G()${Ki`6)Zf#Wh;kbAeUfp}BF{{RiOKPy+%?d+Vzm-FM1QQ2{i-T4?Y z0k{&cxDMDgJ9uBjT4tdpqu`61OJRVfW}$1 z-v!uT>H4r?yHxPtj>I^~EM$W*THFkQ^CEee7$kvTpC;$S-XGRh%II1hV(u|7_HE7Q zkR^zoRpE_Y*#O|m*g522@Olpmd^YeVkY&8mHB`_`E0;p(S!`{P7gwxdi2e3 z{aWtm?Do9}{B!uTIs~!96DK z+-p$jw`~%kc8d}XuIJ_j+W~#dj0 z<;kw{v^qpeJ;{drZkB13&~NUMq`Cwg_>WPkIGj7dh`{&I_@ga zIb~*0(IW6NXmhWMbyw15C=%XFBy0BM<1gif*vskWHhN$h={!N;yKn5c?)(jNFNwi6j$k0gGEkBb81@0C@(}_~#B=F9-O(=GRZOxQk2F8hHe^hE;_oWZ?O3S0RgV zyVM?|J?gdAi~j%$`6Sk6M$mOTXjsQQL8X;KfHy}hr)JkYF;zScaw{%L@X z-0pEA6O3*2uVdhw@i&?9?^x3#irK{L5CZP*1<1fQhA|lm00(nAXV(-hHQ07Hp6>VI z(>(Uj-u=tzaf&ow z7igNb%lK=t- z;z$Clc-10E)fA853?6KpU^B9iamndgDs8S?^C3Soy0`H^!W+#pFYLo@pz2OzZ?@jX z_pKQ}!5os7jwQ+9>;!SsCp@;kzZ|r!dPx>*>+5-gJXcn8hPG%_3^AGW0;ta^-|JI& zcz7ShbHx{g?pwmTqy!Obe2Z}$%*O! zW_^%O<-u3p&Ua(3(zzV~z^a9!+WgGs+tD07=fp(4^0dzzMGmI8Wk%BoGb
|11H z4!2IQe6k&Lje zKT`O?6s<0mtm5;{xWw_e>4S4GWZO1Cget7)@E zqutKyi+xV%W8NPbVgr&DRCZjSU~2|jeM7$cxjQ-(qJoa}bN$96yLkPkKHZLjJc47uB=-p?{C zkl{S5dlwJRRwWWS8RO@W4^C@%#>&QScRiN>0OB`-1hPu9+G$c;v39w(fsu|0`8f?C z4UNCM#yz>N3iDO*CFQ{JZj`U7%Pqmkna#9`ka3^10Ol;??gPQ-YtQd~A^1|-VE5)7 zD*8k7O{b>G7%<$pOuL>k1_&PY%4)aTm8GldI(6~9nZu#2z5B(RApWMv5yr z7!V#zNi$;@cA8Mn$mA6vbK1Gtyg>S(7J8iCdM;Kt=R$;W!2(d^;NXBU`PVHirH+|* z2K&fT?8qb{QY4m5px^o|YlLiGfR*<`SZ>JJ)T%`CA;RToKO)tzq127n=2g+naf{drbL& zDhXl87QFi+0Bdtx`o}!)iW(Jw0_?d0wQ>rz(3~gDFFC9S{VYym03{OH!4{FL- z?X7O2{@jaGpKfy`vKdi_sg517ARe1fsU+817J+|q=1W`Ar8DeAMgs9p2P0zbP)AN2 z{x#0e;k5qJj(uiLQte}vn9Q@WhylhL2)v~P9y!T5#apssgta2nHQ$Hn8$)L_hUr*^ zx6_rRb$o991Q(D3p?L27cpJH|Tky`GVQXg$nzx9)!xJtI zjr?)TbgIk7(-S`4NF+Bwk%7UiYE5o$6{d%VYCa6RONO!W)%K@sp4I76{3$t0JvUu=BxDjqiy)hNu^jnmieyuff--&0bn^Jp_rg9Yj`sIc z)UI@d!TUl$s{;doa!GQk4*<5@^O0WdXXC$zF!*tGf*{asqWR*~G`obgkTbCWV54xM?Yof_+Ga^*`4#4jla46?^5UP0rU`D*_F_MiBT z9-gDY`rKY9)7&n}FKy+qxj=aNq>nJo+>(gA9s%e+eDN>FuZVsl*KK?!@ms|*={8e3 zvT5>KD6*6`LgDUXk^XESm~0Ktv92lmNpGRFTD6iteY`WHe#-v<4ZKZhfAIUnv1q!z z$S|=GTI!bD@>v;@L>Dc*j(n-X>zwdw^dI4MuZ%Ul9{%UV_F7c&7?N0TG;=$(z*2V? zwvsfOCdlD&Ht~(5kzdZ&!5@b|v-iUI;_-fy;v?a0O-OTlWNc@*m@Wtp4a9P0J_b@b zBV+yGPHXg+;iv8Osafi9Sa`2f@Xozw%1M2twEAr1XDaV$dAW9f2{ByCRN;c-r;~_; z)Avo&PAb}--SI2J{{XXJ#o45^)F#rV)GY^@C8IUYpQ+loQVfqAs}<6@T;t?L1OhXP z{9OI0{x5tF(e-B1wZwu;{W-qOjdNtT@jOu|SjxgYsUv)S=ZRUE!1=K2sB^1e}5is==wYu9+??bLQ`e`bWhtiV)dcd@Aso-#Csa zklMiLSUJYwy`sPa_*m}kf@)uezp-cS74bVzpGwvs((I>z6u3ik8s1wHIWk0_yi9j2 zrFb}FUZ3FK+kfKcf;1oOzYAH|+*-WB%+r}5lc+Gt=`hAqs0s>#Fn3p_=)blPjs7$1 zHvSa&vjke3i6b`tL`83`B~XjxRA5twZxD)G;%r&1BnhZNn}kKiBpCO?dB zwRF=wF#1=D!lOc`bVsJuviWHN{#eS>&F8!50883#R_*e0p;QigD z>Rb8u3#n;Vw-8DL1;iItQe9e?LUI9k+Zk=3FC5pwZ;UP~@)Y#1oxUdj0Krss?GIHx8PlM7?4ys&lxcGlXvj-07Tu$U zNVCZ*@%IDf#!nSVR;w!({u673{h#&iQ^Hn3f%3+q9;IkxUJPw*arT(v zjE<%WVsLrqjDB?dYxqy9%i`P1-;R-Lz9qAQL>BF*T}0__tU;8$uw;;cpacrcz+~qK z2j-dlf8)Q3Ul8Tje`cQzYQ7h~xl;3s8>F#fkA-;+ygOssK;8S;_~hq|hI~)(@9hC( zXNSj{-lwYDvxL91vziojK*Mp6#-&-+hBJ(GB;vZT_&Fzb%&JvKN#fz~8YqRPfva96 z+CpVmm1K(QA~=k?ItIb-f>yqMfA~qYaWoopOC)6&H22o~RhSY0-UwzNPMtHwcM)rU z3v4uq{5fSFqol9_(d-#0CQerHnpnjv&E$N4(9&u!E&*kyP-vH zWXPB0;jl><$j5s0(y`AZwnjarjG{!p@eSmqF}$P*iym-tj&jU#{{VdY6IuzS_-<7a zjXubR8UFxT7k~R~DYu>v@eDp;@eTc?QSMXaT1L+-n|*RZs2xZLlhckWfztGq`C97u zk;Y7MXYtAt`I;_6SKR;A{Ec6Po+GxlZF^IKD@-=f_N~37E0eX*h5Q>F`&HinXkHD| zwTFY^m-da8nsX$6d$O<1$qun?RL@+M>Con}yhZUZT)sCDYq#xittpT)Nr=j0KYdkq zo;fF+_QiQ*+RuoV=RO->^mkGvfT%%^d1BiS0QDcQf1`wDC8mKbH$LF-map)p+Ua!9 z2tU~Er(#T!hKeFfa0JtS^s()Y7hM{?jflTFltyCj^D_Oo!y zlg7p%5s`(j&RYv#6W!^n9)Waiwq{77Y{(Ib1ngiy!=9Y+*NWEA{1vTu%UFFV$*BFJ z*dyG>3Gx{tm9S+yPIeQ<ON2QRnnraOjE#$ z!LiN?C?~E3cK#Ch(eV$(R#w`+kEz)kdG`_)Cb?BrJ6H#l?IS%`1d8!14}&_3=;K|~ zd`G7xmCB6Ar$8iOjQoo%dV0`h#8*Wu;5u-81YuZHz4h<&NcKg;Y58 zB~E=*9FW$reXP(k3=v*^FG(D7!Q4TQHS^2Qh?i`})P3_;0JB7j4Ze(Mq|=`V->00Djz zS_v$F;Rs7x`*M+7HQFg)WC}`ecX-$`>tgT~??2KmOC79PTvOyjfsx-qkpo{`x6y0yAJb z#(XbW zGFieG#Fx%A=o|N~rKpwR$iXa#jUfXV4H+bLAXiJHe#Y7|Xz{~muik3b_pt+;WRl?@ zY2Y!-`C~nQcC34f`@SEIo&vcqb)y@JZ{V5HqPlq%p(sY}@w0AM0P?F40QdFsj~V#a z#2z(@IIQ5*w5!!(u-)nZ0BK8Ljmc|>j>~y@#z|6pZW*m3nANqU4wqxVtp5OIZ3jWt zE^n@4)Q6U+6ko=%TRevWLZoG7B8YEW9dkokH3H3w<}267dE$1R%j4Qk`W z*II115qOT;<4?R&S~=n}Svo!rzo3mT_IdFC0El%vtus!B-0HefM!mIbi&&a8 z##GA=)|EE~AM(jgdmgpJc+2)e@Ds>C{M}{<+<%9yRjDprD-Omohe57&msV6;gU5Z@F=?eBe)4_kVPm8|K zZQ$=2>Cs+jip*t8#+DgIbNk4qiT=?OsUvH~gJ{MuPakjmJpHHq8{umygLsot)ZhrR zrje+uaY`7SijnF!D;mzBMi+Q8o~-9J_`!Pj{{Y$&N~qdg?3+Y!mb;7+0o{O6RB?=C zs6Fe5nmZ_NUSAWwp>(omdGnI z;E6@X$Zoe2%E=RfjH$>1zaH&;O4@uDx@U(bOY3Aa->FZg2%!@e_D5*fjdcx3#!HBTI>$C7p@m za_<-?p5u<)Yo-`Ttqi9&{pNLx@dLt|mYTYLg>OBT%yQ&Q3~|n^7-WfFC1wkbHsx{n z)E+AE@5IN`2DNjiAMD$Gq+XU&0;_${LI)r(OupW|O>&l6=ZFoPLl&Icnug*_br2@r z{$el#pz5dBAm*xRlHOn58>?2aL_(>ULtHwM&}DX})f{Ih^sb0fOJg2w2=BD%Qr;Lf zc@t2KpOr*xcVmos%JIS2Ruh0||u|i3}ASfNUuRzuH z9}3%DHKwm@>xE_WV`(1M+yRL^xb~J`QM>95RMNGC&>9Eut(-S9+`00S&IHVbNe38T zFl{{Ya1S}FwYMhSPV3?S0EKlw63%qktgaxIM+_QtTV((bn{MV%!F@nD75WR{FZd_+ zjpCVL@dt^n2Adge%+_wwxndQRx){P>XK@+H=nhSMHQ^r^ccUvO$s>_{QPYq!j92P^?A_sS9LJ&DYu^tpCe^HMrD<)n+s#3c2~P~R_qPc= zi~^JLzHmrT;YE0u{8ZzrJFt}Q^xXG-6ZRy~JRhkg{;T0j?-lqZE%7z~0Eh2dTi7Jd z((;ksia6CflI&`5ES zoD+)u3Gtu6&jb8T(X{)Yh>xgg+OCwx>{``@^!9g$&Gs?r(7|hd_S;t`cxIjpd8JI9&k+Ha0b||4;Jh*WPk3X;hgH+Q9}8_Z z<4+qd^ts~l^%zyYP}A5-V=4<|G4h-N$vHT#U#v@Lv!PSw_pNk~D8K!Z{s(xs!;xxQ zO_N;3b8cQeL+u6UXr{vlhUu8d%FBWtJ~M(iuaPxx+0Rh;8LR1bQ+WE*U+}U@#M=0+ zRWGfB@w&%Rg%VfVaq?Ym-JTc`&!m3Sn(vDB-7Y22^{c-P-q@zdE_BIRZCW6?*>k5A zYgZiOI7eJ?!=-)|{BiOB0LMwK<gC~B5gr^8lSkM^80jncz- z0@|+J?E)OT1_wPEbKF#(9kuv{E%>toQoNZFRym4PqFYLdg_+UXNh!u4^X|}W9NRM!0cKJ+b>GKdVl0aW} z2a4_VUkrRF@OO(6=Icka)>_%Im$w$-RHxmInV72#(r00>r^^T()2Wrgm1xg?30byQC--H8}D!eF1gd;0$XfjklMOU5xs-XMvEsy@eQ z*ut%jK?=ieN#krq*WW!pxjJWr{8a-d!rO=}FGDjuwxI!amFxt zIBz~QMXFAgFwJ9aX66RV2raTtPcm?-sKAUW=dW{$;&t0iEj>+ax3LwbjiYD_X%4@l zBPE$o?7;b8f?IHT93A-_nDPMZPHR&C0O9`tgnUwD)MmQV^j&K4!ojCp89{UN6lHhv z;hI(hJ1`DAlS6ptO4qFJ{5RuCAk(hZOi~*)a|*JsRs=cW)ZmbZk_b4apN4)JyV1Vc z;6ZKTDPud^`$}hoJ~Q(yHu+KQ<`(tOYUkB&p`(2f<(?+9_=oW4Thp$!YgdV)$$+O3 z?k6fdrBE4&JF&6HLC6)>f5KaL?F-_s z#4Q(5ZwW(d_M;GMW2UTGWDV2HD6&kwM%|gn?^k?Br+7zFvUvPTkE7}jgo$pPvEDKe z??BR`0!Vd{0{1-hxnE9Y_ZdQ64t~$!Hji{0SMc@pm)=7-iq#lI!*4j6R`ZAqdX8kjv((qA(nxiI@!R1Fi}9f34_N7TU$M z$5@l^ytYy#5U9up5`iSFLGFQjVS1XcY4I9cw2JS=mNqx0%1L1IBw3bdu%Ud(%VtiEu^uyS2DDl<+;-tLh#7582ppnf`a59iS)=~R{sD~ERqFRbHhBa#Gy#X zPEOvWdT(2JH$&F6EkjE1%%=O!juQfxAt4IHoW~Lk?Tm!U%ADl3a4t&ERbw?Sr;Nei z9~O90;yr#_nKX-gLCwve{p8EIMhO@sW2qIQcY5kyLZzL)k#^Bb@Y_{eMl!Km zA_cYH?YM2S49g}#?pb)?j-sumsp1a|*m=`h&t)n38huV?NSZ}B$PVYtB=p=E`toZ~ zXfxXA7nj}|yMjv@A(Y$4q$E*1YS>}tmJcIs{w<|)r+^K4rN*b@TOHT8ixTPJ2UDnR zjwJ`2sFA$V2gZ5ixub-Ph11vNfnh-3g-X;nfuUBDOH_}3$Iec^pl;A(BCCXambGTbab z>4>VsYk?qDRV79jKQ2hka2Njo7d4G`-6Xk%^~+eofg7lFUB5BJ?5ne$NZfOb0p7Wf z5Y6!5=yn>8kFMWGs$LtEzS5E?p5EOfd_wWw$OLZEjIj!iP>s0+Wa~~XX>I!#IIfKD zE_CtXsYbEkyAyqIK=N-Txp{V+{HM#e^5W<}<)M&zW4|AH@oG|&>Du5~?gYopg3T3f z2_rv#AKkGz$IgFB@*Nk)R~{eK9sE5kmaQI7n;9{O{JXdSd*w*Di!8j3FPZ3vDCw>R;C z%F+@$;rjN?Ue@e3;#))@nqqMfjW&mN4&T@uHcFfPvkc`Woo$G|Qg8Rm|Epe)~ zy~mV{#@^DyyO)u$58mSguWIUbzlZvlfWlnGXu2H!NVwf{tGVNqcPpEgXEF7w!sV7- zL-&t@dhaBpWRb{G{OaY#uXCz4`LwBIg^-2&JYHlhBO6?nL$!U&?dKV_fR>wx*Ok}KWe{j5J{p8)7l_!nOA0I1?!(<{CQP7;X_gWPvP2imAayLmo<}sL zE1KCO=-ns78gIlaNiB`z!3@r^x+dt~AfCGbY_9`p1aV&N;BVOlptxKA01s;57`C^S zu3KgYk8omPnO!6WvT*GisOKx5pBd=i5oGXHyjC{1H@+9Z1#O+)J?y}V9xa>_{Ez|M zaqC`-;fOqYuEGa{tTmZbNUpM7I7>`8JDEk>i9q9xpDuHrwS`J@-7`qbvGk|JuMlZ} z3uL;{^@wg3z&Qr;#?C1$*pZSNwzndy~UI7_IcPd#P!uC92>``fO`*c9YeV10g-KsKdFh0lv|^9Lsfc zrAKZqKy()W!H8+s_Yc-vOtJ^G}zO$FSs&E)Qsy87V?NYpPhR1nS z?WK*QC3_xit9(rG7mBqzYwwHku8XoA*7gxaEzRt)ecjSaDl7nJBLFrB9kX4l_=omI z()4?&^gCORHe@nKCO+2j$-!ohJAl!r95yxs*ox+UHvZ6hpT!+2+rx?A+l7KO+Shu$ z%q@TY(TWg`&>)tK!--)&H2ZwBQ%|;+I9f*akrI|{fnc$X57D6{3s z&@GHUGSjRqG_+u0w7R#9iDLr{3dqkXZgN^I5PyhQlSv-8r$_cr4Z zp9T1DTD^zBHZsm7*pXd9bv&sWs-50QxCY^4%auR8Ob~XA0iTw2JyTuRbup@1-o-Oq zH<*bDWq5;az#DSy(BLzYN&G9@Aov}v+=(AuNo;3<48|27!U!LG?&m+k82(kJeieAL zPSeJ#;Qa1So2;z2F@eZ(xIrFI1LZ!Wo+-kWHSq2Z4V+`^_WmNYw~xc#6V%tm zi!bhVD;aKV{%!`(DNycq1oUNLk^t>rTKH@BUj4N1yePVdjW6!B1$KCn)_a>dF62Ez zh4Um~xp~WuI&}mHJ_dMiO8AxG%~DSq8*8|%Wl1D!tCx=I*UVGpsc@*^u>>GRB;fVO z7IfQx+7m|BZag{g@N180Z@^7VeWu-GY;dy77FfXwd1H)n2NX@T?GmnzQ^Vg0yfNSn zJ~8n^$6dGAr41bKbW-BTR_b>kd2>sQ?Z` zTPt~f7Q0&r0K*al6Sh8W`v6%1?eg_Ks(tqFVc76x@c#hBZC_7Yt6OO8QIy7h*$c-q zV?Q$N-mF3F00;Y~veUvoI|u-uQ^r4oO_Oi@2a(66eTU#{yYGk^RQisMc(y)Vd%|ie zpgIBkGUhXaMsPu5Nw0UbgTQ}fkL^Aqia9ne^;qQ~HaP<)I0m`hSA9xQx&PMrOW|ki z2k~3P+8BpQ(luLEFZ=K;#uavgo4)7{PP4A~TTt-_ihM0>klfh#aiJF) zy_WD<)Xg;NZXhZ{MjLSo4$a_!S^oeJ{x8q*-$R?m{{Rf{q)TOJ)#9Cu62Nhn^HH`* z47ktA#~2@XwP*dQIY)EbSS1I{wmxw9aqvU-T-Icp!rDHH{uc2DpA;85hM#swxwndS zA~ljSvXWa2M48xjkO(6cx#Rx;g&(psT5M)LGwiWk+Di-H!8eg4GC2y(43ZG)d#kR@ z;~g$U-A{ zqln?~jLNK}{{R;|xC1;^UcdV;_#eVLUFNay_8VI*YVs)JmsYg4f#ul=$c@tGUoLRi z0D|buhvbkj7111Oio4Y1od`YmKQ^v^VlNo{P1Q}ujkRk_d%Z>hEVmHF6PJ-l$ht__ zeCxZ3HiB0;uS@uK`vz#90MVBB<0;ZKjS|J!N2z_HJ9}va!{ym4C}Qs8Yez64ayFjz z^{<9LHa~*&*S_(Vm#1m(XF9@XR|#o5Oah^J?tF`jdD%t`Ga{92fXhNC@IUBFoXTcI_%cP5Y8xRy- zEzF3;og4<-NNCa3RD!G==RNa^@r^_F8`E^j-s9oUrD1!jiA-x0@=G0)kU!;{KzBy{ zzVT4m+zA}$x9tPq>9h&7?}GYWmW8WLC!cGk>MbeKmANwuUnrHfX#oL{e(W4D4b6Ei z_w8Bn16x@%tGznwLcNG2S5}urV1gp~DE_hk|dKbe9K$*#{hLubzG{d}*@r7Kbmx{ajn?dLS7~2aY*nSy72H zd6E3_zW)H3O5l_8SIXC^CGD}l(r>PA;*v!x1laQ;RRC>CpCOe`@Ri&%jMrrxFE_?= zx|VlH`qx?b&*KZ5DX%;msa;)15k*@&n}wG8ILF95$!8@b1AwF{Za8Csk1x}{Jx_C~ zuAwydLs+#Sg}9P)a;%N=l?p=S1Kn^`V;t9t=(@#@p>&5?Y2^L(Bul9nVMqHXX3{#2 z*yq~4`#`_YJVS1DeSglku;E4gC1983APh?QMg?*Tu{p=Rb^92@rNtVJyEC}5)qF|u zjn9XCM!Jj__p&UJ-D#1mGPI?c)5-aDKOfTugl&`Y>9e=?;aA5+$x8NH{S+CBpfRS`K!(}AKRDU zcDZ4H;cpCH$>2?ER=!*Lb$<=oK+iqIU@x10XK@^hzqCe`qv>9FJw2Q>L zg@eJUp}LMuJ{bJr^8BSIe>5wB!yZB7g(ALQ{{V-Q>DM=M9WPk7Z7{!)cM8Mi!7AJX zlF;5dd4qS}Av$y18u`l8isYQ5oydpW0Kv8vg)>^xqOe6^@;6_KTP|%F)DL=H1#UMZ78#%}~qtkz8%S zCpGyctp3#(b6AT%5ovmsovGU^iS6ZrHrpUSD=S09@sWauA-e7v;C#p8ZAQ=HcBwV( zjkVU9dvz+4cWb?7l0*uD-Yh(FHqZwAq@j49|;DX79v0ie2#k$`@tCd%M^!x*oLzAgUNI_9ei>7E*ywY@pxWu1ll zTk2kD0|rf!0*`-e~Z_W>pDcNf3qYoP38iV+RifgWSizmI`hCdINU3`mL3q+Ov-J& z61qHR#6K52FRa{8CW9QT_%X>1owKydznsX*tW=Tx_6MG8&gIedeRo|;ohL(|NAt)n zJdX<(JOyWjJLG2>-L#Khxc!T^{{Vt)-fQ|40_y(rPP~Vdnk18PBp;qkszTYw`F>z) zss8|mJd5D(fmc)bmR9oKD}W=p(rl)=b%20lxwy1UNRFo=gl@xb2wY?irJ2)?h{0iG z&%}LK;r{@OuCC19I=Q&Dv_Q)p&9qVM=bf)7?;q-RsOf`S_Pz|!{4?RW^)r7R!pyFQ zIYZ7Ee+~q=gezqI&V8%)-@;$BN5O9vh`dRn>u;=SOdoa4hm_FZWEGRj%z*C%5L%#dgovAe1{w~qSCMjS7j6u6ymKe-~~uUb*f5|*WXO`nbydOowO zGPj0Sd65szZ8Tn70OO#M7ijE8ayxNXHFVc4^)%6S3GHlWk+&#q(nz^otU*8&6OPTA z{WjJWz82_`OYvUD$483RUoJbNJTGr>xZYX@alLwxheAOFbiuEXe0%Y8!WV1yH?&<| zCMup>cEUpHa@&N`ZQ8gwI|t#K>iW+&!gETOE3-TcQ}DzZEa5eXpH{UPTX`lwXOoY= zu~RaE$QgDy#s^+G_Ryu%uBE#0bP-2xjg|7F8^JvZGKO*wAo|y1b>S#Ak1prKRuamN zPTOng-Gpd9ZGo}@1QJHuYvX)7*6LixUVOHY}vkc4!7{Yez>DI^KzYdX0=}Y2i%{Yp4RqOTh$_A>GuOGa7Fh>%WgqwduEC z8KcrKg~pk09+!Cwmyob%ypTp0FDV?8-FO3x*9YP+jlL%Fn~8N>12m({kL~cGfUw5V z=9#2epRZhV*FBAUSv9I5Ld04n)~#!G4c@YUv892-J3#9GVN;OOs*@P+kf)xR75YEm z58A@l;h6r}mq@wQE#hz_yN2On^1_UbwzkNQhjtGxK+oR6IsCiT?{)99M3;J`Qd)5S zZL~t-LJ1=TIY`2e%nJiu-@{*rpA-Hdc!Jwk@OPi6JCciYJh6zEXz1}qv8e!hmM7QJ zF*sPwS-TxnDk!_N`Xu-({{RI`v+>pYcwb0@OZSu*Rs#W#2>4P&>gHqDmW`E{f*9A^ z-wk|Yd!hLc;13-g78zwmz3}gd?-x`+jOTk=-^se}7CljJH^-5;03E~m^6-D`Z}CIK zI$T0+a>K*kQndG8T#auEI*hQC2T?H3Jdut-Vb;Fp_)GA!N$}T%AinXxinP667j-6o z4tbJWKEQD8G4jz_h#&@F5Ad_*zEc&4rk#_q>(j4j-*fg);vel%@NeOcu=-EL=x;P( zD_HsrCu7S~E7iI@|RFs%Oo!VU?n-EZRWhCElQ4SPFe22G2is6u8E+9!TIegdUAI9Gcd_&Uw zO!m5@qr>*{1re&lGn8Cy!m_fW&Z;x8^PjDD-X!>0rd;d35!BAVzvQ?^o=Yh3 z31y09i*(0ke&e>>iiYCV=B)gUAvDQ6L*btjcooNstS+si($EmI+QS>aE3uuV>8Q=-fFT2#`g`oEPDvwRMLJVv>))av@xbjDxJE7 zjIizbaBw8sCPoP!SD3%8N}n?KmHfn;PUVeaPlUcK)*_DIP;DaUt9k1P9z>4kE4JZ> zlpN%eNCmPw)tyIAkHHt;+ZLK;-z05@=703biiH3+{H#%iVtPg~$2EbZd?$lK)9s)r$rc5xMz6m!4LjAL+Y z1uSvz*BZVv@Wzwkd&^1m8~8NY{E-0rKm@-G_VcpEDg&0o`Fn%8&m@z()Mm4E&ll<5 zKLbopwpmEQOmf*PJ-l&%Hg1gx2pd0nf}RdajAo{<;2#uQ=T?Q`x3aYJJj;u9lI}~b z#C`Z;a-(3#19G0El0c}mZ*`*uE>&wBy{@-?@aE}Yn#TE0nvn02qO`Y|FI*KfGYkUT zTL&JMZN4acGu3ar?-yy-%W-%J7q*ybqfj=eC78KjpkTyvj&WLACypX*25k*&?lr`f zq`|&;(lu2iGQyFTm2;nz%Pv922N*OB;3QJPP)2xVn?yw5FLV3?mf zV-vS>;1=U)H1@la)RzX%hfMJ1wW3%)r{I~U*DW;wN-f^q@tour7X~>QR1!A~IL;15 zaS-YrB+|7Atm413w-E$av$utuxEyaUF2q(n*>X7hz3I357sW3R&o#D_Yb>)$5ofwU zCRQrj)SJ125Jw;?va!w?NFu5?#N9zHXVJW8abbVt@7#+le`yMrmT9X=*3o&ZntdzO`uZNeevSK4yVp1Qu|LKwq2h0vw*W;L>=LLh+85rls6h zCjJ2>?{aL8kR?rq<4ZVDO5kOj zi8&|upC658N#g$i5%`KPJ6W(zLfuC<7e!&8r#NN96Oqs%$mccV1IFXSw*F)nQCY!; zj%emzHCU-RIC4~Tzyk-;xofW-T59pA(%y7wp8{_!P{rf1YlsRmW$vQr@Wn+I`mu?9KeefZ63T1%|l*%h>o zJzm~MJHFd5+8vb#AO{Md4?&#$0IT*5J?>*EE9hx>i^UoriqMNVUFQ-Xp3W0AO^yNF zKXuf4;1kxbMewWQMD`HsdL5VA5TJNui!$6WMj+w@&jEo4t9mIIHO6WlFpyp?gf}S_ z)WiX8hAF`8v8mhnVz(gikBj^#t$)HrYYm2#bmf(<&zB*P$f`1_9C^qJFh<-CneC@5 zTMGBw`)lC0>^1P`;y`UjT=<&0iH0>>`$#Rm#r8&*sG2*ylOaaPV#(iwUgP6W_$Oz> z--ocKvEbhk+fOCA{{W9&S~%{kB~@T|rgiqT}0!8e9MPyo<{fGo(i6v4!G-G6n-_)wCg{#eUa{N7?2=`%V!-F z`G-4HdJ<_V-PqC#$mldrv-q|d3u!j8$GhkL%4d$@izxv_+r|btB~v|edFQOu%!zKF z@P^pKKHLXgNO{t3z&VOyRy6lMUP;L$X1u~#d_m$!Jfo&t=@20VvdGe0NWFmwqBJ1m zC!FzF*FO(+ElH!k){^5~iZi*OmVMFyq;R&t6(bqORr8FIRn*O^8>7c>s0hsZJD~)k zMQEYX+#Ya1$0yg1Ox5iN!a7HcZ%NhK9}n80Q2Sk`yBv~yw1fo&jGXQRJYd#r+P}m7 z7BZ2^C6&7F2qcfnah?K|QH{qO3=n;5jfQpArJ1}H4xw!Wlk*9Dm7A}~V?R3%`03uW zo!BR1?#~A4{{RZ~&kJdhX{9_*tE7@eacitc`T*FYfg35y5DeOMQDU?*a&5 zRg7(JAIx=DnTFi(n8j{r+NXgcNbGb^CeZE6!dSBxQJD_s1(*}jv)8{B&nRDWZ)%;D zk96_RiaZVDJLt6!hdv%28raxbV@T{^kg+9;MprVBI8anq3#jNi{E}gHIWHk|%yw#) zhiLx%fRYJ43KhVv3jY96@V1w#NqM7cHw%3LAs0>Absk9T<}nI#J=^F=tplxi($7vy z_yI^pGcdPjl;@^nW3Xen#ytg6M(rKHhJ0V(%@*}ziXS3VmN8xzGDtu;23IUg4!v@5 zUc>O)a%NdM*~kR47=VaVGvv5bo+u!ANFHq z-Z~C}Ro39O7TtK@{K7X;?I=*wu1(HPv zfwS(*Ynx9K_>#u=#0?lX7A&#pm-edUZS$a&R#?LnAQHPv0s%NV_zU5m z#=nT34Yyrd-pbxsHi;Whnf$dt8?STk@{U)rue6u;r|=)fKML7Dk91ul#8w7WWbPXvc5NxJ>%yz7@OpspEYcPM1cx@XFsLh|0-$==M9G+ivMbKZk+SkzY*w zZ1~^tXUDqC-Xiekt)|=A!UXfeLYtz)i2%B~BYsA9q<&alyyCuX(mpn7a$1cY^f~oG zb0FK~1(_#opOmrVAQD^W_4ER|X!Bd%B3f#WW5FK;ue?!kb-hY!Z9Y4kNf^?aOLn(r z#x@Aum@2k-Eyp3<^z+h^oMUt7R+>4k8|x`9JVkRShaOOjs|(3- zA;(O@oCE6Hl#ccG7KyI@$QC{ry3lpaE=>U~@Uug#381}Wu4njNZFOy<#_x&dXDN;TvK3MXOq@Fpyag!5SZ~n!s?%QhV+WB2ji*84 zD+^_5CxUClveXrU3S3#rv6&QO@2sv)P6v9(@#pPX@T%iS9t_faEZ=3fBYaoL%m^6- z806#y9HXh{*Vn-{e+?V`5ufWdbR*E_5D zw#!e33maJ0D=Tt9mrzEYWf;H#+a&?+s4-OvN$9mFn%@!F%kewK8q}8>KZRcNO3{wx zg}mpU+(ZY;QEpI88uj^Aii~uvPY8I&!@dw`b-fA3qp3?EfAj)2hmZKot_$ZpovFw0 ziuuo5@h69TO{qtx>Spup3gRf*4b#Kzn6tbo$R}=bjGEoi{uTI@@6DTPQO|0aJx2FV zGxBUQ1IO@?Pp%0S)Tl)-dDvdlvGirX?K|-=#2Wtqo#Sbt(RAY^R@ayJ2q5w!nTn*6 z-#+BMIPa1(jD3B7M)7q15P~ls_!m{O(n}`Ax4O5t znk*7H3c*|hfJsLVMn>Q>QBt2Ox4Cdu&%%fpuIu3xT+1;ak;J0jO)D>e%w z1X2*3{McL*w+9u(cyq&F3H7UWwD9fs{3IIW4kbx00$cfhS{9QrT&xZEQqx{Nz^%)0|#wj|I=#XEz|JVM3{vGK)9=VP`6=|L$hVomBa^PAB zV}*ej46n9MFhJ)Z5z?XYzkoakf2Znc@Y+o`;rEL58>p|Yb&s+~cLn>#0xI6ANZrFA z2+p`T{w#I*&&3yatMJ~(L;aMrJ1-6DI(tto{*`x`6H%6Vcavz(F57krTW$gL`9S9v z!k-c~F{WF|K3i><%)l0b)7LcyV0*O!|i|S9rUhP54ve zkBfW(;GeSo(hZ_`V^J3=Yh!T*+DB)0?176wr{+9o44(d(KHd#2*M$BlczOIk4cld+j z>n%lYGz;$w>Y8MWX?tNR32o$-Kp3VE5VC%!uEOO_;5cXpnK;wO)^{{V`*?xCpNzJVl54ZM&`b2K|Hl~^1si?PI! zFgfyMR*ASc$Qby?Scg>jGowwa-q_!unS_@2F7v~rLf|oTGX<7QAVx@e7~Aj^N`!%5 zzgmaD;rl=QKGf#dH9cG6rj4ZEEEZ312yIJG)8kbPlLwb^*xUHBnzIIl1t5WWuh z&r1>bH&M2l)<~Hye$-kiBaMb6f>^s8aLMx%oM6{wD(O88Cr--Nx<3{*T~EiluA8$> zw7t^q5XLS8A}C)t9fV6Fvogqj?2)mI_TA1r&s;wclSq@p{x;Q)wWhR;l3Go0CNN`y z6sgKE#{isXur>DwjivZK;7wBR;n#(tzRwejwlAUc1IwF{_4yu~3#az1`{ zhEcRA$-vGsPJJuOtb8wTt3sY4n^l70)aMei#^`%(;ZL_5gO9^{NxUfx@XO*IMs>MB zOjAbc#d4qpWC%;+)V@V}aD|nO8oATzpA_`V>u1xV)7}V*l!oTlB5Xm`rTCiaYe@0?$B3Zgfx0$Q20D_X-`2Gx@ofGcwhM1-1?*sNONkEFRcvkh z#Y&QXq#sK4e+=m#9d+=R-WIa7x3`@J*U9$@YWObHvm`n5cD^{=LW3iz|(UW0dasC+oque6JMn>UR3X6%^l zZsje$Giz~kg9%-%869zg4&If=Qya*wYn?RdyI$wTnkVdU`$zbO*vp~#dg9jQf-IKH zFj&h0!HKZ!!1@NSp#BQ}$Di<&>dW9iAKB`fCa-I3X1cSygoc#^1v0@ZowKU5D&-#_ zpP6zoUtIWa{t6%AJ6L4U{AHltXnMRmnCY`fl7?ImB#F`EhEb7`v~}y#72Rn+w}<>J z1-bBq{udgzhwq`56AzgswbzvEx!OCMLXQ-He{_W$bCt>S=a^P{Bz3}__alS;3HRWg zZpOz#)a+)JS$3!){LRCS>Su6s-~C$haF6Q_TI_pR#JGmVW5IBWP?Ovso)xgMOjDz%rhmAZP z;+;t~9dATYd1-5L6~3$GDAb^hiiQzP@w`EPQMp`|Vxx{L$D)_xpMbn&ZQ+|;e$(PN zhc2cqCB@;6*I3hFkd`vt#-3CY!hmfIimX5f4f7i0`qbW%HgL7vPXN&T3Gu(fI?G!4 z`deLNL5drTfvMb?AiqUcLvI|=nAgk_N56cE5tN0(NFRFE@XoWOd`KFWx$xy+ySufN z#L!z44EB>S3aN1ns4E72kiBGe){IG;L96L2gR4Dt97Zx2)Vc27y7v_t>Kb5q?0YT zelys&0=x@G_(SpewU78jel%%5ak+_>EmK2{j*&jqCBxg~wYxI&wqGzD;bbjesMo$F z@s5XgJ?Du0FK<4E@)m~cOY<%#SpIB0wlX5f;xo&b8fHRSZ+z1gX1PNAvsRuh{2b))shtDWv#tmY z6cIiKjz7K2H9tpgXRFSMLPW?7-oHU|1TXBJku~5-S3gvPLTvwiaXSwj7 zfP6<4zry-}C5vN#r{>2_EAn^Z$Lxvm zQ^tNQ`)9-b8X4fh<^pY^Mi?6v7z8ALc@9o_73|T*!LC^&f>>vJz~H_c>0S@9yMw~^ zzCE+Iwv#O$mvN;UD-X*TruYYV-JW#@BIaHxYmy2L;%Ufn$w|35XdxgWz??13zZ` zIQ@r)gQH6ajdjU2Eh(GqiRUtJ4uG?&fU-^HHYL`+3!p{tHw2H3eC;Cc;*!Ua=V1xp5fNRDq{u=8(DZ0DU{55*ohuFw3Y>H;Q zWhan6QoBd+yRxXS({C7l&U$x+yf=OS00>l#bzt&&q8m$xitBHyu0u+T(C)!KyNdX8 z#6BO_ycyyfJx<_DqeQH}UEJ3~2_qcf4e#b0a5#*E>62Pg#WtTKGp$Lcu6ZS{k*9b@ z>|1KORPx=cDx%&>skdtoHX;__?#m3Q=qvO);Md0Qg|>58=zkNhPJ;&TF5Z16$QZa@ z2Bt`PcJY;xF&e2DVsZ)ccJbDM;=LB*P56!*Q(1?U*Y*~OuDHksLhgn_4^6TB$G2+i zG~a>~$7Y(Z!^?B2+s1*~Thx|mB(RtyvyvHMTp!&tTe8lFZwG3|a7N>jE3dTir-yV~RFlX0?Ut6&hjoT| zJo~5v9sdBjZe;6|v_Z6k_W~qQaPh5J}oh22cXIvwFj5xj} zcoRo5Tg`5=+v+i)`+OVh)v(2}C8{`3rX~Z-ibUQ){m8}+Cbz7#-W-)tQ;KcsjoPPKV-xpiGF(hzTewp_+{D#;^*6XcRpr+l2(k{^INb-tes#5#nU zys;NUZ>1PoX#BinH<=)29XA+blNtM_x*<{zcV=ekw61P=r@>m+gviuwty1C?m-lRk z?J^W6B1AGKz~dM=PzxRoDV`korQ_i1Wp8fYAk(B9nm3Df2d}d!{C3o>^ynn<#)2ldmJt`)tz&TP8zT`- zIU$wQ{HO~1F$Xofu6#7Q@ur&(h`tSrm>egR_S%_}3GO8GH|{NV!PLZzLaH~GRCHs_ zduM_?C*Vy=SmW?@^5{BQY zB+q|xoa?@!b`UEw&jbU;K`!dPzQg^9y%jx-46d#(mM z4^O`(@TbE&gp0%ahM=0vbLKR*u}8U9$5ZCW3{A-x%eT{mUSFJaOC_DN zO7D&a9Y8%9Mr*I}SAo7K9da)YTv=VDD|XVuX#@9`LBzq?JIBykr-@$`a zOAm;)nu6XdsJOYbwrxBTM&ahdvI#%{f>dGo2N*ayP%nuff=?0rCVeAPQJCa9bQ>=M zj>pZB+_)VB74-z?70dXy;eUoByq90_y}iz_sVOQgZ*QK`ON0k*=am=Cn4)}&avNyn zjs`w?)IKtJgTdOL_G;Ma{vBBfXSBA8NYS>EyKZMfC6^>;FEIxIGM+_JN=x^DiAol_ zlRK?5$38yN@9s5-XPiTD9oPFwfQCs-gFaxDi{%Ctngu5~&t55b&%|C8@rCurl@VE{ zfq&98!6QbpWSLS}=ZPedR1AlWIVX^M*Om{9ntzWK;PFnEajQveyCJ{RAS;5b2G%RK z6<^{cGL8mEHOgxCemB0jO=H4Zy_A+#aG?`NZcW<-W4HG)5tqkVT;;kku9Zt&bbdt2 z%Tu?#w!VtiIdnFI#I_30iOlVAkT3C*0{KEV2PYm`IrpwM-{V}e*~y?ys107!5UUK! zv@iq!Ayi*A09;{P13gA^PnS20FD3DeUMH~C@2>S2nnN^qL|TtD@$(hWu?vTr9HNRB&FTOwk_I89AvYQ8AxuYCpqojx+6-<-bNK2-iNj8vv_|}(2k?v z*lx6o(4~!@u2rpHE%Q1!!y)bl(d&S7UOA|x*M@Z{<<##U0u)D^2>htoi92yLh0Xyz zkbP^2w!XT12(=vwDYZCmL+p|>CW#&JflY5sJvg0Nlreq^HqJjnjXIU_t~y%7G+ zK0dR)H=3W7;){4fv!oL}z3UI-NS09BPEI}WcJ&p`_+Q3a=8>V#{{RZQ#@5zHm3Nzz zd4VKlnN$=Q7{(orGr%>)YF`+BCml)K#rmD?y`wRYbP=eC7XA>3yAbEO2d*>HqVniC z!5)X=UkvDe3)H5!yVY&9>&ucQlF_6FbPPUU-6zbW8NraQeJh2yu=s4ZA+9Z^zPeUC z#uhYUdNW3n0pA$k$Q<)o7uui1ZyxEwD~o%LQpzmiTkt+#_M;m-+!ND0VzBgG7eVmb zOBTNO@l1`yk`1tyP%yrHqjY?AJ2Cjyu&DZzQWi9|%{RuH45>AqnQt|Ne8M|uTHkUL zpO`Ba8?bYZd*-?S0PL?3Urim}ho(ts=H-NMAy|+xBm&Z$kiGMkHQdedZtBVZ0Mc}5 z^t3%LVPh&|JTR5RIB#4a^{h#LA=-G+ypv|y<3fWE3J}o{NEx`uQb^<;htrCvxjVZV z7WAKl_d2{%Y4_e9Y6u(UnmI(*34xrt z$~eIO>ibrG&X?jXNmyw2Lf3xWLv}^Z6}prvGERDoj1gUqg|29tCa978K}+*kIl!ueTi|cd>YhrAMEcNYACT^$gLYiH<2L? zpWX(Em6)jG9D;h{v9A0tFNeInPvOsr?X^87d@aO!sTOl=;{rLt6_Ysmdy>ACr8M1* zJ(EoM-&XMMsc5_PX)bQKV-ZCrJ{v~*q!#*C?qwwX0@oSi6BJf+H5O*lrt8mE5$DT@MeR&*b zX&xZ(s%iHYG3oHY@3$AQ5jDV2Io#MQ)aRx>J(`bO5ajJ-dCUAW@kPQzu6P30@>3+7 zA~Gsq9z3Rz)JdKJ+!P!io71D$yeX{cjc;Y~67N~Ew=jV3rzCd&05cK_`Rc^T$Nh%q zfr2ZwxAE7CybG&tu-0unLM2V972w$*Rv;)6#&)UW?)5%{im9yr(VisM5`PZ(cSc*g zf<$wytP(bDfR$q7VNP%ZADgFYmgLrjrrxaQ^!+#CXNQr$v2P}|S3YA4(&FJyc?p$4 zQgfZT$R`BXkzQ)P6Vr_M{u|b3BH4+MIB^>8`{dlI2%$MAa6EUdPZIbu#y&B!noUaP z+ROr~QUGpU5xG@QJz6N#9+(e|02BdOx_5!J?+I#9>Q@$jZ6%gFd2f+!88gEO<$T5+ zvG-6%&NZ-; zcA?IG%QGk>k-$uK2h)RF$HqSqvET{3^MM49YEc~sW@dB8CLi0T%NOk;Hh=Ry!dH(cL|$ul~fr~ zB3e3`!Ter^rgX58wC&@*$;#Z&u_hgFNoCZMYpOpM~M@Mz*uq-d#_vTwQ;!is@X+WIUF`FI@D%`kGEETZ&V$ zp{;yG_zbb+5G-8okI|IuQ;xx$4P5}oeoL8gxg7fx(w(y$j7uwy$ z&6L9^x_zN*<|^BYzm}t5Q_kRh!y5AM6r>&!j@w%Be~F^D5I4(ZdmMc12O)N$QaSCl zj~OPn^ncnv;;xFe^XR&4Q%LN)=eM1b7Gxb~mG z)n}K-TE*4wt1^k?x702qS=g!CcNv(p?7uDtnhEMpYUY1v-v)RO;id3T4tR#*=H7VG zQpV;)W+9mG5P3r)H3Ord40r^dw7flYY`dWDFQPCuVF_*Q1B#lP=3=)9K%%_i( z;C0-1&sE~j+2+r{iZ$;6+-b9%W?!_#?{Vf6fE3FiX$N1MDtqFy<=0Z@Hn(TY%cJ;; z^TW3HS`r&z(wFn@Bo?TMN6UrBndWDsA@$;>vC?d;?Iy7C4e8a)s;FWrRaKY}aVzE9 z>c3v5o8rw!UbxbulK%ifhIfWWV|{Nd&R7GL47-^LQNY3iGtaG1(X?+8Et$I2e7z>t zMPjKs{qPP>6mmv5ECKohU5P$h+`>-o0TKx;ru6-4Y(6>Ft5H~zG>WZ zoxQlv9O7y|AMswDr*81vucAc4mQw>0JWH*0q?>6Y{A62M24 z(xj2u#v1{c#-(K`xa6rldChcD=|2)}tuE$~pHI_BD3WOCVL+-$7J$03^s z@~qtx;njhX4MX8>uJYa%X&xlRZ3_c{+sXNu;N?(Y4i6+&Q)#UZg&VWhv_Fm!*(`np z*8DANqeTmT{!4pwkz`=Ich2vbQ#){3fCJvVUsGrJiK@u{8t~o4&|J#g3#ieaQ*&qx_R@||Wgb;EI7CGyK(}MVebMU`K z)Dr8%TA8=Lft6xQV$zVHZNlsvVYwT(jPdlrAZHt_{4&?~kc5<}lLR`D8{i zGIpp^bDZ}ib6i=LuM{#qmo?SnIQhS_ z(pbW_dKqF^)DSw2!20qlx=)LKEck)pt1S!mh0_iFq)Z;mTUmnZdU!$x==_9`w*xuI zVz~!7&3xti7i$;4wEqCX)zqxKN&SQIbHF;1=(p>r#TCYttiq}$x4U~zAIy>m0opBq zNHesq+V(FN>Rt-dwOvENekSof$nh?dwXSU9v1nl3BCGC(YtJlC78@RHNKsBul0zuB zN-kWITBFldoOWl?_x}L2f%q5j?!l(i@BDA0#~bd|ty&n81vx5biDP+{kq8ZlhH!Zq zIO!~Z;H0`P^EDl%f{#6 z+GN8;9jP$vJh3w~w1aMNxL%)mt7VbWhVl}}Nvkw|llX`J3UT9~4^Q^b1K2gIJR5DU z?{7CgGm_tDffWwyWr21aah|pFkBWE8HqfEH ze-2&gu-dsv+GEOqMIyA4qTsUf2Gw2NM(<;u zEB4>P{{XPJ!~5b!v!lnT#;dWUw}~SPPgZ8mTZ4iN9^ZDpj__CPArHcz3hO$L!HqI+ z7HV!_p5I5my7D82H+E^`hU7&vMIsZGiNIM13bKV9Wmh<+m69$v4sC9JA!+{rv7f{r z5v&P$Xtvi=z_Z)jPY6gza9FPIFakzC`w-33AlJG4CjE*0Lt?kuN5nr5Uf9nG61J-d zf?ErS6|uc$m5V~dAT&~<=lB`dCckq&23hzM#-1FO=DEhfJR0M`d;6{o(OLYU3fFac4Ju4E(sSYsg-!zCm14Ab`PASf6Fo*2<$qJ zDw)KsqB>)Rg3+x{$zK8f#hx9vi8Q3pt?q5)+a!@(1XU%J^vPxgfahy-fC$Db-~Jo^ z#C{a;ifUdb_*-pubv23p&*AIaksXJgNO2@mG;Si__Xu3@4pe5o#`tsar{RBswHxR7 zYvN12W5brVOz~(92)uZnF}waH@=GE-QRgv;i@mxHzF(H32dVJi?Mv{x#9GXjJ{AK_ z)qJ5W`bF?aHC0kJq=pE?d9k?!M>J0)ZG*MW4m|pp$5GicqBJ>nS4YO)1JZO)h#n>X z0E8dn28F2TzAj679kdNXCsZLDY4WZEEEg#r1e?%|0=WPjA7}9w>`m~G#ab2ThdfE5 zMR#pE5?vXbnMVOe-KPu30B!(|-RtW;BgeXixYznuinRBTT*2kq-b4~Ri+JT2aUI4< z5@2zL$RrZnSAuG7@oU0*>>6i-b+~oU4N3|y)xKR*OO40Ri6m(a+7rnyB09IsjKhJ? zJY3@WY!aatsy{lkANVHLvGDK4&1>-cUb)nKF$bG0+8ykEY~hL@DPfrhn8*$@7|UlF zIP2uk`#O9s_@k}g_$~ZF{{RS{w`@$eIy|~lY1gR{=vdu?(}T3_c-7Q&CnGib4XOUv z9}E?4JUYH2I`*NY%jMnujELlo&l_C3BPIq55DqxTYsmf}e$w6o@vn=t$UJAES$LFa zmkBM7pqEBlxJV!gZRPUNlmPt7#X%Vj@6`-VOA$ITw!5Doc-Qtc_;KNDG6a`8uA-Z~ ziFHh}$2P)4vdGaEU8j*42e>A_au%LB_*-pjtN0r9>$-iyM%UUM`mNR0nR3T$$){>? z#TmS^L57RZw~B4K5`ubunCkk6#amnXyg<>f#194;joU|}ta@&(V-7$N$zV~I@*)d~ zB}vj`02GQ+a=xK9pWrq2+dK~{2Usqkch1(@Jh?LXKR3#N17D_EAAo#k<0FwOjaV@?AUjdD3h(tt#5xM0$XC zH_G@Ri<>zX6hrcpo$s6;0mgBu&rvHx=%Z+B7JjT__KNss^ZZS_jvp8JXVWreEO`iM(Xz05m;&Vs^{PuUlDvE;JXhJNe#SLo;L8-zjGzxK^q;mmsZfFymxG;E}>4)q_UMP zL5GH^8FI|kxXL`+AJrek{{V*n00KTHYBu`Dp*8N0;wxK)k}YpXxsJ|ATq~bVc zRzk}Af!w?fMR1=H{{U%iSHSwRU#EtfUhoZ!QSH_ATf?YbTU?#qaJkarg=LN`Dq=w_ zrJ1tGfMovwJN_~Fm+@D_+C8n`#cf*eQ3XrKac@792Eoj1K$3QdGM)ib$>SC1zYjlX zpNhU8zO~f6C#^o8Dnqo-8nw)hnJ*lEWW|dk0y0`WgUFvLiUj z6~{U$yFD4L-skC5{w>o!A9!lQ*W$JPm8|n-REJJQ0d1XvBgu*7NSurz3o9!*+JS{w z*1jotV&6qgH6mkWa!tL=7m~_^2gp$y5ahW*$Xu$N0tI|^qKS!prN@fo7f;gaRvSY?;Xb#(a&h;EyHNCf#xat}aj$hE(X z{{R^EK;9L#+FfM09yoT zMhDLo)bA7LCxW4veJdAH_|E#qj|H#6Yb}$lu$AG}8dlVka{PsmksE>m3UJEU;GVTp zPyLnC#9EET{{V*X=XAG-NHmLa5x5(^W982?B9$QE5D6XgU9FFY>@06!@TQkxZD*@T z9@wq*c`n}JW>9gvatIxm50s%QM{;tQUZ*vl?D-c>zxc)R4)*Fl72Mv%cNM`| z_?yFefz|hbt8WNc=Li>O zNtDJ0IU%+jjz&#bi}sAyd<%OM==MvZTt*TT_J)$>_X1Tyf>IVwP+|ZRpL+416HBG| ze(r5P!(F?zn|40OV1*sR9sqd`{i<`hg6+ZR)o#zn4dS^i(^%0Z)fU&~GC40Sau7pE z@y4jFg*jC$9mM${t$TQ!?!DI;Y~_?BU%ny zCS+MkZQYX+F8#U2OQr(;nEhIse&exfZq||Y@_y8R@QTUyp9x-G%XKhjE^boIB*)Ds z$o^v$&p2qoJY%h5+=7sxhu%> z-xYi=@js08!KVC9I%ctc7=Nbe-{{j4F_z9Ylb`MkXSQpP9{~Jeq+C7cgmitdLJ?%0 zAr}|1$iKqIV%&O#MJJ_1DATo`=1O*Y8vZ2FHC=K@^{)-;^7zwBytr;r{?O!%sFzX%@ znm-=3v|;0lOOs-Xjlj$it4Mca1I;n?>)O1;d`0n|mv3{VXwzGGb>NaAHLbi6$15_n z0v|nOmCA9KBya)8YUZa#AM)$*{;p{%jjc~i@YMeR5j-UmY8piUbTS;&ffT7qKkRERRjzTg-6YfxTR9pwDe}Nc3K|ibK>6}_+w4}-S~Z_+C`+e z5=W@%ESK`0GI!4~cLW}Gs&IOam7m}*+C$=P-PP1>X?@}yBKIvUTF~F8tfOvNO>)u4 z5&jZXrZR9d#e9dMd`j^zf;CM-0CEQAdu`GrStEt{(*u$zN`p(^j-C3HrrX^6GJlL8 z1UyZqoij+);qXSG9LTp;V){F+HQbg`EsfH7VY$I6xmUj!&nMy^7wI1jW1i0Q!tx^& z-Dj*`N;NBKpN8M$N995W00FbX=dFH!>c0iNQJ`vf_Zr3hvD59qiDzFjcJgo^0^Ts~ z<_?S(<>+z&AbR(}qi1Vjc)E7Ed3|C62DY9>w~I1l21|C`CKnhjxp~HOQ93Y$eb;kX z%gFalANHg1YeM!q<;{+%Wp^2lXcs0S7*aMyZQeJGGh;22jB=*A%|pZ&+Kefv{3Dj{ zNw~HiY!;WU^G1>mJiB>L1E@L1;gAU!#eG$xe#~(AjtH&(8eLgYVvD2DWDaGrdQKdU~KMu94ABbAcw{L0UO;B2WrOLqtq;e`Ta(NdMkf+^a>5Kth z%{T1vt>4?nr1;Ll%Jn0dNetKGOSzg;%7~&n2_zGNw4S-gTK&8DhvGlkhU3B7hll*C*q^!4!#joezC z68y`iN@KQdt(<9RWSE6jvWen>-Ix&GW=zeyg&~-pdh=gzd?ol>;eUx* zOr9V3n{WRB2_#UtVvlDmn-PIHOQx9~2S)q5U6nw|KArs1gl^7slX{;xL;E!Ni^sN~ zXYpmOpCs`NDbr(dcZ1It3~=4R4hDJvYPFaA6Wijqg`NCUr2IaA7irF-Mi$b^E!DdL zz5K{zcUg!S+#&?$fS_0E=fNL^9}T_&c=P@#-x|+#pxoO^lW2O|+f6)IN~&NM_c6*D z@-v90VUj`KI2F^+{{RIx@E?XX<@k5vEg!@l0=a3*#x<*PWp63lxlO$G4JDk#jydz> zC!NIftYM9BNG2&$P2Hard>Z|fz8QRH@O{T3(lb7Ok?W2r(1$t|0t!`M5tXVb1fEp_i5-uPpB{4DXA z@nwW<{{SAE+!A@hAlrb3^zDv*Fd0~a!@peJCstn;hBYFTebd2yA$&5M!8(ymerx-y zbl$G95o-#vaLXp$>`2c$wkhZ7TAvDTyc1=7+H?rEse$r^VI+7Yt1_IIQcejifKKCH zZ}Cs!4~u+3t4MWSRdj`vyO^wCmRz%w_oQ~&7{g;~@Du^gGHYnMg{Our{@>ws`&_7B z{VL!I#v~YGdLRM6hXHfz*1bqVPB*>GWfh}7z_R!sp!kj_@jrvLPxwc5rcLq5Z*c7f z?&jeYUOXOz6P|wPuIpLwb>5xei|G|zg+HBdA3ipn!*ut4V!FGTVY<7!`(T)EhUpom znQo?=>F%672jk#4-sk-b?$3Qc&$qAdbxFK4(|6MIjd;%&&(hg;{UI^wPehlsbqj{< z?-57XC%6_Y-N*WPFpJv}pZAMGn!@^<&CfrOZ7_KP=y`;v<%7|7%exf!Y6#V!b}=?j zH0$7cr0mS`X+bj93SM4bYM3qqOWc88dsSJ{xU?I@_i@Roo5l!rk1++^#*OWD{_**t z4id6^N#-rlD1h^pN?Y@vt`bNatWa-MA|PgW$CAA_a;IvqD|?T?g6WyI18l!B4Qs}# zM=Y6l*I!ffg|V6X3`|6K)k$aQ+a<{X-kl4XV|~cKo@6iE@|puR>!iOLxRehG2Ds8= zjODLsvHz{A7bW~%*BbS{UA`4RL@6E0WpK^5zMhsQGF zoXl{)cb);7ZByr6}Q>CPhH%Tk{gF2XoaZmx#>(XeBd;l&Qt5IIza=yvkB5y=9=8=|pJS}t0Cb%zFwue5PpVXT ze@v`L8#b$iHMqgqZHoQSx{F%TC0+1eN`RW}hZfmzC`n zW(9IMNL;Gqa1j^h#(2-01`n*&F=rR8l?0OyaT&_@bhGR^5_%5wI@1&`S(lq}jbUfu6 z9wSUcVi@zi;b|MQwx#+#6)ax^4c=KgvO;m(IEuB2o766?(#9I;Fh$sim?-?zKHM1HQwzgp zY>8hy(=_F5uf~Oq!oQt*Q#rlt&N8X zfk2!Q@v7f0^li)79i=UWERhGV1rUHJwrju(NdF5+8=|&f5LDpkqTxeDe!BSYLcrw% zKSD}okJ9)N5lbSY~$=0OJ92H|q*o=332xspG+ zNa$NNBqH%J9*<#{)$Y`M4(*PJyM9(QgL!aFYJpizKCs1FKOD|4jDbu&N!wS$l0(#_ zFVvP!1QNGlb^^1^ChF3S5vJ{*$aOMh8FM>dwP8wKa!H8?>ghQ_S$6NEYkH>k6EVE~ z<=n8?A;CYk**%8@Y+lnZ*WCW2m$gwhLn;**$?;I8e3 zmQ%%%1|pk6Ou}qjj~9i>|5T|@%p=KhVY6B;{UpHou?|jKdE+mr6pX?I;=Sbk0Q;MH zsECG~dNxDf$WLWCGU(>GEZvN86h`eJ^KFx?0o*mf;P#tA4Ey>++^O>im7u#9kcdjV zwMGxJ4t_x(io5=Q!k|3wsvqe5@7C7FR~*ttKRjd(OQ_si(Epq-)E((|!*a7KDDm0Y z8Hu6~Wfm^+#+XSP5C$DelaY={vANC(Vr196IUsh9GpbSe*gWkEfO)F9bl7}Qa z)tqWhu(P-NLJ89N8~6uDf4`CRYNOWoT>r&ekndji9gtB;f9_@{)|hB&Fy_ozH!5E% ziAZ0uqFyZiV5l@D>ddqpzt=YJ`~qq*R@QcPqj@bxWXm zhdKMoN;{3##s{BrceSOK44w9kyE}It`FKZe|J*d@dUieexqm~w^Ff|`%EXN|b-TEF zU>g5I^TZ9G%+mE7(fInBQtDTQVVU7#r08J8#**ic4tqbXTtnU!KGhE|prOagb+9y) z(jFReB($F-sCgKT)r9mO?PEvSy!%P8)?gtz{PlOnz~s+YrgH}kZEZpRi`KSdcGy_R z51&b?`owQc1)s}`%~q)J1`yQkODs)1jQ8w`F6basYs)_U9gj9_JJP+@TlEOUyG5`_ zV#t6>>&ACk{>fDKLF(s>*RKWDgHy95x%w*286oKI%C$rVmBB6%&-kh2Pdg&Kl665 z`Mjq2Z}cBi+uxiDmZ2IHMr&1BS^^~EX`x4=QMf~p#89Q*(#Xp!oEq3bk{m9?tn;}d zvLBsOvG)z3Z<^<;`qnfNOh_P*lWRPg@Xda4GwVK1!Ox~)@W3i( z3E1BC+eKNtEH3){Nc7y(h%CLftZCC^uPebN(6VazkmTz&;q9zfA*O9pZRJ&=mF{{x z+*;VUJT$=uEr;z7EUTA{_m#9Y#Fk~db7yT;dlc@0c6?s-EJHZqg#Fh?>Mb)?<@{X+ zPWDG@mhSG-UlFIgWHVw>;#_Pe0K%-0RDu!arCFa}Lq~icSxccF&YzMdAw`+`_3>&( z1|?WtGP0S55|%2#0w#lMBgmHOtnE3wBx{H}T04;cqeD;Ue*QDj-| z$l$0C5J(4=0$>2nTlE<7b%~mTzW}{a_$d?mV5Ee$5_U{wP(T7|h;z;(&()oD)|~;w zH_`)&Cf%{@&F1s!eoz2|Ce`UCw5~lmAg{vIP!5HJaz}5z18oGD`jL~wFhf{k#5B|V zBR;6!`x3(vnB94jX%m@78A6tvt*u|l2h;6btz`{t;dCHoX6c@EN+&Mu-+#%Hk=j1P z8zp_cjuv()RL6^-T*W5TS&)HV^)z}Y@fMB_j4pGHBq{x=T;4b3mJIscsZyr3Nk`h` z7$+2zBm|-m-4b^KvVNkl&fUIJ_y?5>_P!1mL9K?2?**tJ^aED(B!$;FH^s%?*0n~) zQ1w{;88nCEJLB$pNXEacd`1!M6eB8;E3eB}7UUQ|CXxiWjJY|;h$?pLDE(hRdra|8 zG;e6l&`h7P^aUXV@iy3M_hPZO z%ti!-oR`(_@#p3YNZ)P}^y@MeXrnP&+o7c7%&%_m ztxS0~hm`$<8v^sOWH@hq8#N`YB_|>`o8bmfruJ_kQTxT^@umY=i9u+Ug#y|cE&~3q z4h)7hR>P`vp+TqpH~o$`>Q?2KUpY!Rm=9XopJe$ONUhAHEYhC*-e3#`)J2D&z_y3B zIC>#7F|u0Hs>u?~=U z4CCB@WQ&LWgNrfG@jysZA>k*1Bhg~gd%@(#BLceWp%#=?EfyYEuuWXgq^ z4ic)Y;U|j^SG(#ca7;*H@-bBXGpFk<8%(8oi8x(M#O3yyn^nppGDSk|qFDIQu$<-M z^*$BvVEm3Sn8s!I`GXx-bt`Z84SxQ>xvkQY;6}TpSV@SG@wG7PoMw=t!wlO~^)vqB zC4&1T@MNB|mKiFMlT%$aEp#H zOYet|xi`0>vF0Sn#pCa_|y`8^35Jtj9@opuibfOp<9c35Gb{1O}8JaD-dtPUX8AjJyF zs!;w5L|a0LKL@f3QVPV1uqz>0SvCz#h#ztv+v}YC#-BtpHG~9arPkCET*I!V%D7A@ z&k8^LTv@IUFSg9wA?zFSnb#_W{1Y~6seZIUaQyWY4eiJCV#mQAP5nT>ewYpjNL5+u zAyW<5?%)AbMgc5CMs8^S9fv-3z?z#a%*zOaQ5Qj{eBr|B%AHoh(Z*zp?XKjmP3@Km1+BqFrmUp@mz?;<5H9u#-HZVl=*&n99}`{E*2Ai33QHVkap*U# zatN6g`s7B9jyH#`6)f^gRjiNsm$5nSXNwDb0yOOIKErUJNh1pLd(bZz;$3AyCWaYt zywNi=zb1Pu&=^wt#O~@cd?X{c-a%B>{G;W&e1^zZl{OpWhs z8+niz>t?QsQ<98Yl=kC}{?V-J?Q@jEGv`i-YZxrRcU@6JeO{mOD$p`-;k;%Jr`z~& z@LHw6<`J~U&1Llmam$~MVfwFVTUBpFC-afi6t<+U%F18^dwSvMy-={gu}X$m1^t4x z67zUMCibx>bahy~!AWVak~5M++1qmMBWz+fD_qQN2dguc>0gt81B1+6>T5n*c?$jE zktC5<-O^~OW{zOgh(1FH1Gt2DJdWj&fsMY8PWFD9R#*!oW-5cwt&q5H`j8jfE)YKv z3=3KYv;QPnr?pcH#l13R!bX7Ne=|Lklip)^#&BI(0ZAMc|1wX>}|u!L#X)fTDJw*>qxIW zg+tzx^5l8b#oeP3jfnBNjq$t5I(qpYDK1M;!;&FC?)5U=upBEKECQ+5Ot{YF?2tuE zls#)lzq^aoe7>V{niBJX?wdd^YQ_M*Z>thn&P#$v1H+5q4bW6Z64js`=s$+nf6e}L zS_4==yEU3D+qCPu z%jxOY-4gSGpC<)&2nlxZ>!Zj2-cYyG?!-m7{Sy9LbRRqt+d~dBGfkrMGF)u#gdJ(T z)vX_`IKst!&v?D^naA^WIBL+*uV%^h+`a~zL#gH?YhQJ2=9B<{rSg_hbPILP+=Pk+QmpRbKnUks<-#U3WKV}hH^N%#FZ z;%;Y9;G&S*7LPT5wti3bN=0|8j{2q2+zWww{DjFw7cx29PeQs3XysEZGBnqB2VTm~ z-Vze&d&xuPmtD$`xz=%YsB?X>MbX9a~IY%GP$@8u*F9kjks7_ z3jKFrV&rRuxw4wzc$+5PkNYe05 zMdL}RwF2f~17XQ-XPlEBvQC)#4wc?8;hC2IH4+=URsZZ| zjyD4iBT!1(<5#z%d7XhK;o{_U9H4fs(D!OXr7*>+~mDfSxZo=vS7jR*NJ z@%zcs2;b1O35$D{CA%rHq6; zs>1&16SBIdb(ElZ{VnMh2MfGT#{**T#IH2pb{%r^w(=u_$g#psN>>+yTkuNx-E~h#x(K)9 zruw8h)tY~V4)kN7>SY!?@%)AU`}4Al_mf4^H>O(~sI zyQ{Ef0TEZfh7YABamElsCS(EOF>9e2+S|INDGh`$?ESFY8(5H9^sph{I%8ar7NwpB zc4bt~wk6LfmMeS3uX1*xKd_%2{!a?a)@}8fUgZX(C8m7Sxt@Yt76gOS5t6L~>_4+37`NX76|omw-d^!| zZZJ_%j$-sp)K{3{Az)1P^8j>V;|Yorl9PM*!JnWDYhNC-1_Jo*|5&C^Cf&^wPGRAF zE$yF3MSfdZcYv}+{&*W*>gjYeQpwnX*FZp=-j!@QC7d= zU5Iu*>dO7J1Sod(zL5_01JGs`r@B~(RF{&gf&hvll(!kQx9gqeWS49TcyIjchI^P* z6$UhW+w)Wpx{7{UlK7IAeB9xplG)+>2sO=+>|1{Ukr_B zFUq28+x)EN{&ue;;E;;>7Oot4-Yhovw0|`KAcBnh*MP8V2&ra|_w)QU#pu%JGF3^~ zjp~(n5H(Gbvj^9**e{=dGTr`^t2JATWE9#cmBLytCZj>jCE~(ZVG)+1#Q&FM`*l-Y zB)Ls=T=Vt?-dm3I3-n-od^slB7TXPK=dTnEsjNG1jD+fDIU_nk`wC+M#r#ZB3jW)Z zO5M0hvg-*q$KsiK8!?GpM%s^k@J8Gt!+r#2Q|VIY4?lxhLC%J9eCtt*oLrYv4%g&R zWvw?ANWqa=J$lEC*w`MX*){O@kb~BG*wGfDn~(G9^xf_k9-f_4iGNZ&3TlgER8_Ym zgN2QD5pe1FPXDsG=F+vhz_<;ihGJ}y>R5`9JNB?1`fmppD;o<-5}4f(!GV0Bb#wQ* zU|iINsH^Wz*=!<1(B;NAX6;i0_J1awUwvqAjGLXjL>ql@_#_3yWS@ND(l8Tn4>5!~ z!orTJ60v((nt8)rE_wQgo%>tg!7IG^W{p3Rj#ulgd$|+O>3fBO((Ri#LtT)5=K`Lx zdld{FnJ+cdX-%MS{N6-c&p}EWAFB6SB z8j=jS;r`{s8+djBbMv)jPP<4@GR{(hy@(ltT&Yh%RUwU_Hy5CyrmIMhVO^S-qn z;xckxwig^IT-PP+A=!^o5hU6@!d=7&!Q@_F2{QzE(&nKQztrQ zq+d!bqDfKogXhmo0eHGRY(jeBcy_1oyw^_pspYdWq0xsScv%#8i*(9HCkb_1ZKY=} zb6~gH+LC%is4aO^R%LCc7b}D8+gwhf@B*&yoLLcxwb~*2wSL3D{3ix`)T(~^=W0Tp z7{hdjA2pcs=~z%0!hBQzc9M}jZJvd}l=o>?d)%lB-}nzxJhM7|wLfleiV~_`jd0WD zcIEl0Kypjn{O`$z|h#9cQ7lt+)q$v1~q%*%XWhJAU)krM1yG+V%GwPu?OL7ihA*Pu>h;k zx8}CJ`7W(p(F*X-R{b=@rh%tYhH>G$G#*?ji==;#oe%|1kpF;k3-qmsF~lJ!ykuVd zFG-p@b~%-FxcJ=}u^Q0^8`4V7aL03AuT3%zT(${-LaI~aA;Er9#E_rjyg#&QZI89a z{wrRMD(gm(I6A3BYV}*Fnwo8^lhVQW4_vP8Ia|^}{moyMV94?S$*bwfhd1jA>)9k} z^+sfUm{kP~6+_l}3u(ct_`6edujs2ZPLp$KJz#&%ms^Q_))(BrqU|i;$Y>lps8mAkp*CdN$H&`GG$RsK9aAxq@QTu%1~!s zhXQ=aQc>kqe{RM4xqAd$=)E6-kXV8h*#ajnM#>A#7=!gdw=H__Ni{oG`FcD&o`Bn|`z{(J zpyP?+ph6++K=si^zKZvztgoUBOmCFiOAJ$P(#JP2(`{CM@ABsdnb>mM$^pRV^KDn# zb4Z>4Rxh@P%wf3$)6_rw6wvK#xglOckDQ#B+ZM3?TwJ6_*5@-L<~8M?S^Q+!dzbDQ zg4aUmiW^x=szNxz>{o2G9=8?B!)ztw@1Q>)2_2A4JMq3V&cGmId>_xXn zGTGRv4b{J(_fdPoqx@B6lFdb5e-zY9_(NI3LTLRN?nBVSFD@uN8UJG!T5mF+aJWmW zf3rP=FZ!DzMK9VIP6yp8lU(ca^L_;5(R~cSqgMg|3Nz*tuF+%^R4O+Mevh7^v1V_a zUCQ*u-a(7$23HkYJo4HNq>()4?-TGAbcv#e@rX}|C%5>Ul2nX}hq}uNnoE-DP~P z`)y&!HnhQIVe_OXUZxIJaA|03Jo6K0Y*|NJ+?%0d){}&6$_I}~HT3ZeQ*o!N%rX<~ zJihKr%hTTVHGTkb<+3fk+qVCe-D5q%!~57_1xa)d`JN^05jSp46?1Q#^Wt09r{KZZ zD>~hLgWh<2R%SFvf?b)Iu3ARrRE+||e3SCOT-kQ9)&2qJ60ga0n`BGOV#_9y-JbZ$ z{$%Q!@{>vuXNZAwkPgSFPFpG}s@Lf23MwZx?SEDW!5D|FP=Q$MefhYD=joT#?iJdFn^oJX2-oiHjQ+KCsAkatbKAj29!%Y}HS2~^T0-}aM02%(H zd5~^itKP=6Ku;O+7WsmYncyd=Z=r2kC30-_L56j_`{|_-MU=?9*IRW)VAhT~M)ISu zvP5^Ii+}<>huUuiessb8+?})P(x#22U}lkX)k-*jXIp(M0qSbS@E?Q8_%#-v8|e|U zkL9-)Mr%>2-~DC$Yr5dM^Z{cL#tbAfH!mq&QyL*Igm+>~wgtzwek zpuL!QdzT4ghl=Y%+COV414T$ZWu8am%v}4F`VfG--pJg->db#_?#QI2ZeESV*}m{W z0j&rv0UGxdG(bXw-E4wO+VtNNVrw%d_VT^NA0p@g8Ow&Od7)ZQ#(}8_#Mn@Nat#q# z3%^HW9=^93rHjJ&bWUSVv*9X=?O`xnsjf}vs3LppqVk)3t1~m+biu9Z-wQ~`$`#u8 zPMoxVKOSQ*hRvKQ74N6Ty#h2yYxuwx$Wy292BGG6#AxXqic#?5U z;d47o2@R|y6=DPl{5o6bKgl$d)63MbH^coBh_jX^$2EYu`ngZC%bbFAnrmFW)+wbO z&?4ORVBew-SBYVA9`mr5{g2hXGfT8@hVN67f=-9F^%5#BMM`i+Z*O)*$(ngfJ~4`eV7(vVz( z3KqY@`#EYjtqK;>^&!h~TFQ`rhb{FKGEAwOp(6V0j67ZRy!C24)r{cV-<9Oiu@Vf9 zjSm{et1!>yuXfrh@v$l%kqPkfvhq=u-ts!-S@t@xcp}TE+5N_tdiJ|i?COwGqLyb2 z7NN78vy%B9sxx}^3+rSpo(hxtTBaEtda2RWHi_z2_vJcT3TY9}P%4_;qO&t zm3@Oz$_xpwNK#5kUXvX}z|VZTnD)+i){>)LYkWb{;h&Hl+ebkOJjpVvN8Qo=G!0oJ zIKFqsT9Ed*`49g0(q0uCvcMpmQg$)Q3_zhYO1SE;Pw)@@`4z5*?7)yp%3Hycb(Q) zx%zR!!;MDDA6CLFwEL{QbM@;xfG+P}Wb3=)fv z_1|G6I|}*5@_1j)Z-x2#BM$5>&@0gtHn6ZcKMgQYu^onzotk0Y=9?bf>@2KZ=0PVl zHC(8w?+9sWiKDjmRA=iv^1mc~8=X`rS73dlVKxbwGJO77f!q<2OcF=Dcq5+I(Bvnd z#4|?7WywAKwg*jz|C9%8`hSLTtM85?9LG@y3Ccu6-&$}mduewDORhBTW$6K3Z(YB^ z_`=wnL7O@s!G`&(UjyfvQx;xg4r3?>Hf!4Z-wtRCqH8z@%+SFct!)Oy9R-K$1je!f ze5uBr;ez2d%*F-B1$jQT$r{Z+%O*uKTFNR+r}$g)iF)$!hcBeDu>&@!r0ABz3Xl$N ziSnxr!Q$-;rYb>I6BS=VeNqSMJ|DJX_lRU}aGyFYNw7FjziM2(L24*WftAWGZ|py*?lwq^FLHI{dQN3Q3Uq!T*qJSaXG(q_$4V=FYkGCMjfI)pft~Kq0134F z;Csh-d&%ZFpgiA0xlPr$<@gwZe<)4(tw_%u3kCVY87U9tVxE6x`z-X0&Ae-K{1g(q z_e0i9=FzfE*Ld{vv7~W+wy<#NNG}_)ES9< zRzlq+GX}MuQ;|YExcXc(EUO=YVto8nyZg+U^(~F_gahVtR7cUTT`{FLO^pA7MUAqLFgpXdgS%5 zT#w;7k3UCUjUzN@A#W9t=lsLBhMGkQm>Rm$wnUTZ{!n}NnJD0@_4C3F>f9hB@KiU9 zV2&5i?FYSNu=&$k^s`6q3rLDTgUM?)Z_V!LK0buWziu~(1{0RU=AuB+zFOncO_vPt zXE6?mXAT^g@32+ZqEE{#CoAu5$8;=bd+c)?9ju4p29>y(57HB zUqLDZUuwgnjXK!NtO$VuDFo!(E6%(0to%@O%SMnn{2AKtJ3f*IB-1OT{9unOx)Z$^ zwPi~~of+tTAUb*h)y4GY2h%}|5lVtFq37h1x1P(fCY#JnqUBfPejzB3%#45%hmMvE zI|tW4qscU1saip=N!A+)+|U@VUY~7sQvgG8oE|E$>g?A1-r~2Mx!Khk&IJ{h@^w+F zF-F|BObKZr<2ND)x{`g#4@6?3FCf!jFQAjJbsu3xz>WWgZY2W`8n{>4iA!WdrEzn3 z7hel!f~h-|JFfNALtlT0I~i{E{Zu^Rfcl3a=W>mwWUMe#FNq*vOG81`P9GM%dz{Dz zC>w?pW6yt>Ji1oi%^^ylb4%mXHyPha!@At3kPqsO>^;~LlScrOACEX8nurO>Yo?Q4 zqvU5oCV8?>UNhUA_PgPlT`HB40;*2Pb}W6@vY2RKsr;;@4cRD)R}Vw0`}k+0+dgPL zgv12qL$+k3uX2UHRub8>(9m_~P?(RG8JEr?#%r;>7`A`Oi4{7A*YE&Q5B~e0%9omdTYn=F4Jnj{{!Ao?DJ(6i{%5e4jjiH9K@YMbo^$bW&LbwU zYYn`oP{)H=$s4%{8ix_3GPGjezG@HHI=) zZi&5s(!$C^mjw|T%SPc&TA4bA;PM-Fl4MF4rxuB=KUk;OoZ6)1|9jpRt4x!7EohO(hvq%ZRKjOFdBQT&s#mZdI*!uo@zBCwM0J;88_679% z1@tE|L<_D2oK{v9y1dwXo!K;y<0Vk*lS4Cz_x5k5pdn}6o71J7CJI}XZF;{|3L-4d zOhf172U{ivdr7)AN2jY*yZ*JzxU2$DJ$dn{ARq4J{cKZ&yM3`Iy+sn}op`+8ULisj z2;`S6H$;ypPZEGh?4mt;Hot&oZJYDj`C}veHsDiGw5ig*FAE%*rB%Qzo*LrN8>m?6_(VjgzfHUJwSoJaze%#uPwQl2jkJcQjqG@-l z^5>H^l^<9hnYO3e1b?pe6jU%#+`KULApXhjL;P#rIm2{zCS|#_)MoS{u8N0t6mg}s zl4|;F@d9GdEq(!+S-pVdO8s#G;xPvA!su(qTYCwopC04IDzYnJerVNTN1n#t0+v>z zaveEpmnQnP9feQ;(BDql=aN(duUvcq{k~wjC^Uq`v@X+k1}9}D$<3zWzRrF%mT}5E zc9w3c1>`NT7f=NJuC5l21|Z=D3xQ0W2JzWRr2fwHo=>nyP<)zD1#;M3+GvcyzA}cl z>dKF$|Fb{|{=0g1glT-RUc{W%+qjl1Njok0Nebc|!k+;EF9`oe=rkqHDJ@-UJ4Lhl zp{~7AAR!$a-cv(W^*;)G0WnShP5WQj2$`OCm<@0_qTc7Ax@*JZ^qw;1eN$nW1C)0_ z1l%zk&7fNR%j@o)fYAe6>NVm}Hp5IN&?D};n`MV&h=zD;qQ)UmyCpT^i2d@=S{oC| zjeFEw`l0df?|*f(Zls~3vB57O<66m5Uj)(%D3#|7(Paeq62mzEsC1cnC}Sv6myclN zd%i+4<2A?3O6k-b5(nQ`JV>N~o(rpN#OT=X>Kc|ilI~d-4!A|pia=}Gu-KUA({VJ* z;kq+u6m2gqSbD8Fv6;Q2SR)5aviH1qPfarBA057)v^ig}fsko89RE;OFFu&mK za}HlGjg-;unQ7~4ygUyn@uqpIK0bv~C(m@(F^b2`B#+Z_>2q=uhoBlK&^yn3Hx?)C zN<;8nqtn9|zc``*^SDgQOaU<(dW{7&dyWLE>sg`|#1hRl(+J`UG0z|Q@ne%r6$+Vs zNWj?(2y_AGbqT8cTfmlUI(Pe<0Lw(=@6@0&Yo-r3Z)ALsCus|J-YYey!x;j^QYhr7 zbI~mLbK)&<98`cePuel*) z*3_Q?i~Hl$irii782C+lj-&XpF1iq+0Vp%fS1*~e&7L2$=D*IJBFj?g!0!7bB2Pb_ zzjw^l7IelS8mYH@<;YeIwZ$RF(3xT_3h+KZJ0E{uQh6iDG? zmD}S^;plP&jONM=cxu2fib_c*De2pbioDszO1=J?gx=LGi?LL395!!GXYtP~Gb*aY z?iR?$9OfAOD)002?2JnaQw^B9x)9w>jx`ld!z1~dxlCsM0&+_<+-#MQdB`krv1~)i zR+^yB*fS&UT|P1U_qQ^XuMYI5_rN3cN9f7;-=gSfR^NqaUCs3Pm1jF@6jl5r<-g4}glGIE*hMDbP61_T#SYoMkuagvtGLfxbsqf~Vtqtod0 zdI*1aAS`6-S#o^1kA!&r0!o*Z{I?i-i&9gF8af&oiZ}>;B-ASm8XL+{k&C0|f|P}a za5}y@z56ravPVjm?-PpG&^(()myI&1A!nY>D$sgG_U#&dM=u9%1gqP-`qGi>s7pbuJ_GzuQKtNeSk$zVj#g}v@rzFr^FDqn%<{Amg`TS)c_;vA@mjw#1 zIX407<Jt}NO^3z0M&?~0Qn0#y413}m)sD72x#b*WOMZi|I8&Z0 zWFPUr?j#g&o3ISvh4qq)k?l_j_WNbdI+4pV_NaCK>^xWf-Y~-YN~4g{bv%6Y^HERr z{~ZA13+NNz;7XSfaY@}bJ7w^gN4saA^auU`SjDwT@dtO~>W4yw1+eHYtn5#1mvqT! z{l@3v=?iPoJ6e}x7m{}K#cES}~E(qgXNixz`5(Nl~3cxo7AcD$H(>!V6_dBRG?p?A%IRU<7K8Saij&&A(D4*r18PQC zvtP*>zM+FI&^K(oP6sc0@}-6I&b@Hi*#rIxQWg*&(LQgdHP%43#?|Ai> z=9=haKD_>e>ZLKD+Vre@bo^v%l#ZA6E;5#$F-a3cULS(xYU&qhQBZ6~z0o{jfRWmhnCUk@4@ zxi{a#0Sgb3wqaCpaVx7{x|^4Kz@99f+i2TkSIkM&-%cQ=gMqX zPPeTxl!P`;30X+T=y(RsQV)B{LidNa5$!BDj*Czlc$&E28IwvWDg52`nlNMbRVW6~ z2DzdDi!PQzQL8D1LN`4<+}>VQE}z=DK*W! zbC+PXqp4hLU{bz)lHHR~KUTc$D<*KGMX?TAc_05n9Sc07I`r%$l%;lLAoQBD2gr0C zR8GGhUhx@MwpB5l_fmj0p~s1@js4S9@@<#oQ|bL{k24o}JgN2LEtFV_q|j>Afy5-L zRpOmRaHXX&kHWA4x8obZjWk8pFa>20VM2OR^EiwGp?rA%-1Gukbvj7~y@1BvErg!c zm!}`j((wlyoEKt;+NiHOsbdA8(0A}$^wTyDOk2Y1`9 zs{?YE%Q2hWn#)tq%;%+| zNn4uVaGFQ&M@4y)6r3KjU^BpZ)FX%v6QJ%>=%wwE2ptocliUh@MseIV{h<3^doFQi z{Gbzx&EpIWsq^qT6j3C~DD6@p($BIi-q=W|#q51gf+_V>9Jkb=jx^$R2*XoJmg~;@ zm4;xmH*`M;dun9)R@puFo29izV>xtx?-{SPu~rCNpnqO}VPpNd1nA{& zQ)=So`Uz2+YyAe>Rg6yAIZD1I88A#Ua@rzpQW5L&G*)C!n+WHstsEEpx*JB&f{0Y$ z3fgo-$Q5D(DNyzHBTtv$-aS0PFjx^0&RP$L|F2}pQYXdW3$IA7DSk-$qvl}*3_-rp`32Xu%0-l zh6EA7Vb)CK?}|aUMMg8l&fS0U5d8;&ar`a{|tqn8lwGpS9doZE9Qp~$=ut+z#VlxCp(BFtau(| zBLg9&=Wn_@ZNMd=csmkVyH<8!?piHWdA#Dh4m=HYl6+tr11hEY2h7tSHATY*&t)99f^aJ6+L=zhlZ)tw zjzZ0L+XPuZP6~fyw{-X24$sy7Kl^9fuMktcq!L0O5$~V_E9v%h^$sMBaFa5_w;2^j|nWJ7Q{7VioPE7u3le^teA zR(o7aGoD^i5Pn`^M{~wR#L*z6TP35f>xPoov=CyPl4ni5Q%;$r*<+oJ%JNKB4=DQeM$Qkl-~k`6UpxP8z!>r>kc7mN*pBe<;PT7o{ED5=hlt;!=6RP% z%6M!UHtf9+0&(=ZHVIB(Yp8VOISB8WeDUG}V52nSJyq{NXs_AdaN`UI^4@O$ zrlx=e?{lL5tf}Nfpf8qvOegABDZB0!Wt}<9upD6IT+X?rA@E=BBhK`7a4MhLbFyC( zqH1b?+hn0Ia4a_6$R+etiRzD4(7f{)4Nu#Yd|s8z2$<>Td_X&;`S(PFh*~^SBXJS@ zA!>tSwE|b)-;N+p%JTZ|uogVkv>tb?5nz>)l%PTS&eWq7C3hj`VFO073vguh24HaB tP=2vWpJ7Z|{u2mz!BCNwEZlMV@8kt$Un(go=%hR_5NLj>uB-Vsp{ zN$4PiB26TL&{3)=`d)tj+t#|D?!C{P`7~$FI(z2Koc%m||2OmB65yP%zL7qFfdK$u zI6DCU%>ncPKt{&@X=lN7)|gqDnVFcF*+C!{R!(+KP7Zbs4lZuKb6niK+#DR|1kdsE z3xL63PM-4+K>-M#09fFE24Mi6-NVGp#>~tnz{SBO@c%jf>jLnyGLA6=fD94>Otun$H0k8Gt}WCgA_f`mA^K*>wODFEgLG zq7DncdOz9vTJZW8@u3nh|q;gm!+;q%Un}Yy^cVt>D|1g zZ(wL-Y;A)=+uGSXxVd|HdfoT-!8{BJ4GWKmOh`;he)9BLN@muJ?3|amukzj&6_=Ej zl~+_YG&VK2w6?Xs`_%KfmqhOC9~hsYOioSDe4YKiOr@=?(%05Ee(vob93CD2I{E!S zxERiG{-5!`1N(pA;yuH~$ixI>0{ssz2F9?n6UfWNEUw7Hr(+4a7sM~26vrx{n^9Qb z%_gaA^#kl0{DoZ*cJ2G6pZ^2xeo7qb5g*#C`d7QhK)ID2_OUce2&ufHzx z8zAKNG&q3O-qlkByS+Jl8{E+lu-6qZe@LvGD?F)>Z-Z!V_8W{{iZ*`6!3U`s<3?&w zh<TY=CY^h zYgye50f8>2szScn`j(oPMzYf|q3_fL{M{|1_`V?nbf`hJC~Nm-vG2C9Ll_-gCJ|uE zU?5qdKh`QI;V!S`A@bACtSX0&lJU+xKQ%8ng6SUi=q$QyG{V}Y)jV(feFUG^-CjwI~DQ~j?re?;GK zRFkkjq@T58temM{O6=le#H@`XF8svWEFd4K~M=t5-U#*Y3#lpcTA za70$lJCTf(4YRD%Vn%Q*D5diI>zCcYkJFuo?p8~U0D8fQ8Vdd5rC392xmNaYNa$O? zsh6&JNd*fIz6jv1HT3Sv}VD_T#{TI?k zUU+1XSm*ZAEUgkfxsyJaY2A%X4hj6a5_#IbZdui2GH^e5U?V|{VM3~)RHHOlh~^1T zawZp+Tp7NQEdIo*w7`?;+_QKdpX4l0-&>w~K}zzhNG6xxQ>fESMpjHbfIePhvPE;J zMMhp2?8v3zsTae5B_Y;mM~Inz^x&H}6@QiOTFpl=*ds6z^NMF^-tBUuC%lPJ{1bP$ zrlWp8V$%6bwuXW54@H#r)ys-$h1EvlvL6cAYg%Xs|Bk75%j5N=Ltzk893k=#x|v{m ziDy8cRUwwtLP6RO6pKlD(%Lq!C@9;fmV2$6_;Q815Fo)X~a4%?d00ADg zKaEpR-gE|M@kjwh6GTi^oysLO13-AVoj-$P$hElu!XjC*k9woG1Tx;7ty!-3)MtX z<(f}&iv`gBfvbm_4fL;4_9PKd;O1dz=j>ar8&OEfx+=w}L~2lf_v+isA2%Sb%D7H% zzbA#wN6V6+vX&^6`tv1$Oe3j6G85YAI6S)#p_f|8{g{)&zo5z{DQh69`NyW|RmpGg zp@Dty+%pkX!sN&hEg7_vkXC)S{hd;UaKG$4|5y|{dORZbNB9k;6nbWRm3)(@=D1|b z_u~%tA2O^|NhD-8d=#Wo>K~t#Zgk^Ql#SI~a(bpvTIiMx^r4zS_Mr?DH<=R^9l4tK zqk{!0*YPXb$BozXQ#k-g2-1#oc7$W`hnr60&*l0`2lNFLX6LEDt@Mk*>Q9COQJpeJ zpV4VP2Tn@_H42}SSIoCuXdqmO?95W)Z|^FhuWdSm?53*J`WdALr8A5~b9s=AMf&IFIw)e`S<;)ni~abR(^eXAF!$wvqry>>ei#~2l`V$ z^Qt(i(4yRyatTMMEZLOKr#bM?uHv(jC9kI21~Pe-Sm)80`Yx+@jDk|)Y9bIMBpJ$H zGwt~NJV4=&Do)v!Jv1_3V}Z@H&3W_N${}6FbUwALa8&H*)Z0DxTxC}d6ndb-#=f19 zTI|9w+sSS-{)JcC-p;61P%38YXRjVC!qa)FZFkJGx(WDe2CT^{!mI`l?%CO)-&daZwS{_7{NE+S>N|!qprDZXP@qy!p4~OK%Vy!Aouk^qGA{Y> zkjlGmjBsRVsrBjz3z7&l|6I`_-X%tjOY>P@8h>SPBQ1+0g{w$^d9D^99!iLDBKXW( zdCBDCwCz3eBV^3~?#W!;Dur!JJuqQ#?)q|6Ct|n(*pL>!b#m+{qhQF`zQVEQD>PTp zqnfuVCRuo~T5WMSy+hp{x0wA^_YtvV*bA#1MDD4C`>I^1iOBhEQ<^89+ma+7(%@l5$SeUzTd&0W6n^(PB?= z+gf3!KyZC562)ehF%8`(99ButhUTlP`sQn#2%3E1g2oJMIRa@v;?>(pUc&o@c=0(%^EM7tgR7@?_b>_5uT!DBr5>MlB@{76E3haAAIuA<)^OPY5H+2Jy?#wLkzpOoIv*~|Yp z(nQzKNd0O*X%5XqG98XqOEyY_7;-s7%8XMgjd%Jr(v0I~WD{MIcQmCdoVj{4&r1PX z#?;v}uY-U^VXGBQM1oq}sED-HchsIC2*E^XZkTimqfFjTPU?dxN{YTm_3dzv8Nm+0 zDVgajt?qP4MNV{ah1QZ+iSK!T6UQ(W2b$D^7+v|RYDE^X=I2=cd$ zc_7eT!EJrr)}uYK)S<17n;QYy@u_;5i{&nws??wcZvUe2E%4-eVKH$w*{w(~)Pa<5 zybY9(%Q}aaqnWIvJWnyb&isOk?sSUFd2*hnyGS9pSQx&=WNR)>rO1 zb%a6=$~k9fG1)I_aBV0`%ft&LAYFaTtDn=c+FCv=cs|m4O8f50w$p?`y>ZD_mR!^8 zHc-~|wO{y(ZX~d<@^z1*WCavx-7DN!ez5Dbhc+OZl{`1T9Y_wFZZ@q0FHynH2&Cxy z$s~ZZSf=>{<8Gp8v5ZrouwvZdXe37Yfzv;sQm}d;8M$r+Kxw;;?dJ^_Rhaq#GY2Ap zzb6Lsr*-;*nb+++ctZ@#bJ40JhAl zRXCnBe@Ix=Wq>7v4$4~vX`+9n6^!8IUzD3$esS>#jVXMKK7X~py>c5w95XYa=K|#u zo=~wHU{J^GNPh??hm2wzibvG4Gw$sOCp0les4%6qL0bhrQKg``Ne2852>>duUVgQL zB0e;OEf%}SMp?n#w_O=liSR8UCV!SDIaDNclW5N&*vBw+UFTYZuz#qRLuJU(!*1MX?j)=fpRA*Ha+UX zhZetIXeY0aKB?P%u4cBs!Znt&-@Vq^@RF6ayCG{XG-_g&6g)ewR<2&~!CkqY__2iBzH0vVbduu_uN>vvX7QXdwZ5J*HqV7u zR$RXZkZ|8f)7&cWt8&)I*mug(`vL^Ef2iRNG#%!S|4DY(4q@uk;2j-^hUb5H5?KfM z*{72D4byJ*p{!``t(a?1-~uIt<;?{l4lZIHpAC*xh25~=jcj~nWCrCwK+Sc_UBYHT_4cq4!K8?$s$!k{GpS&r=(W>Yh6Wr+GICeww_By zoqb(CvU)K0gA?|E#yi_UYk@+UscF`Cb&}sjz8Y9y|D6 zpSM!p=j7ImUf*Cm@nN~4oOfI>s#TR~^2oexKV!{s( z3w9arsqu&^iUc27sIP0z!E#9Q7J?lx7dITCa;bWx&7j?bEd<%`@==il`fi-*rq0$4 z>E1pp=Pc$bSm}^HC^ZTU?xk>M9EUT$uS_j!P;|b1@&0IE=+Z+iO00o6_kiPri&yAi zs2>pFX|TJ`oiLv0S_Q^5Y8>788#$u2&PQqFbVWJw@z<~Upk}+fNdm{Taea{RQTd-P z70v8;o-8RL7@kUav2s;{FOqY1fIV4|VTH>=uqRvtxt1mcxDAtAj{ke2*_RI*UkmJJdSU?+6gU7)P?J2(w zd8xp}gstCUs2a@uc}`!*b4M+2P5z$!`IYp{a7oeZmtlgWOKoBdZQO4=q!-Z-+~UK+ zkYC4Sw!5h~Q!7J3lOI4}+^*w!)*5FsmKgue%Er$%OY#E1zFHmC355HreA@Z+jO^{E zak5VoJwQp5csF3G&m`U(4Y1jE`7oqrLH7uVUon~=xouIA-hm4r6;MlzuI|4#sxg)H z+0G9`V4a&sc^A!`+@l2gWKH)i+;ogMA6&!o4=)I&ccHmi~gN|8_h{crm30VXsO&K=X z)3ym1Ab4RUXsW%0 z@5MlcB4^|wS+nvpWMMBIutll_d$_Vfe=w-P0=Jo5UTX4K-Z7JSiTP57p3O@i(Bq6X zM4rm%ze0ToxPAjTOPm!GV#>lHRf5D^=eb!}<)%Ia+eB}h-^q6*)^oaCmvTXpRro(( zUi6b#9={^KmG}>E^<~HXg;A67=c@S2lFUCx!Y}5VTxpi~46RtROnpQ~(Y&Th92}iQ z23VAkhs3EP;amcgUQb;|FW1?)e6Y8w#&yCci*DyHND@jsM1~HVA<^GdBh!pKSYE?Y zQogvC3X8=w6S;FZIOVkh$~t}K)_tCj=zy>n$2&I!wleFNr-{R%ZaQVVPT5Re?B-&q z@`F;1v{yPCo$!D~ii|$s%E^dIGrJ>Rxaea5h!&{tDj zT`{r}&6cCI^q3cjAThIk?UHnhO!54ZsdJ*quv%gi<3FQ;cfsV9REQ>iDTl4tU;_<) zM*ECw+nW#egp|KE;nSXthU>E)OC*q@L=R1yQcQPrxZagHDM)1#1Lv@D9ifjCA6ASr zxfIst?54Cn@T=45oIW}geN*!At_K^h$BEHD@zV{q0+JIx z5^HM{GPax*hEyj4Q%B>q|8(|xEO=9WsA7@S%N;^nQ@Yw9rXQ@U<>vQUmUYP501`aB z#1~(K_TqS4#+b<6c2EVebTTB{` zs~xdE!!5J&-ico0v-wpZIH+4wIz!*&sUg6N<^4&gS>J}KUoP;*6_MDp@^EH%*TP}x z+-H%5Uk&YQ}Nx(VG(wlG@QCy@*ABrn-ePWK?|2uThcgGRK^-prNfVmq+C^UV&o zuR%_NHGfZEE5)sKFZQx-w_usONXyEUiKGt}iq>SGDZ#{8?YT6EyY(l2PIy8a@J12_ zSv@i{G1hyZg-tAD+s5PB8+IcLt@Xp|lebFl+n;bn;5VA>s4EPd82oXeX_I4_Kh`%} zGNJqK@~A_Z0!JT2lU(iNd{cL+VACE`*nbjp=kxDr52PPvI5;~`51(~LtFtDg~;&XW33t&eYx>fBi8w5M$pMzShrC2cz*kvi^e85%J;8apAdPoXrw!4r+lD z3<{0?1-}?=bZD=mm(+D@lrZ_9h&`Ve`d6DIEEJ-SYeG8+lQJ*smBi+i~F7`*o3#Bx&PwCu9d z0({s4uK!7(Tu3&UjdT}1e3$&>HTvUCr~Tld8JoHfTag!y3u^lTgvbjWe%D3YG>Oh> zN+HOWYlYT$9P|CzLcS|fVksjnoI&gBj)Gq@2m88iqzdX^zb0p~4Q1*63)g3|PSYc6 zxDp+&=o&#loqi6V3m9$JEgRdj!q9Sdk8p)CXKd?R-AzgH>o<-YGfvQ^#o9(WKpbJR z)3w8~)KLfl_ViqQ^8l6@|CLz>0wKX&CxPb-nGFXqDg6zBENnsIlJ`d;l4Z@lT^q## zz%qykmr7~3FNnAH9!{4%-u;_6iKkG-qr|`PITsnpgbGP_*Kkia=QIF2C>6nZl~~VK zsMgQQJ@@>!fw=sIRP$9Y8b=yt#HZujh+bd<{$RE7!COr=u9lRnN7VaHt;dbyjR8eG zP0|3&82UvY7*#(Klu{IGuTOzOshzs3ju-XxuRG+Pzh{#JRoI@byUGHkdDWUtjxBt` z?4Q%r79k2-JO^x(#v@b`-(V&Z9VzRQZJckUSM+uBn!%)F7-JDGxqY&Y9k38LmQ74w zkZ=*CA~(m|!C&^Oi%JXzvIvR^W+@i!8*CbcnW~5HPmoki){P+R*CqGo9*h{SaRrUO zx}REfrPQb7le}?e%{(LMV!n#|eJVtZip`p8EXv)6$HiiX?Tl_qbFz%zxHH*fwShL- zEzA*g$HewpUi(__ z0tY<6BB-sleA^`N5;k?gToOtd{`2X4{0zxJ==ZpSX?q{wP%HbridzoJf8xUwtGwv#sdx+Jc@fjL~}VYaYd~=iHB|S!Orq z0dfgAR-x5A+Z^#U5k)jVtNt>o6Hx1^E{d{mZa;^8+QY|oaez4Q!qxk}%Rn1B?0m0D zesWeJV)Smcb-Q(2g(R?C{S}mOIUm1f} z$scTr#G>wZd8@gBuz?s{v%kd2tLLZh84Zl1@Y^aC!C#HA@;YPwq);i2?@=8EKbQc&J>}3u$*)6fJsySQ4G|RGF=UaPr>_PQ|LQ;Ho|gn2xp5^l&mTb&70c6qqk zn)oX5rJGS13h?AlPrAX4e9_2sgmuD*8KUafsxY7*zO6R|Zs69`sYyIUloiH!gK3s= z6ZQW>cCVl9t`Qj(cOAHMGwN4isK&dH#W}{3TE8{=gC)n05AREOF0$ zlnlAo1@Aws%4G zYG(J%Ik&j~E<6?4R9)?VPY?iu5WRwJmxICtleB^bYHmDfiS|`;*3uQR$zgK$?67Yd zRWz=1qOv)=JBl#9Lb@OuOaS(D>`)t7}K z*Q+&>-=s?&73)VM6xf1&d>oWZ!9p1i{5acWRH@8%A=S4go@%R}_k*UbTa&$47)4lO z)*-Yg%8yKFg&Cr1EousRH6B zLHp?ZD&v959-!Dil?}@>3YBPG7(7mf!R6U8rhBB$%kE)ZyypKFQ5oX-UT<~f#*I&eJ z@o>IyX8A=*x*Z(QnlbFMgYeZf|8;27vyYZ7TtN5lJt1pLYWJ zF&a~u(~PUOI#O10Be|uIWiM!V+`dAGEd7~`kO#Ife4@V5_Pr$T8^r*@ zcRpJIAXfqaORCBavm2||cXaE^J;dl>*>m`GV1N{`zLX~3&y+~S2eM$A_?rN1>-H8pq*ar0P37T-4v*Q3ISIKUw{C4!Js2Z!NL0ny zAw=}TSemN0QQGrB!A+c^wY_5@$ETMXf!qdD&2m)@`ufX76Ihf1+T+R*TrVJvm4kBH zb=C5))CiCi3$3RG;is0nFTRS6b5zf*HOpt#J@T;+ZM95~)=Yf>^27{#yK#qq8h}Xb z&)~GOnJKb;-_5>HheL#_k(R-OJ&H7DO+}j;i^oyw`r6$!ZH<{mkcnISw0eYP0`@|M zX1nQJ!(fLxX0~aC)$Y6#XJFJM+tbTBtB;CsYW&XZ&fK;lCQ*GOfq~^w+{yJ*eQL$ZZ zR;+(-+3?2dxW!FN*4MD9=Lba=8~7xCn8+;njc%^{lnl!O#on?^R*1~Be@ zjZewdX2YEK`(}yB!u6xDv8nG7h*2Fvv->^J0w{ICyI}zMs2Dh(NB8dEczF43(-l*} zRT%}q?u3f&vdW?%lj3gPRH(37DD7v0&8n511n|bpJD0o>Nb28h=HHZmKpSZ`|MnX zk#H~&A6z7~!M0lVN?R9n=M%B7s!s_?MjX~8;*92#4F&_1WG;h30g}L!spc>8DuNan zncNESvG?#F=Ra#nWVv4m`P{(!4rbTCZ_p|NH+V~-;oI-NqKyXhUFVAsTDN!|${uz` zgEQgVqlp*mws$JZ^%qt0=6VE!ajn`-2@^EvzmVBGFZ}YBE)gD(N|D$ZdXg!tQ#hc zkGIvz9>4#sZa3lD0z|caxxcK#g+s5@d)$!tO?XtOJS%=7kJSA8;+q(gr#1qTVR|^3 z``?i?QTTSJ+mY^%k){m25+Ah}chD%Vg|LTKA%6sqcu&544%g64W7ZIU5 z2u--U^ds4gWC_KXqbH+*5_o8^%&DY=aM7A`HA5t$c#BCTBAaEE0LMToG$BNUr0PpkrR zyY~UtHHCb~9Z%zKGXt&V3grq2oidd)rJFya@$jgMWTA{pC;)!}NAqZf`1R#Q_h8@D zIwZ}>8VzfCIJxujI8=?;L6_(yewp<{8!cQ{EOekY9?zkS0vv_LkJ z&jVM-de8AWUw@B&Ng=3f)3!}!r4hNd`W7QxTsN`y-c*x#TDu8%!q`M-zt5lZVgpQ0 z0V27_Diyv#P4`}FxC4}jJ@timYb@wu2OL{6TfD_V6&`zgL4E7HC_Z})T)}0)9i^_w z$*5<*04LrxX86FDEMdQN={{;o^|dPPKOJ!+S2|PldhY#Tm}gA3bCE=+4o8O36}zesh7d0>Z$5U7r3-Z)JWh*t_fto}36Ec8SyjZ{~g zbt#ZrG2H;16-GIf$ef~?iZN>gSC=S1&69B`dcLZ&p(75#{7x%z8l2G+i}kiz z(gSo;kMQD0!z6Q&-n%W?l1LbSe>fQZ@*X2LGjiXy_eAJ~q~?EEt5&PpnJTeORMSD^ zNEjhB$w7amV<@4uIVCSM4FgPFxC|4;pwxuCX0Fv0>5ltM(CL>~oEJn>rD94C&L=BX zV({Bg_oaO_=j#^-3)7l#QRGlbj34hbVKP7vkRHZ}biy@XTiS^XdMVR*Ele=fzxYF- zLiq2Ak_=nf5e713Jue<9sQs~vhxzznVG*~0S1Vd1CVSm7^;n{~LoxGS9f-51e7T|@ znkpp~w}%f(nP6Mxc^BN&@5+Cf=iW5Mw)s}@T9DaW2cgsl?K)7mOr!0a;mX%xIct&s z&ZHFPSaw}e zIq>N&y5r2bvdBJlS%R|AKs9J+IRiv2dIKK774)opMl2t$CuJJyn_*Atn4N;}o$WL(`64VThx{_GGY(dQDkm zO|{9%&(GR*3uo2cO59+xb#PT3VN6~s*R;6`RqM6#H)hvdYnpGxj-=>ccjsfHf*tIcUWZFg z{|Bf=!#tn906^1v)-AHqBf2+0wau=}xbpsLeO)E2q0k7!VX6M$g`K4k0Kb(iOtc_? z=5kQ({`7@p$go?42|jC#-pyu zCP%$Hskqa<@8U!aheE0?^hppvffu4nlEdv;hKToGeb78<0r9uvnXHi!|e+{ucfpfcLk}12(kqGh0it zTgsOxmmu-)jo9l#~kJtEBaX3Iv{SL_~& zzHO=cfx2L_BUP||P^)`|niZxh=O&D*n}L>-3|5>kRTpS;TpgzHtlK+0j0&2*rwOlU zh--lh42JZCFm_BdAYog_ax?a|Ic-kv?r)6enkO2b1|?|#Ex-1BP#AjRlV5N(a{A_X z1Kl#DdjY;PLfjYtiE|=QFX`21 z1?t*J((QV0d`k(-iDF$HAmp&d&EjeOda|nxJ-)O;6#p+&7k*B)P?dp=zyYC{oha@5i9Et29n+Z z0L$Og>Z!P=qwRyvfn~ZbZGDL{e%>)Jl$xIkw2vTdUu8t39iowGT!-j)>Qx;*YC2 zd^1@bIY!s*RoqIno;)ZWsw{5gd`L~$b_}&vcq#>;=7_*kCw98nG41Dfb~Jr`rcUcE zRBL#UyJi=dUX8;OXHk0R6nYILrwSu3hKu=&goACISWwK*f<{4czpS2PYs?IWfVR4n z)hv_hD;DHI0G6)F<+;#u`-)$h^M8UzUZMq$1kauN<8G>^X`!#vCuod#Q1 zq^3(OL+yp)EEtfIQ%zT6)9aw+LTJl3pBgV?1DGx^ejTlTc=^@H!F*w8Q3UdE*nQz* zCxK-#ch_N>^;{n!AaY9M`zv>i0I75t;j!|?Rp<J)YbnatR6mUJ8uLYm}&9wl^ey z08~zZEdXLys3OFi=!^+rr-yeX-G47dP0rA-e0aaVZNCyvDk~7_l-`t|LTdbBr@law zLy*Fu(=z&D(8vp&BA?#5wAkLba)+%7a{0`FWPP}a7H?bp^Lk~gnJnX5xMp2`bRGBz ztH&AFv{P`q39!UxFv9GqGWUoasyNYU3?&!{iI_W0$*Zbbchq0Y#9fzX2`Brwy~oa= zAOM`9(5ZCciRJsVJfAX*`%J?{03NESI_6bqW>~E&0nh13ZKnAQwY6GAAi(u7kH=|R z&H9Gj!5#X-*t|tGN$I(YG!15*5f(L+y#&{qc||ocn-#`qU{0#cy{`T^N`wslYPM?~ zUx@!9Jz|7duDs1l*94hNb@Vhx@*nKl%09XpMx578|FpdtJ$7DY@}lHO#araV`po6P z45FZ0tbk|d#M1e#*4jsRr>M=zmvLWNmFIK&XF6jn7dnYw{{wh_o7poL^l6bR?}gVt zCTJ#K)0vhfQzC8GuRhTd^9B@=#@fm^C$($zPinN{u(||w>wZLIfO+4q zKVAO;8p4uJMM7BH8VS1=1{};U>P{NkMQ+Rd_?Q|s@4!q?esIUSLg)L*_&f}aMz2jU z1I3sv3}Mqj%hhr*!uKolP@f~+l?VR=@N-3;92*no4G6a{>{I?0T+_T^GH4&(eMwx5dm5SGD1$aP%5f$M?qr8)yh&B*4WFiiK@Qd`sep*{f#-@Zm<7i$PAIB{zlv9P1B9!%kSe45B+(o zNN%QTQ7$EB;aB0@zPUfmal2Oux9TYe&pSBYtIgS^f6_RD+2;RBxFsq-;XtB(KTJ;J zP+GQs@Mz(~g4*O?_!6nUBs_F+WAdS${}SdMTvh$(sK0!&q&&~D(Y#0S&eg{Ee8!uwk#Y-4SP~cOTI+#0RHt++6F_{xw@P35ToFKqj%XqsFzHz(#Arby@d-bD?>9r zAQpa2_x`)_H~Tc|$EbZCDO#j?kdNF)k^7R;mq6sl!4EdkKd;A~>P5Dw{!YryDJ^@H z_;0KBUe20f&GViQkg03cTHbPxP~1e(c|B;C*7x;MNcAM@lDB$6BHeT?MsbtCqKY2W z@P=8#1#)Ixf$|^=wdHnHlFbi2+s%*U8wrchpR1`Fc^{IzOMONXFPFK1m9*EiH+=US zsTaHZ9`VMEuOB~;VC?NIlH`-#O_z~^G>11%_E6Vu|qU)e{8rf2P%ZS!3@>+23v74ZsiRb~z>&q&kud1#$_ z{S0NsWUfKo81E*U{bfe!9ZUBDfv2_8#XenI^!*!$g*vAzbKQO28k8Mj;T8a!Di)eC z-R##@w3EEl3PsDkk~f;?(S-ZV9mkei<9`rd)cPCUpK3|9lLp-|{>Z2pG$xaYGI(6= zHCr~M^w|xM`Xx+sZxw=KtfMkwddrvD!xU+R`Jd>UugxNVD6}mX;80>%{w?XDMN2V& zz@C{;0_FCzk5eoR!JP(sy{|=0FaHxmlue>Z^Xqg4HZn$RL;4$u2H$|HDA#aJpe#m z=hWvfrQoxnYh~nWL#-4%gA%5aXL2a^FLq^Tb@-U2*VjSnGL>#08! zshcTaqKq!%>9I;-eSNww$mrE z;gECYPOi!<4yhRe84s8b6I}Q``AAXjI5=DY8n>7IJI&a+FDM~l6TN-!GC+Q^du~pu zPzC%+Cc7Vyk(FTbsgbueGlJ<~v$B3BrpKSEmEEH`(r46M;cN&J`t4SNkbz%SH@f&R z? zb!KQXw0YWPLJYheiXgIfEhzx#uv%T2E4=x({~`{Bwyi5u+hGY4XToX#UHzTod4JMv zMX3Ni9+TgbzhGXj=EC74a4jVPQl-M;s|~mgY<{RHGZMNdt^Uo$0iFZyso87eAWyGb zHk_QxPN#(k>mv?`?2Pe1rSEaG3zcjvPDDY;+RG*CYo_hZAQypibK1gBck?P9ZuIJ= z5^=`d81sSrZEQ`TqJZArvs+o>~+`5C1x%BS3dIIK8NKSAjUi+ttr~Z@(hCr8D}^Wd<&M$PS+E)?>hD#J>;rS&Ic`YW<>0s{7C^ z{O6K1li?mA^}Uu($y1+upSai^LZ~cFvfF^iuO-9os9A4UlMGQO$>vg6bW)Rei-(aV zNeTIT+V7gi2{FLFDT#v!s3MZ0T;DlsDF_HGiH6g`Z~Pf+7$1E}f>Jep+v=2cy)J@x zNxq5Pa)esPcrHBk?+KK>9+bDXRfMuCH1vxOuZl)6->0;&X>fl$Gr&##eEMy-#W|>r z^Nh3lZ`mbJqC&4&{SpfJMa*5e&@yJWP$rpHJM|iBlrdeSgy6fhJDiRrk?|e6+GdCK$aI%j4h=m@5 zJspE!{RHjxxRj@5nSo*MKjYR9tBxGI7ED}a!GG6z& z1hPs2k?W@CW3w-DYpwmZy`}l=1ukV=yQcDXfLxboc84nw7N>opNUpE38n`8OwnN=% zCxgFu+g^Gyq@al9Ze9Sn2j8}Csdr8g{txg(2s~jAeX9%zR=OUlbx2-6&(;)RWwy{J zB~+6$vi9x)YzO8gndO|(h(?r#s`Ax8<9LK?cj@}+@YC4SioAMaBn!`p-&09CCsOhZ zIqE}c{(!=|VEw!LiROiUr>X9)S}p?yi`@sm(IKn-*SShM@S(#-M};dyNoI9JNlMgI z^iYqt#muO-jQ`A|DSgjkw8YFjOJo4O1oP5NV`XILNxB+(<$nXuKrp}AbGR zip9M=Hy|h+gWnzN86#*joWI6ND$2RQZs#@4UqTu;LUIpd>s>~qRnhQy3y{RuH+oh^ z-Ha2DyVAMZI@GLknu_g0q@CY*ZR85%wTM9{AP{&J*XmKr<`qyAExU0&vC_GnRT+?w zioE*wtR)_%(PtT`ZVneDw{R;C;IWXLZDZrEiCU~mUO1xnLmo~&fAxUNP)IM1ax z;aK2&yzU3n+L#kOkDKYLOj?~C_JOTjZk{*H0s-hv zVA8 zted%rX4r5s)OywH(5h6AloQl*SyQR<&ryTUJ5=)~3ME^KvM>S9eD$hs+qSP5&wpBc zrI_-2bCFcq<8KE5dit7%%@Jg>Y|q{&j2}@$zsZJRF^piLaGvF`3m)FUSBLye_sTlcHz0XRy@R#;=xVqEY z>s;|0+Pr`XZQ3C2NBhT)eXCdDFM_Nzy-q7lItPx{A?A6qIca2b*Bz_q%{M_?Yh&{H zPQyN9V3GOOkat=GcR4%l9^1iw7+pV3vQ{?e2)wwBc8+`3ka(u%{{UFqDyrYJ4B$(E zG3j2X<4H8TD``%ntG1~lxnHvxcCJ9+5IX%U=8uT_q;OAdaMzl2Z7JHZ!7kQO@6V+~ zrrRf$`EA+EJzOoidBj3D;|H2_g)%l zZFD_8(-JNOx{4W?;Bs?ZPMzWF-yU2|aVD$%oFtUU(XFtKGRLV6*AXC^*v4k!_2w5jSwC&(2Uc3L*S2wKO8kr?GFb_C62cj&vR*LE}Uah3ImeQ`}nI!ph>?_6oH+(MAd`D?>t?0K=-WXtU zi1=q1=-#~dtNN$I&kyML5?X7T-jAVbakkadh3B0zj@idU?M$-pz4n@t%x;?h08gGm z1gJ0o9R@y?DY$8SvRl8rm5+k2ZQd)FNteoCa8M7reJO@9!iUHmtL)zv{>ZWT&h7Ob zBK{a9oA1C*2t9L*dRNcC5Pk^iz5|j-&IMwz-k(QN_YZMV?~OkeTQ~ zHAZwKb;02J)|`E`85x!L=-0;j&1rt46|(NIkJq$Em++$2|xKjP^CRXu!r5&f(jfV!5kXys{4N z2*Dj|TGn9b3aBm*T1{!GEa>gz$m_HYm^GnqH(j7_>OBo}@Z1nVY%YJh(zGqES9k%( z)bt{eHZ4HC;SL5{nRIpYGU zn}N6m9Q7FKRVI`$IN*25{#8|^*nsai1Fu?7W3H3d=8?Dvcv3T*~$&x(16OM47X2eyBuJR%bnh=YP?{U85usp zuf#F9a>RO8O^ewwdtK+|18C0~6{7@xUm%_b89i#OF?^+t2qPa#yA*tnEN=jFxYm1< zWRCr2h`LDA zNmuVWp1=@!C)S?twie@%P6+g>5H{_^gS5Ba>Cd%2e%i>s4h@&!gKpMcndhgyHq+3-D+&lX3xIv8Rg~d+gV!DDL_{&lI3)3f z;<;wiI%87AZkYS0X~r^5NQg@1PS*Z@wDpA>7%b#;z~ZKgAfTuOu{?iOfW=f& zyEt5eGsqx{vuxzQQc!dR^HFr+LD=fHDq9$v71i%XdFH9SWLoA2;sQlbVc1 z0op<1H7XkDWd+Xr?t^9*6DrV#QX zQhf=siwZa1Dsz@Upr~2=;DvbWij4_7LlA*_FQKIK0l*B+lZ>#%cg1u$8s^lBe1OVO zWqanRNEaJ<40@WmaHDol7#YAdLgn`cBjo_#kZYkL97@j29Y@N3W6AH2N<${%4mN}O zQ)C0?Cm=I&8s_aUMxM6xVE3)z3Pmu5~t2h1v@TVo%Dw{>h33CnR)a+Ep~G1%NMY>}F$IaWJ}027MIxP&Tg+Kqve&m-yWSEZABH?cSdvt@M(K;=gO zXQgK;8>loRU@JGtoe~;2L!Mu7=G0)<@sc04anmySob=H#btI33CJCVbIIRhRJt&n zk+cFbqjpF5)_v{5N55+oKA5W4x}!%CDa&WE;;?lMV631iTmX2_Q(0f3(-!P$OQ;24 zTRTTjbL&+`pbKDSn4Sr)eqB*X+&)$u_2V@JdZRRh<{*z=_0vrmP2BISue_iR^&k!h zTI#fm#&rq^5_;E~={k=iYjd7>uER>W7y+LQMoILp$h+=yOUU$FE3{3Z?TnBJZne=^ zUlu6k$j8fodi&P}VL2}wBLk6-t#r0mA=f(KOq8xI7xg)LZ2UFvkRzKj)=yU1Yf= zV*@zvSQo0Qn8^#%cS0)}XlRaaQE?z*asg5T0K_QS8?+22sk;e zb5dkd6J}dE2e}@#gLY?PXFaOKl14&hlr95jlU&vOOk+lkv$G?f)zn=D4S)&*k%8ao zSXYq%nG961=tmf=S~ZEr>Jb@{fF#CFO4mPgX^q@%9WZOQx`9)6NXu{t?N~a6m@?tM zY;p?YZ3lZ4k3%CQFrl&~` z{{SowF_5(AqXBYG`0G1~>=OZ;{Ad2X$69XLo0EH?e1|*UR?mJQl z$Qv7idEJhlly9(_u_8#S+zvrF$?5M_Vci}_1bg~=)CNB-;|B}~9QvAkUpbR3!GR+u zj-9IcpF=ja)Y4m}G4ILT4o==qDmktJcMG|PKpXj*t(J`K`r`gJJpdhv z6>d?x7{hvTRc8l+$AQ|dK4M%-@-ZBO3j2zQt_Cxaoc>hVh{FNfoK;7Cq$wFaOhun!)v?IV1!1g%w^CaD1p@kE@>Kp2E350eE^1M^uF`IEkbSCdJDr z&u>#-dHf#m)E*1ebZB(jUAo=h?<0gLCf*WwEPE5z7113P+2mrT*Se2b_$%;jwBNH^ zYQU?+s@_x)?ZNMsHSc#mAGfrV&$YF*M;Kq;e4r?)Kcr5l8b3?7GsQ&$m%mna^?7SJoObHi1h7b+{-G(XR2Dp&&gPina6%V zO7VXZ{6fFIzPQq?Og6$dDLm2@cU8||GmO`r-haZ!vb|Mr54E&vqA_uEID|UqfE9VJ z_rjhQ9vZfP5qPz<{SMCDw8GsSAefHipPkKNCA*ReP-g;#1$bP z>5E)R>E*}Y@D)qNpBr_H%Q@21M}03&Sej3_>ocvpPCIN)xczI+b?=E#>Q8&(Z7WK% zBHVeg>bl8qEk{5j-TDO;JIii?Jg>KI)E$DtXmaZib0+ zp`If6uj2mzi8{@$k>R^_o*^Mu7Mb$n&sTXmpU$N52ZAiTHDNNx9j2kHMIQUx>6n<~ z80I*}d2IJJ?iN=*2=LaM2ZB5+d#6~$ZQA2i)LuZZ0uD}Fu*Q1V%D)#sX`c-0H=3S_ z;Egz4X%-C84OV5!+_~VDUagb-Myi~t$4Kf!jQ5dGPma%4yl1$(^KIY?BZ?9i&2~Fg z7rExWX2u^6*k4}hclzC)l^&S^ZBl5P%rJ4EDLWYm5W+b1+l;>nrVxFhJ>VFd+Z7{!v zbVbp8GYb_umFfn%x;g950Aiqv#~v!Scy%bXd+YcR{{Y9dY`bxe0a4Ib1E+jwwecLb z_V=v>Rw$T3Dt_#Lz-!d}F>QM#h`#W;=~rd2l3SvW-kfI`>(a7SVJCK9BiS9VfW9+b z>-Scc8up!N*79+(ctbN}A~yI@~)pq)DMUJO{`yNI?TGBld8L1G`B!* zJGz6Op4F7=I5v~}%T&3yM<4Oiz|!4WL91+qZ*=k&<6s-63_eyT6-&Y1AiA}+x3-y{ zNKeeDPYcI-($+4oC$PRtnTvV3U!AfFgTVYNjFwxgO+CXDLI>U+mCZa)Gg=)H!!;Lk z*}N&^sI9dXjpN$R`Q8`+7rs4z8ub4Fh~EUX&xk$+o>}yym4caKyNXOW$o8*`F173H z8`x!vRGF4AilunQa%<@CfnOYFghywpN>GkVkU{`^W74}JUy&(pjv3PLq4AH#?}1ld z1Zi$!o++6KRanUxAB}ia(~dw~Wd6SO`pNN|;bpJHn+Y!Le5;r*AS&w$Zt|FOzuLbl zJa3~-;m;K6dR&Uz?WQ2J zk(#kvp<*7H&URKf73Q zJ8)FsbjB-A-f<=|k;i(LR_67q5`dGytth7Qn}hYNI2hmna0g#{(1KEOqy`^amwkck zZraMAboofeb5{~$xNRGO90S_1QNCO$12twdAV4#N`igrIxw5ya6vkAyuQ(MG-71zl zP5|gR6+p?iZ&GoBDkJg|%2*IOdeD}GxoTPPfpR(yqcuk2VkBU!tCC0^sU?)R%yXWg z`qe2Jfn{O{$Lm5&XeM?F0el~BYPpnxSbW6hlXAOtH~^2vtCu(dTb}r&(X@MyXDT+6 z*V3BO0N?_8bDCt2y9BT;+t#Yi!T8QW^sJh(-%6#diKLLA!6AXQ)1_yw=cuYDr((m#4?)dCg5(Thu1LFi+t;wF?3i7^VTU6%&fbSy zDGUgE44*+#jm*7pM{!9Q1QJ^q^c5s)#~2_D*gRG(+hftEM(Cax%$x$to@)KOj1@Tm zow@X?(E=kGAaT%Dm{C~#VboU~)8*GcV{{ZU>oE04H2d!wMAOV75x*tl!K0aJ-+T9kPEZ7eq`El8KJk=-=;QYkn)4fj_CpZU#*EQP%G=57g+E^OBqM?j z+;9a-W5bu)SRQjp3}HS}aNR+nzUF(e?1e~HJ4pi=rx^B(k+VEuy=lTi`N%l^YR8rs zFU%z*Zoxf1waZd=HiEfXBIE=+J;1A0;Xo!+^An8w(6JLT?F63S9xCinFefd`ACzLe zdXjC}^=QV*0jDSs8TW#^S2E`{QJx|iAF;ZmCe5+EtD(uR{ zfSm~LDs@32Di>)4@&{k7S%xjl3Xl~>7%=42$s1(gsTlO+S7d3l>~T(_waQXP=@4%iN0*-1P!MgRPouFPUMSuaq~8DTe@A&e{0nhL~cujArv<3 zXMieIwv?~jB^2;|v6|G1$jcX&ykSN?ewBP_tsH6s;1Se}kz0Eux8O5E7lF>+wMuIM#fVS8KcR30!C)~Yso7PUt;cWy|=c<6m= zI$hB&RYwQr2D(eTCutXW#~iZutc#e?vnvdYj-!%l?Qq)mGA3oqH{HkZ^~G1h2PK>3 z?m_9rUcG=sq?Q9L2OUjjG>l0|#xuw`$j5qDxs;X5V+FtjfJO;Zz^NU~Q4ze3eq4;z z7$t4PkWP6XwO}y{PScWl=clz~*K-{bCuy(|hynT^fAFhO-ZBvwd^qjv_|{>$&eBFk zazGtDs$`i;k^$i39ffjAD*BeC)f#O!BL$o0W$V`zeWA9TfxG*l{{UK~W!f;KC;;ou z1p;PvBml*_p4Bl-mh4>fi!kZNa!2y3R_~M!TaR;DuuHfR!u~xCT(?mBI&Q(h_pZ2l zoKto*;si693hm<<`c<1%JIO1MK3;k+y=O-IR1MkBRqvl#(Tgsnix%V#1~XmM(dX1@ z?rmFOl^NvlGsSM%$RcG5-*|OAS3PNll~kXYaJlbSZRZ>Y3ykBQYqc#8I&G(OuCurz zFu>y+WcRLy!gf&4oGW9te+uHPZpW3Av;`m^U&^;Euayf7ump0#hP1IAZgpvTR#Y5d zbsyHQ3)KK%tcR}L^{!eCLODo{%MuioAaVHA@mzwdleCYOu~{i{CLB0mgC<9Ou@tudi2W zkAgaT@H*5tce$06uwMMvKYwx#;H<1c`Mz#zE@N&+O?#-QUAb)MC*HZuLsekJVMqH} z>+MUZTnlv^9OHmb8LoQv^CyfBIi}boCGN2Q05}AW!1p!JU+Sc>W`0MZ&M9>Zpt#6x z`8ndTX5_XyW1jh_ie{S=$*L&?DJLzToBCC&E2AW&ILThkSPjgfwuUHhGuT$GjBJ5M z0U6F%R>`7VZ>id8b26fpJoV{ahLv)|80C&K4RHEUV9|n3Ta$rZhMKP$mcSU`5`Ok; zq7O?P(p^sbPH=`IN`QC)SRcl>t!4|j5xWiC*C${QELcH|^c)=5N2d_s7=8ZYx|3%- zk0l4e9DbB^ijT`F9sER$zPZ7|w8 zMsa|Dy^64^7ly!gIo;fv&9u2#*J~ZXSLJQQjGT^ZWum${^&NQrTdwBX zLC$;o*EfB11dO|e!pD=zHKTiV$jQsEHO^k$Fk(z+0CgRWX7@B&85)er-zu)ntO0Cw ztQ)yJ=2zSZ#sM91S~rrJ+_Cvaagr+@=0+s&HZUh~JrAXG%FUfI>}OrhNx>{xKo}iA z3dX%%yQE;Hm-xEY%~hjn`u$<}e~sQ~~SM{VOK%2g<=q zIX#aCweAm^kGl#mIbUB)R(0phz>YhE&6Ga~SzLC3ynHp_}uEJiqG3Q6^= zEeSt3JPZOG2lS|I!Ei?7)SsAks|rNG5^&^p9cd`L=sV~kU9HNNUVZ7y5O!9|BGb{*e3F$bKUqNl-p5%XYjxWM{Ul0tmK zPT)KH3U5=CA(V_9pVq4VNs44MNJ14O<{be!AXN*KzB7}!9;eo{_HB0DqF%qo#eT7lX@28QqSV6;YZ*N}b)&n{slBqi!+6 z&sN$#C*d& zy=fwWhDLLYagO!1q*)m+5CzNj$86SIspxAY(z)!P1bicH4fxburOX2H_rIPvM zYwL(0AS)26qxk_`ZOy04pyLSNQ_p@Re$hJQHw|?6-efR27PfH=uqXJ5JlB)z{{R+l z^*Q0y{6QqM0NGe4kdWTkC#TY*@wMdt0NOWkSxALgY?4G=jQ0vnd1kX?Hmj*bW|5_a z6#*6dpQU4HuT#Dho6#PNfAM3)zZEr`Jt?J@-%5>qrHv46f-m+*Ku@Wz-^ZU8FFZ4$ zLE&!(+&z*7K`O?e%QKI>@=59KUR|NZFNgH>v9#K^mpN%B=v8|PCb1UK{hHr0Pb(Li zeB_g!05tU3KVLHamE~(0J|fgLuMla^qCjLzNo3!7BqaH_j(@&-l^ro%UxhperhGNi zbnR2cU^G)nzS~VwH%b?teHY%ktvgt;_-)|1btx@upuYipxvx2CAF{?f``6AsFY!LN zawOzKh z>7FtN-Z-xwxVxWIx$|`y)tGU(4itI;*jAmFhi5;#2oyIs#xYkQ@I>Zaq!#C(6gj5u z+0wADYYx$@T4NpuR+2vjp@OAfy{#{vq*>%^> zcHG3Cqv>BjcvHZU$EHS(I*u0{5IE`STmB#LjkcMnnIM5e9B>7CcAW$f`A{{w6kakH z2M6A{r6k?@9kIcvBgS=a0ZVrYjV-txz~l0+17Gl~X#|-ecLlbBsb648rQ6%a9dI%` zR~_RY0@`?r$)i;A$R&=`$^2_MB-njk>Dc&#UC?EQPcjrNnE>xR<2CGm0VB~bH8hU? z{M)nOd1ZHCR38w$4EijgRak+R;AB`T5hnwT zS3@FiIKavE^sM`vZMi#nJ%wo()fhPTIr-NI2hywgd$Y$LwXHRvW*;cxsz{h5e+_Dz zW-VBtxELMjRd$bY+*Hn;@q?bUo>q}yK8t}`P*c9MEf^hi{OmIXL8#R+myYD@>$+5de&F z+Nnwg*HQ~{!Okk|*eprTe~T4bROf+}_3u^C6SFhfhTBu=JJesQ`b&uY6oYz9tA2N}gd zCQ0BEk;V;q9LrqSm1D=?vQ9x0k7a(xmhow3}Nkx@u*o8|``R!lO3 z{CPRRtw>y=p<&SEX12NJSiF;NPzmRu3ToUzPBN!rjF3;|SmZw)F?qXDd z{5R#5A`pGoP2`J5MK?W6q=zv;uRJRo>-@+QbZTkMO9^CmX-uCbqc^Us6?K zqzqwk#WWBYXP(E>p$;D}gVU!@lnL+y07rZ`PH9S3B}T>A6(r=Hr1P}Y$~WhLFnP~< zq>`v*8@B*CIq%-B23J)i;IBUQ#ZpbQZwMrd(8A8^kCl1oJJtJm_Ku{5_3KpaZUM_P z7s=x%2c=m{mB;{r>N=75*9@=CW4bbmj*Pt;?K^{Hp1(H~*`3gBURNVM$)-g!oQx3e zq;NY@?kvlZwZOo0QyAJ>oAX7)k|rY#0_Qz@0au}j%H4Mod-fGt0p~J+3Ml7j<29`c zM#HDi#&(<DN5hq0OfikL64rcYm}!>lSS_N*2Zm$t#LE z5z`u%Ryd2>Wp=>mGm<+F)z0cR+ZdC`Jbf#?y3?cB#|A$@MROXhuu+0`>^NNa0;QyD z&T;p#fGXNt7vOFf9F^!Rn~LxN6dWf}pIWsI(Gn_?xm$uWirNsi z$1LZ4j;`A0fJr2eewD8U%H}?PQ`)&pIhSrpUBLCtUxM&&;f_xuHQyAHIHb;)UNWO( zh8V)*u5(t!+Qf~AS7FlwzH6CV{{SQ`Lp*bzTA6P)g#hFej!sT%V~wYKou$5@X2IMJ zI+MrgSK_}-q=vyKBpmQ-i?!5407iF?I&`fFFM}1&ET?xSrM{)W?9%y)WpXlcgU)!X zk!mZGmctxmXRp31k+i-!0a$>2PfD`)!5NM-g~m^M=9I0u+ZY`?>RX5bnF#y9o&l;d z>Ms)&Y_olM#bAv~Dxe0!{{SkEOSURT<=O^tPb60q=8@S7ZJQEmhZx}Q=sl}C<5kRV z0)fv2@@pbXwlU<8a7jG&^s3TNjzg6Ps+#6?F84asHl5LKYo?KRDZpcl_32qRcVal* ziXFWRFH`jPsLaf=5{-o!;d&a(zL}URu_`mh59v{U#;P4zTK@5mR^84sSXVb?mGB7t zb64)z$;cQ0)=c}O$t}A)l@+3Qc4l0b)T41HnxGO``c*Wxb}fPrzo)fblMdZ6j(YW{ zMHwlA2^s1KTCQs#QPiTAq_OPY)umx*#fasPN_j-QWOk&*IQv3jx1$Dry!{4Yn{_A!2=}jJzG8Nrm#S82i+WG9&=rAXB6XPbed>V zs=jC$`LacH7GuehFgg*~S36)g!!OD?We1#&--UI>yBxBK(ppY1plt;9 z?^_mkDpX*yp4{U#%358k#v30mJeB7)rE7K+Dkvv!LOTlSO6Mewx+#oDkB~nZ9gR5M z><;Sa8F=^Ru;aW87Y@fgiS(z7;k%a@Zg6X%Y;xOO3zu^P83!MSdcwSwV-Lt}-0iCp z&E^8k`2!@9c>e(PRxS0|ExQV%Z*h~FhFVOcaOF=p13W0tO3S%#UA&UaD%HgKV4;h1 z*F8S9pC-UE#H4`Za6N09US~pX+7vDrxeXyW3)N5EHIaF8thp?B8(ScbYW?ixX2Y;- zak);t~qO|){+wezE$P1RrxAO3Nm=W$;C|UrDWJkDxTbCqAIFS&O4KiwGr!cRj7-! zp~2jv)ODm-h+(-|K{?BIsjZ>}p5I!PKuI__4c9-!Yee1jG_}^Gkbsaj&{;vr%}EN3 z;GLk1jidCYM!8INBer;}3lfm3eozNE>qj+#MX_VrEzto4_1edrRIBq$sw2V1Hv`(6 z2uS&00UzESL8+G@N3dge8|qWu#apQpm0bPp-93#k#0s_)bNoFjc_Z9L+=lDH;QP~L zO{%A$9Ag>AXiUz{vL%aZtZaMt;}uTvFnI$%(yx=qZ6+NPgfDNURCt$h2&8AQ^{lzl)aX>(Ga4)c z3Je~(?^UJ9V5IFQpsg1u7<{|?5!`)iKIM^DDo7)d-l`7L*iIHj3y`c^<@tv^cC4F{ zLon<)#cEu?85qguBl^}&+?+RJM+Ee)>SrwCWlfUVBQ-)oMn*dKJ*wrvATtBF4%zQi zr6qD(xUSgiq36mvkq??d&N0W~SFIQ393D8uQ%%mxfC(LOTJ~{%P3>D-K!RvnL1!r6|_EFj6)SabaBRO)BX_nQs-XPZQ9Zmx&emeCmiSXq^!`Zq^Y)j>F|@n zkoafAI&!9GxVeNFgARc5a%%Rl$6eQzqCs(Grj4=37~*S#-MyB%-vId5_WuA0aIO3&k?uTE(Vmes`SDwE07T zI%N8q<32L@n#M0LXu3Yp#Hu?H?kcJ%b6c&NU{^q~2bw>~cMr-qJ`b0xjN zj%)Ls$e12d`+>kd)#dtsg{ND{i{>!T`@-CiJ*(AVzqq>&ZZNXU<(#m|Irpr|Q_h_& zkm60*b|a_KyjfLUkE^AGaD~~->N+{OyOwl~VvR6^)UG)-Np&c+%bm{zY_alVZZdJj zc9X1D8fK@tNtICGp6Ae4pIz8$8pWKKDFVdKKIv?lBQAYWlquNrj~VKE$Bni5bqm%E z6zqZ_#BiwWqqcjRqiLgFDMPaW9^;Db?0gfd%8b%RtFf?mH$z*Hd^*-lNbx`}Gn|Ou zILB(~oa~jEy(&&xvj;%XZZeY?npe*ONayQaHm3}Bku2ib<8U(B9X}eIP0}@c3$3qV zENY|f?a$?1CaryF+FmY!iLf_EAjUIZeMzgHq$#Ui8QvY&?V;MPLasr<{+0D7!mHt_ zLk+xSs;~gGm^^l`mnQJ-h3%v>+sFaVN0-q3E8hMKU$VfV8*?r`_`&*DHAb{(?53Me zEgwcTy^fU^pK$*GFyN3gx~I`K7(7F%n|M*mrUA&u(w6-FyG6Ho+Aoj-*f=E8csfgK zJwYx!_>%yzQr)YeN-NOhnrbbjsp)a}NZebiV*?xt$<}-^EsSXRARZ1Ysn9j}@6-V_ zWkJ9KuIoExh;%F7y%$c6N_^2iayYz2D%`R@bNH9x(|2I!XDFnG{{R~CuLAv>Pw{2B zxcLgpg>l&CzO2@CsI_hJNVvv&b+0edeh@(xv2w1mOAE=lB*r$J@GF9nN?IG^?I@`q zLb4(7&)Ke|5~Q%(xVyrGnznqH_6wSD=ctN43Dn@_*p zIw9K|4o2Kp$v?Iqg_>^!>sNLz%@W7v9j7hVo-5iD=J1Clhd5$p*F>_jD#ngj%&R&9U1CtBvO-WoMRtO)lOJIS0EC4j%!h( z8-_4F$fzZVWP(5k__0ki4mRA!c(;ZG40{@*acr^Yzdijcb?vqxTyjTHDwLMY6y-_h z6{KX|rV?EYNh4KacK7K}NUN4Otx0Y2a8FJ#ROCRU=O9+;+{VQ+yySH6QpdZfAbL>X zXCM-KVw)J}9V+(`I!H@($gL}BHth}3t1UWr1RA?;#EcSsg#$M1=|D#~}Uztl0RChD{cqx2-J+9TGN&?0X+fd+Oy@4aBlwqN>{n6b}3Bx z=Q%j463d=W8HcB(UAR!7?a0RmjAPofXBccL!S$)Kxux0NO&6Bg8Dt!0sk~vy#sKYF zQbqvX$s8^ZYJkM7y@q`SeEWS*)zZGj1%Q$moPpM)5e8yaNF$yqV+r1I&rWIE843-g zu_P1PvgWzEvP3|X3Jx%TT6}I*js^hr=MGd+fR^v?imIo}KQ}r3DoCd>9FxHQFG}fKSsyiswe>E<%7bVG_UTvc!DQrT z)0|d(PI3-M)K!S?8*^c>7+?@Mts2uq%6c#*6Dr#8WLjp2RGm~8I_VYGcdK{&wA8NJZ@d63u7uz7^^m@suYFGk3q@xtI7!ZcCc-~ zbc%KkYZ((+4>7Vf`tey;_WLFZ+;8U|wbV(Cps+01IrOYs=#dx<61dJ!N}i?O<^|Lt zGr91)cYFOS67f~!GHv-l`B?S*YoWP>hEO?;k1PdZY6?S~XDz`erxaQ4Zx?2BH|iZY zTmn1sn&UOKNR^os5xbmMZG2Kq#zJtU^U}CoV1D(1Am_J1Qte|{$((Mmt>y5_InNz& zT+RDuE4vB?eznnRvI#ff{JPGoOIeT;25Va|J0hW96fkLz6IcVuK|gWT1a?l&GA zZaLAc1WIV4f%4hZYt2C;?C(0_{*;W%dCpIQ>LA<4ZBsOG~KQ}d1vdsWEp za#hJ0AB|u`bft5j$KI^Na?C~!-($vWM71+Ii+j=vw^8Yeyl;1|cal1E#tn0DTpR$$ z<~-H7?%R=^FRgBjcRDEb68Of_K;)i+o;4AZkGwjMPkO>jhFmE87<8%Bjy5*a+$DdRuPt%HMe7PJxM?iTM zA`={goMWX+cF=K&W+1s82iC3I{H!z3RScG57X? zHh@7S9OQJZ+X;^HM=W{_aa`53fl7jK2R`+&1?H4 z$-qt63-4NX*DRzy?Zgg;y>m9VJCv~`FvrS9e+tpJmuMg`$3E5985~lDj)E(#(EZ|= zI3b6BYQGiKh@wc?vkn*)z!lCV)(|%YuLKjEeQ9C6V&w1VfJv^nzUL(4bF58GP@qyl zAf5+WMb!Mylo9hD-Nkb!QZW`(Ja;Gg3S(-@FgF!Z(-qRJ_Bk3!oh#odk9v{^BWJy3 z-d|w`cqIMcd8~KUF_Rx0;FfL$Qf*O6ICdUjI3uyClD2}?>S)V&8+J0lV?L+ev#&1< zdvZqq0OG;p6_+lmv6evU820a3bLveZlaKz8uYrVu{ z9GcC0oU)zS9)_WL31U>Q%zEazWVyFDOxuoL17s3?IH}^fE6#SFx%Cx=9m53!Zh7OH zk*=CNg4n>%J#$>nHd-A~lG5g*T?ba`NjN#L)4 z83O^isggZ_C?T*2BON}KLNm8&@J9!cnz~3u!3P7A&mPoEO@yO;#A{_6Uy;smO#~7M zBoetf1E{5U5|S4xpnTl)rzB!EZPFDaagX7sl8fefAkOns%G4EAnCvMfqQOF?vb;l{W9nhwtX3rU8^Dzp%U~^PuQnELFzjq(j zwVgsUN%w{^$9l`Ujf!O85=KYAYLz8>Hg-mx*2Yc5j9Bmx0zp32n=|ZS2^l^6R;|ou z4l(m*9QLf6r3bDT9B0J&#K5iahzzvl7`b zPTU@PR^5bT62TpNYdVG0OOA;~ZTc@cimZqhgdjf;=6VsaMEMpFY z;~44%W9jj>6pRda#dNkY6St;){{Wp#-j_S+K{EBNx@|$(mpCJlTUM5=gyaC_KpMXQ z9DVQURBV7UDxvv^!994byBJ+d1|^6i1bSC1yE`eERq$#0QMfHx;-mdy3_QgxbqHLFsg)ef;lzHTI+WfdMZPBv6*Hup*Z+u>g!TzUQ*MvyO{m%+}C;Io9Q8+QFg$N7{~*6-Z2f9dpom-Vh;Z1!6S!@Dl(w_z}0O&4bh@w<@Owk==41m zq`8eFj5-sOyRYG1M5;#GA6*!`YGvv(&vpz^K*lFIES#Fk9tqdAC>6}hr`u)YnhG*1oaQfnHt%WZ7{+{@I+dn(tTYaTJb@ny1K+`$f!;pp6~K4`?VFVeNc)QU|r zlC38lR)?Buz8HqZ76|-RB$Jf|fl>h^{{Sy~lSI*VOD9oft-$KXfh46#KSFAbyW$TI zq|z0E6a$7}=W`F~S##@}9n@?s_d_ENK5XKOoULeSNw~gddW<*v-kB=3<(Ziff?6(e zJDk?7-OZnfbcxu-46)=4O1O3W>(2Zo}0PP2{f*I}1xfSmO8sc!F4rOMA+(sdhKtvOO>mE}MJ zhHvLx7l*8b_Xxwv3g;t{kUi_md^uwU+(K5BR>CzO9S~X?hg?eVA=RpJgmVM-8 zg^=VQYV@4B6_1>i6;h`&{7EftyxXS=ax;TnFNR=Ci>ZU6vVp({y>K>{E*P|ovB)`L zU6+J4A|?ZZ9P&kTQL3Xf-q#*sLHj3TOIooJTFQf+xyE~n`0w_h@f=B7qLOL*VRjZOru)rC`ga&V;l)ww;EB_O`-y!w&tPzzNE4mccS4%NRs zv^~f{Bw{>Uen6)tt2Lm(908sMZ?&En?gXjhIO$M5rsgAP3cLyww6Y&1%n1>biP1?_DM2 z2nG%}5OOOP;w1%#AP)Yu(H3(_E~Z@lFhB>f6$e5yoOY{nu10ax<2~xDC{_c|*2$S{ z>^)x}De2ayjDX9}T7?H*hNOvrZ18<5(+e8b?URGpb@i+E?VNFu@6xdmx*pljUbUrc z`*!olIj5*t*^W+I<|=yC=p`g>1Oj^>TEs~REtA)!Xxqp^9Wl@hRJ}-O+D6_3W3~<{ z{4#Je`1Gm~%m^otqp7PZ;EXBTk(|{feT5wIxESFU-@K^r#ApP9O3 zb52GikDH8R>+e-#KOxT^6@Fka4l~o~RAp1Q9AmF1ty`LA2k#6Hpw&6XUpYDb0H#q= zQW$^-$_M9FB~Y)m2Vc^!TsO_h5G91Wqps_bgO{NYdPaY;NITOn`&$n8^fjX8|t7|SWj=k=(W z4gmfTa5<}!LzCa|s0hbZAoVq^%qvM)fuzPkI3(cEA8FgOBOOWNqm(KfhRCRt;h2IL zb2Jhfu@4Rw{>cpO|Nkb5l-8 zV8yp*xTwT@6SM+*9w{cneTY_k0;B*z&myhEI3SXscHBGE3m)<^Gm+ChDn`lx0pqR( z4JV-0&3kDJ50q~tdg88$Jg|0Mm$*db(s{P%}216c(wzSssFw6=8*bcSn(uw5Mz0Rt^%!n{maB{$O zHMwCD9KK7T=NJGA=51U=5q5-9G6Cf0wQpI<#o~P6^&>q`(z@dwrz>Y)r>^~k9-{zy z*F$X*ov1!h!N47Bo73dl7nL|UAx{FjjXe2_^KC9b1g|82g<4iZ^fj#E*%s0PVoy$( zsoY~2Y>+#iY9x6gLC$@;jMUA73EQ;e<|K8gx?@PMa028KGI5M_HI1k!iE!kCTO*E! zweB#hxb8XSR~;)p=#@?eNnG>JD>m#}MjhZ1m-$3%o&g`{n&4`&P#GJpT?r*isvb;OSL3pjhM~{QCudvcPUf=D9KaYel^@`ueoxI zjE+F)Yl+q)Ra1l^`Q4MjslQffw#N;sjf}u&EHVZ<*B7cF2d)U@jC8KURQZ{J1QEf> zt~*d)^_z}yoDTG_*okU#+KZ5{jCJZO9pw4&dC2Cq^#H^8ai2=Un0aLLxxFhzW?TFFMsZ5&m&<|+W=81GimDVLI9a=9S;Q`TZh!RmUCN~)6j_o;K9 zgz<`xP`fTcE;jAwj1Wa#g@T1*103|M$mIZr$Qj^z)z}GMxa9CfG$xHpX*L2dI(Id4 z46^qbIT`I)_Tf%9ci@`83;B*WFl_PZ(vDzJF2>ur&Oq(!R$_~L?mPwVf2CH1c)-ev zxf<Y&vrEu{kUMLzrc9UdeXNnilZRo(Ek8B&xht-4npuit4ibzppI}xO>xS7 zPN-A4ryauyLu76o=QyeS&@;I2Zo|^Efg@`IJvijmoy_2s#{>XRaaUB}k@@q+g(XBQY)Q)|zh$s}}pAv4QDa^m>qX zTs8>KPB;|T)W9Dzag)Y+n(K==B@?8$ye_zoQMaxIQ+-4Cgpu;DGgyOBgt!MGPgc+L zsGCtLa4_Zac=YX>)k@nl7f$A++LG>*=FU&AK~(3|-N6O6_fwDOSu<)pkV2PVL)Nn{ zH6T7*u}?unrn(XA>}p)xz;&cv08X=Cf+Y zTx6M~mktW>3XXSwO22P(ag1Zps})B%86&9XtU*2r%N}|t?O3$VgwBc!gPexo4nQ@l zX5=7pdag2i*Ewf%k^wj%@y2Ub+UI^(U^qLEYR$H37~I#2U8})3Q^t8UZDu=%A1OHJ z`PO_g3Cw(|J8(z8)~zFMDx?F?KU&WAG)cY7TbnDKHr}VcX~Ef5n*n!@{i*}=6bFX- zaaI;FfI;AqipFob*$CZ|Our%_`>Y0iDyt(0akK?ff$566(3t_q;2ey9DvaZNd0qj- zWY;s5`ks_2c^UIYa#2pwM*cbxS=S5#u^1f*3VNE>o-#{dWkJFB6-Mv`!t5gefw!;H zi0F4lK1K!dyPm89o=;AHtz=!Dr(*)R?_0M}<+qcZ4WJQ_YZ~Wrl3N9RE20*N%?zprSkTsPCbq*sY#_{&y^K>v6k?MJPZNXJ?m!7F;S2do(UD50uTo$IrQSS z?H6dk8PC6J(MrZ{m5psVN6LpJcgK3^tfTWb5D8@fV2+;%f^T%(ccRvrwhgY^6caqDuaO?AQu38urL%SQ8ACsH_Yo_p|Y?`!f z5<1G*0|%+~u3ECSIY`wAqyPmK#(G2Ejb-Tsy0o(u6c z&x^I2=+Lv?$n7(%YmAb50bf7qpAvjcqIj=d5!}f35!^AEA0#h0#eEgJ1F zpKp<7P>B!9zh6rI9`Rz^>iQSB$3 zU0H=#D5V|Q<7G6bI*(0{F}4X9P#0sfV;JdOR+Xw)SRw-IC1d-rM^ztrkFc%-Rl1fZ zSkX_KIx)`!*0}E#_|j_|YujsACRx+$?I=8|cgNPeD5oc7x%L$rrmT+7$KMt#yici3 zb95PeIis(aZ)^_JbuNDS9+?&MPmlg9YMw3Bn`S6hoUvx$7T zv(Gss`kM2*n}Z|Ds5dhA^{-nEhe@>02Nzk%Su!jC02yCzD#Y#j=kca#-ZzpNa?Wyi zBQ>9KZ0a{inBC3^suNl(4dVd&SFajW`FUEg;JMUJ_dP#F@q1i2e>ZU*KtI;I9V^6* zJcWdfjE|Q!;1z@tWz`^Da zCRm_3OtFCI>QO zcYJph!R(fnJ%x*IFPQt=;a`lc?`_B0n+Uw0n~}$A^&boPhfcIVZk{xIagZ3k6dL^0 z(|$Hy=r(MY^APRSo^$I_#rs3}m_v1K1^j|JMpZ&*6`G@0XrQZ$rjOL^JL4CFW7L|? zNn?&?f8CbG6nnLHHl8BWH57R?>$qJ0@_uJIdd62j(m4e}i8W z^$!lD78-<8#)oJ=Y7kH3Sfg4J>>RbP~$IT_!Ia+fPTXv3K2&9ex;L=L^Zs;rjBZtQ&7`@Z$o z7y^^>jlGAZP`S5Q&Pf3EKT6NqU1~pMlR4XqRx6eSud4r^J_ZBN@d#kxI>6u#b)h&5reYDahaePpPc9obmw#dm6Hm?Z<4@14$bY z%y2japVF^e$W@7Mdk$+FIj{!P+dP`PZz&v>&`?#o@=nzQ3?TpgDV&aLWl~`wsHv z%gD&*-@PQ04(F+=jXvjMPI%^xcP(l%?p49dW%~X#l{*$1{nwM_VwEf`O&U%j3HS3wW9D9WKy67IRI1-ZW)LJp2D$5S3S07MDer8{eT~x zR+c8*0T{^72DK%HnT`{$ap_spC@s4j4h3{1b5xt$&zd5k@s36gMO0-Wim({18CPnY zuO5c0$r%lupf)=C3h#yNcsPl3F_z7OIXFJmp>7%Wj-IC#UKqf^Jb{DHYQGBLbvZri z+vPZxtVKfPY-6|_X09$%lB}n=Z|O}~_H+w?aqdlAw}Dg>oad5B=CX1rBtx@t2Wcmt zml&x8E(R17Uc6P9q9I5Hfgdm!&otP;79rIDln+GaR zJL8jA-X`St-PgTEDwaSUzG8W7^r`AnOsVCnfX9wA@ANdP127=2;lTh@(yrh#+-DgA zpm1RwI6dj5eMX5xkbYcaJo?qxix%Mj0KiWbK@=UFbJ+AXp=~DABpjdij&W9(V72!& z?S@nZ$ZYa4+O};;%7B?9l+F};m=fQ`2? z3cZ2v+O#cXXk4q0lW5M~$FQvHHjY_v*kAy{nzy({<%50haCrS|t_>Vj8okboPhic7 z816%~u06QLbQTho+M$pfckB39C1)c>;DhEyPe6Ot&Yw7CJxrO&&ls+iE{7zM+-b7! zY(8+uZ(c@gsIt2e6;u2p0~``Zt#CSR-U|)D9ycD9(^+5O;Hb#wCy;2lDmFViNeBzm z1Z0wO1zNVBF@^w>w=6TjIj(xz=@JpOlmX}uKZRVjx(aYM`-9fAax_T4F%5-Q$sC;X zn$Eb&q1l1YMX3$(F`(YtpLR$cFZuHqGkO45LQQI1!qK?i0V1WJ$S7Pc)n0ZKp%S|vXskEvoF98 zTo1;y?Sr!9U;;9Is->VfUUsQGan`Oxg@I9%yS-s5t7deta=ckYca8}lvilMCj)OaGP&E5R{-QNG%}Np z4({3S^rx&Oe54({$6Df)+>%FpA#H#Xk@YQ;ka64mDTyg=HjTZ?^r>Y+%0l3Ci~-1{ z5ClRIhUV$VTE(NOM#Tnj&9RecISY?kfXquTW#kizvexARCExwi+M;-5#?1BT*~V+9 zLl+x8H!FZcj2!WfKb<|4ToMQ&v)lElg}zwu7~|%}J{Suq2nfbAj!CUlZc9!(l-dQ$ zF4Cs}0tOG@YE-r=#~>avlh^523?IKc)P~2*zMhz=1+Zq$NY6u>*O0-bt1>2w5w>y( z8RM{~3sE7A{JW3gT2Ba)KrxIQWPe&oVaq88Xa|xy8qr3}Ln&Rdn%k9SRwq46D99ev zOKpReVpwnoV_E?+epblr!kcw!LwScBuLnPO@T{8Imle>@*ti9c&5Vpx#$ymxC#C_w z{#6|0f)rx_V3X42d1Y4;ndY@)AzT0*yTW|P5%HGF5gVnLtGF>)5$&RKeP}@CxOZ0 zx}xv7gQd->xVN7%Jn@Wzc%_>65%=WuQ;vVavDSA3A%M*iT(-;;&tfXMk~W4Nv{|^g z+_)LR?mcTh;^mP>I+KEPie#4sg7bq^q$RVvC+qy`Iqn};<<`SyxL`*(B=f~pdCQPM zC3*u^3i*wKMsZM>FoUldtfeyE8Si4hn8~}IPfU)K838%={u5G&ak!8PKBkz^6-PBHbV7BZ})@Cdl9L}Adq-qMW=)j(o zZaa>7rxe|fOmIao@)Q)>PkLM2&evna0J7?so4ow+O~2ZA%F~d z6|D?}4dVcG#&KNiEwuv%;PuUFLn%033aHOK=B4d(rYz`fW;@iL2V5;_L$o*pr(Q=& z!_#N=1&>YzZCj}-!S`+?V~(P^DmQbkl1&!8hC(tnhxwNWty>WZ8+OyP0D^lOiWu=A zW;q~Z2cW9~RmySqPSKOrux{^D(xn?{iGr%J0JaVR2fq|FOUYawoO{&E8n9jDs0WO5 z%`cVaE*KIyiFL&~cJiqWGe;}|`S zUxVg2Dh4}_m0gr8k{c(W^{2tZk~;d1wXZ|YlS^}RPsacp`IXEAsZdl4s&RaP-+ni>x=CRQUSy=3CCMwFz6-ur#*EQStF6kiD1kdvb<7OBh z++*In-pcDG9hfg-E3fd=ytfj>fH`F)O8qO&#P*YC)C(m zO|9w#8;gelwyEkpDo+5~X!<9`h_sy}Q%zFAN}R@Ws@dsYJ@Mzn2TjpDJ*8V*Y;?CW zuuP4+k39Ob0iv+rMv46yh* z)?%$3NfM}4{u7G%N5uNE)u)hpZtW86VdFa>mCDk_aAv&hP*I#k@It#ex9%GUBM%CWL4^2_qEKVJ3X9})aN z3{Xro_YZLL5Z+{1(3_8wdFt<6`!5crPJ$!h|^5&}cH&g#58 zlXj7&F#0Nq;^_7<)1zS?a{bi4A~Vd0yn73Cg0 z@O*aSQ*CM)Kf8{CzKai58#H{iYE>q$6TzapxVw{S`J3LfZ#1imNR{w9bgfgQTH0JJ zh_0lQ@|8SOHB0uIKPUl!xXvrKr0&#Wk2W->z9w@<>&zQ+0Uc`Ymdz}iM+c^Av>I`m zemMkZfnCRiJS8lyvJ^NyF;O=dqM)TI9Y@1&3O=)}!D?e*WII*ykUi_a_~Y;^RQPk@ z+kJla28&IV)-{?!K4v`t?_DRtJrL=7&8*W#xCC-|CccRMw)`1$@fY@L(66i)%%4fU zN#w(FS8mdA?O#7RIag77CVexbY%e<0{MNm`hTR782c~OJ!(JxQ?5-nwh?ZgxOyeWg zurPdHrt>g%J1*+GD{vTcH8j?g|U>NdGO7@-}pC#oh9vrL6GEOkL_CLWbQ^1}r zF`YKz3oS)XJej}+z3b`td^@J=npfEGE#sErB-*ls0E}0_{vG|E{6;PzyzuF0=0Y7n zJfHGw>wkq8o-($(w$nUEVD?uE05YlKLEo=>`P@}G!E-`aN6_KrPP}bWJhDA2!5V$m zjU2Z7*xT>GEP7X__&WVRwV}u?H{gsK%!@>gErh#BS0J2>l4|dQwYW5&6n~|lYe*61 zc_hHv6cd{96eBv3Pr5s1xp30-K0*Ds{x4hj@5fgD8`Cb@TWtZ^B`J`6=KJ4Va4Y9a zc-7_H7j`&ojPqOmGq@Vxj(kCSr0lrVCzTn8EUd(iew9``P(UQ$XOKJB=#heiR98dg z=u^CIFk6MdE4X7JLEsv+umO-aeF>?dhht-5Voge@OXOrJ>bU7yPhAg6FtX7Uq==Zo zDmLSu^+s!Kh7F8$1PmJ5bSl7S<>`aov##R~Ng#v$-1Ml2W<;TNrsq9z4hd|K-9=~W z3PX-gFfwslcQIhG9Wj!70a;gDyuwQs0AMaFqAOc7&yv$a0`4Vk_~4VCD-!w)l6#&x zuAbr^U!sc2xVGTuj>5X)tZ>df&Qk6KRmLkC`q(R$UzfH&!n(`bP7ene@7}R4ZApB7 zHPp2^UXJHAE>(ye{uLRQZyoCG>@1SxV{K7khRGQ@#TM=)(?gZ;`A{%>Vy8*E20UavaGqJR8#z!~@tvAmI$8Ip*rn6S(dj%j8Io(nMSdG}mdsMbG zwa~Qh^ENPf=M@nrXE+Vl*wT#jBLlW++Z%IXcXr7YJjJ9{aeD`UQWD*phH zv<~DRqrD-Apx_?-cC7h}O52{iup-Dy4cOz3^(2vQLA7y=<0Gw23aK0^J%@8v1Ykr0 zoOdAq0807UJ#{|DCV!R@RfbgOoEl?Bid&LbIm>kwc1StKNM13X^#rScppn$@d(<^^ zRZ_-Xzzkb-g?Yyuiq4$}@G0PTt(jp1Y1+Y$8OB9Ynj#&E8+q@|ZyijfE3HgfqHX63 zxZ|d37K!6vNEjV!N<{`nJp8TLaZ&jtw+xe>;=5rd%Vc?V756E_8S_3+AD*VPppExn zXC8o3+SvjbhSl}&T9$UBAm9LfDavLrhjIvD#6ij)PB3cB7Rtc=dY+YOVGIhcTrVJ= zm2L}087GCtKt1axE>vq6tCWfaDpYJeI^KOR@s3?ZOj{|A6je6%MMujcCBM|pq09mm3N#4+B3mDjZ7Td zh2&?U1Emg%xoi=+dBFWCyGGRryNdEN+P9C94?@CRg=O3b-Oejk(&PZS`Ac!lbLltD zoSny@J!#>mt*NglkK=}jEXKBrM_K7Lhjn@AvMlUiPJ&&v2-p-ppE_hb%)jA!39 zsc&$lfZy|x#zucy+6^rZd8EzDORz|0%C11l9Fbc#mv~*EHv@uCYUJ&%)d^ly9tZ=a zYFk~ZbS&U5G1UENy_v|<*z6(Hn3V;C5(3~J4R7iCi@Z#O=TpHWsINbc;EXWZ2>{_c zR>iKMVZ#!<@-x@zQ#shu5xMA*>Zlw2VP*hum2m2askvVgcV6|v+G-8*vk)+O_p6Jj z61jE&=K%MvXIG)rsdYzKbz_6Q_Xaos)MHQ|bhD@($9m?H{{ZG2akr@4dQ;n1BY(F! z>?;>lJ&uUbwye<9ulF-<8CK-(>T93W@77rg;}Nb7(0MeM8jzA)mH~G1$DEqQy}S$@ zkC(pT*06PLp|`OqY8On%Km|j8jdD8F?96gVAm`;ZuX%O1cFdOk@NO%aylv-m;m@h# z6e^mM>F8l{V56@zyP`H50M@6$EePIo-|k!a`TjDSxh zRZ|f`#{|}#(G|uA2CGQ{$RPKw)hisfxluE?o^U$iuEu!C3f)dXx#-ZNI>QrO$u z)}_AX#y2OoQMdz>o}(3|3}yIb`7_RHtP9_7UbsCgM$|R{UUB-?4ocb`5rQql8QNQE zVmZcZQU%_JTYjYQgd;1>k2>JO)C+50gCP<3;h{{VN|k?hZyS+GVnkF9mW zIs;1D7}45@+l-dv5PMY7TL7DazDGqJPrYfrn5qHJGuH<-EOv<)&gO0Ff=^oPj1{yw z-(wZ6j|0p&$5-UkK0z2_eq}iC*Veij?a;hsc{uE9yw<9#f)3z&n%kDgGmiH%Je4Xz z1a9Q8^{D)j3S~y+CkNWK8VL6StQa1dJu0k_5O{uZ&U;bI(8kTPJwR`|a=`QhwGyYw zdiCTJ$F*78v4AHXy{WRqBmk)Zp0y2Fn;7?TZ99nhk5FqaXFJFDK+ZANwl0yQ8Qe+E z10L0jbPi7A&Uy;w*xPhQ)3GIY1t;FA&u~E_KQFFo#>0RLVD_rBL&*$BUX`RCT zUS^t6NaSF9cdHS{yr{!x*k?6nJ7#dqbA=zBV(NL1 z>Xt@hSOyD|w4R63n)bm#!Q(uiTHl7)%0BKm2cBvPtiWKZpd90+VeM1VoVBpJXkope z1FjD4N3}%Ie9R9x;8x6#1blFKKS5Fbm}C-04;4|Ums7*7Nv2}S30=5eIvgAfQ+%Zg z1_}1fYBUNTrqPe$=~0_-ISfh9euA}g(DUU?%zJa5nFqFMf*i;&JNB&^pvxTaJ!&aX zfB?exrFWpcpK`jP>ZI{jRauyvg(TzK6&fe6I6ZS$BV01B0playp2)Z+iEY7BqaAvR z(zYZ5yLiaQxT+}@7+_$N*0inA?H?!^;2xEZ$5b@f*0cw^F|=iCHJ+fWi;Bjp{9PSHC2tIi202CQ4FqwR1HpvNMl zMBN^9oF1J&3Q>c(-3hesb1MLa3XLL@)EdjahkCIpzMFg3MR2FfWB?yvFy1mmSpxKF*2h8;z0%k#T&k8I|%ZURU0?I%2QT?#F96sHtr$THa( z4n_r4Dr4+elY{G5r;BL-4X4tokT6{Qq;&(mbwyb4aT3167twmMg&T?_CL99z5q_{i^Pe4cG+oPg>fuW4~}a7u7^=$aftwbmB$K4AlG^D{`_0&dzrad*abZww-tk&}O|~!u~w)UyUc!Kd`N( zeOls1FAQo%+#XANSIf_pLzV7&Ri!CZZOb z69dx~*=XKBiYT2WxOKGy7-sM7>0FnKJUMe@yN@$bi+oq=tTHPE} z>Bco3IsBRQSHiD{x|WeVQr@WYQAXW}0nd8&Jzmnnb84p9T3x{v*nGeqPc`HI3i!=- zG|0{Ektoa9;j7TRIjGI9TthK(pyaG=*9(5n#E%zg>#-s^rMyD!6cr?D}83_NM$SE0OVq@=e9h^GDdPev0cU8uyoy( z36G&aTF1N7cP`Eg4x}~!_O6Jek20)v(Bp0+VqCLyKXd`e`qowC860vm&$V?E*tAG- zj1l~KkMW0TZvlbkoV0^!sYHZ<#yxOzP{v5l)}9!t@>%7fV`2~n_@*c(_kED>r=%ET%LnD&vETq5!i#Y?8ztct1(#Z8R|*>D>mC^ zkzGniif~GvIK@jGZJxNu#xq(`*k!rOJu-xKcML$N*HrBN#XU4WQujlULr?_l$TfgOi@7 zm}A3W-~o<(Ysi#reH?Y%&XPA}RaOdm=dDzDlZ6U0asea0d)Bn}`{Ya@9YG}YH3ZQN zg}^-I0(q?UIvQ3orHO*$s*lvw8--FwCzGB#R;|+R&K!-Sx2;i?L%&YhBONPMqtxb= z?9rCVx_mLf9COm6XpAQWmL&A66EE)w+A*9FO&Id|{{UOtK9$ve%X5*dj<$_qi*oam zoB_>g+9ETn1!4ilI#n1$DH!HG+0T4dy`!qixCdw(y}ufl?&>J6XJX~8r3VVNz zUKZs}K3sGItuits17HUu<{dbzk*H9+Q~)qJ?eAQ!S4UckLq)J|cJtUCGf|sy96{HN zpI-H3$AxX@b_m7~YOiuKqJQ1Fi)dzhJX*&&)PdqYiIP5^hK`zm>5?>(l1~6%@D@eo)k7LC)O}oe4 zV~z+tD|Bb2uE~pUjAghTx_&iC`TY>`;PUFUL$gOWUb1btk=QKz8jbf!NgukYk?R>q^yComD&XK^tTMSp}1hM}KchmJU?;W4}!Hsc|)raZ)ccc6sBF1_fJ*6t5~nV++S3nQIELS0sk} zz#en>RrsT3P+1s%4oK_oU2sY47jJU9TkK^R1@q9Jl{|K&le_{zBxjnr6kC{J1}CTk z^{JwSIr9-pl6VAHZ8)c(-iWN0XjC$&Rz3QRQa!3C%GeSffDUVBI{`Eyc8#RUYc**NsecUSG422yAd)5B{1$-p&SH!(h zE&My7UFvL3!qFiL2<}+$1!VPC1yM!&%=4F0SObyB019!kVmB(}6V|@c@NfJRk6V&B zu6$wQD@_(Q$!j}d=gJ4va4XgH@7N#VC5j-ry4N)~AZ5(LF`r;NR}ArVX=y1M!OB}j zv-4(4Mv&owByva40s*tSmFygfDbLlRB%E3s_Yi{BP`yzHBypmQ_`nO6IVkL z*tA5PZqM+JhOQ;CRRxr&BO{8kw&@TU9JfqolUCzbP*s@Z_U($~l&)N;+eT-tsAm{o zsP(Akk1ox;fCoS`&uY;St2Ws%FnZ^jt2AJ62Os@@^!%tvb3)a6R&mZ7PTAjl7Zdt)7as;t(Wu~j(fir(@GQWbjShV-g4Tk-|~W7LYM zRJ+t2Bx0;%Yi9+!FR80A+qn!0QPU&6IiV2D2s=)C8h%s~z>?>jR??Q58OAE(qWk4Q zKF2kqV*wzjDoE~sol}4olXEFRc^p=~ji_{GU>gTM#*=o`yIk1Qtu`nG9Fd%g-Lkjb z6fqq4AI`JvVh#WpKPdy80BcgqbRd>bnQ%8_h83DgyQ01V{k&lYcZ`FA2+nHsF@kVI zV0Il1MFdQNTO;P_fmb#d_7XOdaB_J1(aQRpMx;v%YRteMK_C!MPs*%D4hohGs&F>0 zKRSpIGEg56~Mo-Rh*rs}Ynuden_3)3rh4uJ1~L*p?Ur zo=?)cqfHr9#Wb!}kQ^VIB%j8qPtmcH@9S44ak-C7_o{O-DYd@n&q~!o$nt996^in{ zR08*_rz@Ht*F$6!0vSshrMcIZ84JP|i$h z7oVkfUIf*l({=l4r8^*286nY!s$Q@*z(f2Ncs@_=qGPkf@J?-x`7`&yOhu-4{ zZ|7WwhpI#HXT(-M2C-Nt@dmKUbqT=B#Tn!-duP3R7m2y=@$yN4qqnVc+I@uAlCDN0UUB?}E2H>;%J-K}3G)kXVbJZ)Y99!pzY;Q` zP^+AE{Oie+(D$PrV_M5du!#-L>`^nkMOCEezTkS-i}>5Yu-$6He<9u{a;l+nK+hHI zx`w@DqFpRf0e09WfLw&Ga(!YO!2bZdMm)A#qF1Aae(tB4ijAdYc^%Kf>sxhqk|aT% zbM*DEE%9ZQu8_v1+*^wI)5Ln*v)q)Bk{FZ8$jy2R zQfouc<&8aL$gg#r%zU!Qfc8DBHeGR|I1ETU44TQhzua9&&%I&Id}MV19faWL`PR{g zwCHlqx{l2Dzl7ctzwvd`P8I&i4-Y3npG?=gXdVudEkY-Vouq}qU)H>H!v6rY6|cgd z2>Yo>M{R6}oItARq zDazqXw?UkAu2WOiVDTlR*vP_5V0dCVW7yZ0{?s?tSH2e>Hq+%;H4AB%%u)B94^S)5 zej9vFmr{};B+^SEATVRW`d2QkbH(V-S`^^brnNoCz@HCq#(lblBjM*<+lHxr|_r+*MzMUuV9tigrL8oRX*K&#A7L#-b2NSkDSa2a{Dalp9-| z6Kz#dJjYm_u4eu4FOKHC$Kzh1C7;8mMPDx7@^p;@E_X0Jv*}*5saqwjmzuzg^=^cF zSI=J^wTU$?Lrv11neDAm&H_GEGY&ZQrIu6nn55rAvDGAAjAa7Y~Dtt5Gr9z4R{y{qpUPebP^tu-WyH4(AK!a4c-)xjAU z2dNzGJXC23+DmO!Cxy*jf*kyyj&Le@7UyDi(4%u|r!3f0+@60Lp>b?T+!?ob9f$t_ zTDlLiorQTqI4XIl-o=bzytByeI5lyN-i9t+4n|ubQI6D*aNL}eS$FmxVK0!M-u~@& zQ`su=zc68d02PmMVYvtifH^y=;~u9SN$7BQ)(oVwJu-S%cYk75 z7k#({=NQ2s{c7i~tcoAFFiuL3O4>0@0Q^7$zu=TkM()>Y9;B%1PjgszcDo!8yTKX7 zc6WAT3<=$n!kWdrvo22nHts7$=5S3U&MxxY6+JoXFgbr+7q#n4UXImqL@;bPMus$2c{iZ2)1CGEXBNs@yg*ZouFkySe-;O4~~y0#oPiJ626IXp|$d zGBA)2ls9TwY|9q;aqZf=7&P;RY+xrQt$${r>%+Gok4nN`XG~q5hZQBFx5`rouQgSy zvc!@Ge(4>nues9l&Bj|DPfE?X)5Kv^F5{o2Lf=DV)6n#lMJx+rBckJv#-frl=Hn{J zMS}Q^WXZJeJxHu=uBP{9huSiP1R;Rw z>+e;jknPK5heEYZOO_-O!47a#pURoNfh^21$i{ja5@T|G2(Jn{H#lL)6^>*Cft|0O zok6Qta;`{kn4Yzlb1%z)M%-kNdg-TRjyjFDI^sY<-MFwljZ~iSV2Fqd>)x3q!hlEs zG3%3A(_M@M>(aYaE^|7L)-+?eGGLyg)1juxI_@95W<5ykwUU>GT#etmCj-)~OK`4$ zu0rFbZxv%FI~N{NB8=lW>JL*)Sy@2{cHrQH-l!$cByZwV>UvX!%W}EO`jJ}M*yfb( z?r8}DY;+y6YEe56lpUwmvKn(2`$x7#Iyv_S^(XPB-$6NP=4~y{BoZ4vYH03DmgH?t zfL1z5xZsjUcInMqhH}R_7$X@KjGHxG&Dbt-!w=MPjw({Vbg|trMfpHpxKo2y1@2?qWXU7~#Cumc4bsR6-GD#5YEyJd zsSHUx;dol}=dP!t30WNgy1-Wh=OFS;HLkfNzV4fVtdV|FAH9OZZ}~LwcFGU?Lv961 ziz-q^u#mG5NyGG@OB#We`GDjo&1XkBDgi7AIQPe`J~;%8dEcCZ_{Kn~IgoyqlI(_ISM}RdWx+yWl@R2Q;(H;)g^5K>SSEJrF^sVHhOzj zJ;Z#-8IA@B2fcI?p+mcCV2-}^m3?j*uvb3dR`8NqBL@rW%u9%hHWZfk?TXF0;FiZD zfN{lb-Nm;(Fgn&%+kh~71I|0v>Q`W?-B`q)Hpx8?U`158jFmX$vz}{Cal80UvhS ztU@Uh1D=GAl~URun|CP(8T!_}ss>(E9P}htEar`42Xghg0kn+gr?zWa6=f@u17oNa zp8!z1dFQ4(R)j0T1&eOyK9$DzyGK%1*uMxWyfT3;m ziMESI$)Vj4QUK05&ow+CxmH26zm3^*_q5+QF6t6SQHDI}b{<^=CrA zE`_UDisK5qdNJwwR-L`D1Lhq#;7X zIKcbdb60HbK3;mBgE+=2ac&m_A&BUxG7Vpb9k?tG?@V|8mC%}q+ij5jn{)#h3xkk* zQ{%MH%DCH%f$oZgz8UCx|5#=ziq9{$ytJ-*B% z5g5nI#den#8&%{Z+~9T3eAXVJV;eU;58cIMQcGizG}*~qTOG)%069E$&%H^fczWwr zy-QnUGRZO78b+myx26Sc*yuBBSCLvlZ6uS(L};KS&1&S z`4loMDBElZ=s()`u4MV7qKvE8jVQLyDgB@S0AR~`Y@pUWex-w|v|3^u61_(A+v#6d z#h_{44bs<5(X{)0Gf!2*!)X{nk8|F;VxS7-yk~y4a1zG@zrACauVqK`JbDwATFaQ@ zb({2yA#)}_yI4)ADI2)ZR&e)*i0Pt(zs&b}& zg@>aB9a-4TsV>2>@^SZlD7V$2WmI-3#F5WR;%+s+^k>@W#IVGG=U^RdVC^kNhD`FWX0y=;|F?jf^r6;A_s{*RHha?q;6j?NO^` zQZ>UM&rf>MkHqaVx!OQ0@-aPrl~I$io5QE0cPsec;D5uPhy_34An?_xx++jz*^eeb z52;?JzH{-%>>Y1sGf&{Z614G2|Ppa+G7X+wAu~a zncwT=xBkj&*`bT4U)_33vDXa3OWm%t?!Qyy*?vX?Y z1J{FAqFt&+Fh@h{QV7GoK{)pnN!hC(Jz6~1L|+G0_C~;!;(EJ2#>_cb;%Q>F8@j^w3(G_K^Ta zNl-W>)~%o{512=m4$vQvN3};|Zm;vR5u6>k>sr?E7GIbzm`@9i-qlFwl6#@i0;O}g zmjf(4eJb6mvxQa39Fv~h(p$mhuq>nyaDByFjTxV40EFd=oaEN(E$WQ<`B0%4N=|ux zrN5Ovh_DG#>W5w&O6qs2SUVZa=NGnCn3oh$EHnjTCAa^Wl1cqEOMBFf0knX)n%qi3G=5tQT}gSn_=mBTQ` zan#gJxP#6ITmxGz5zSW;+ZI_c#Db(}t!Lc6=G~GGJ$R{Skg&iwC!xhsoSp`A>yGtp z98;}*2t(~|a7TKU)_gfU{{V$ll!hQ=4`E9TgMplL)7GJ@oc`6{Lu%RZ!~#xx9@VRF zaur5-^{#p;Laq-^pGwoTljho5uxi&eg-+(}yo!yL{A)%j{Kwo_k~;ChtO#c1esDqK z+O*)gJC~4hE1FBG*9pCjhfkZ#UBrdwpzB(&-T7l(ABAc` zbO|8ie`asxQDke)bz|8+5o*6_@ePfw&EDu5;Z&eFPPO;n#vcz_UTHc-)yGzykP~UyU8fp zinuSl@1%SD70!5Th6*4)x@E-;A6SvD6P1O_=`ahzA6&O?$SRtRx7eOio82 z^vztc)Ggz@k91MUr;gQx2so#h!WKABP&pitm=v(#m^_ggl^LN53`g zfAG&g5xh4Aymr}gbH@_DznskX{k|7!r00RTkzSg+A)29*7 zaETyS`4n^m>s&Rl6r9?WO=$Z{ZAmlWuO9pw@m-g~=C_kjlJ%IZw>z>}sOX}+Q^8uS z+6JhRHt3{{z~r6>(!W%^e`#f^X#PdKD6vR4Fj1WPX1_SUX%7vr!|#Z&T*n!T+VPG+ zA1GiwYnzr1Jdr(!;^w5L-;wr5?Ee7cBXes4Z`&q5@o&9~_N`Y?k5c{XfUHhK9)i9i z{hs_{w{03_Re{Z0NC^yuc)#-=t)~vT7uur!vaPKJvcROEyEUIz~?7DKPqj!MT2~~JP@Su^sNY?U`_|$ z`EUnMrB$>o&ibx|P}-iag;T&EPW3ZepL2nX59wCzpc3JqiBEEKfmUOTv}K6M?UB~B zidz};X?Anh8hWuINJ4S9BcG*Y-096Ak#=Vu_14R5WNs5Edgq)9u5{K{8$gT>Mk3_V2Jd58G&_`Stk{A$=e=!O*a3n-2N>e5 zSy(X~rac91S?Q#($jXkZS?cs1%6583$_n90HKw{6BV=KKUT`@TwPmKFgOz4)O!TcN ztW>xJG84&F>s4zq4qbF?f+HDRXC!k{!=^Se7mk<%y>>C_zGyps>afokCay=LnFj!m zY+wp3n44!KZ=xPJJBB{#&TC%VOaS9b{5|p=!+Z5U)$0=INrz&p8CQZwBC{`aHDcU>xMPLred=6p zXr}r*(xq8R++*0Sjy|7SWRj`e&Vf!q$T`J6TWnONrJ0j+2`JeL zRE%@m-nsom0rH*_Mo?NI~dY`_Pe-&|C391~X5=B`fG#z_<_ zp~tUId(zy6T(DxLxgScc=M1P`8~fPlP}R|e`kGKM!N==b79^`5;a;OPk!w(10y|8y-Pf~8*#=!1e&FG#7M^&QZbQK zlf8v0b}C%m?_vid9epbR7Xe+taGaQVXg-ErtEtw%$Ry>%{ zhO%DdWRrjaq>gdP-Jeh^MdoA7_cs#cGXkVGe@c_>ISi+v<381kYT03ff;~-NSqB4? z#a_&t+`Z>-Ae`iN=9MI8^%>yRWFvNeg!k`OA`OBw)}gV-eG8DW@7J~}#8T~EIQI0a zXyq6yboK34VmatZ9V<4wol#v)B#=2=;Xv*^YSc5~$2g|11XrV>{* zgoW`N<&<>5{VK?l_ln86Pf&51%UKvNEJ*LoTS-U&oa4A11w&5S7b_ZVD-F2F1-@cO zTAn5j*ibus%luVB7Y{Oq49a;Gb|hs5m*r9U=Cf&AaNU+d?@R(02dMR;WWm@8*u-NT z^Gm-ms3i3Hi*PBx$UwjvX ze=lfHMdzB#mLq_!J@^=;mr~42n8-#2K$khobjB+ZXo`|sn(4I|h&WM}#~zi5EJfLh zsBHDe0=gqDE;g;Pg?STX07xBi=~U%Rlg>}AYE1{4CQm9#Ro%)cmd z2SeQTs}MoTWOM0Ln8rZBBy-5CFhkg!f-|2=QlDa@1h*x&fCHVtzolP=2H;c?pI+4z zFe>aja@_$O)#Zf6ka~6eYmu~Q;)&yrEDkWkjyu)cbSnFlXD0`aDhS3KAmlLb&0c{> zJe)TeQAO;uFt(+ec%M9HuN?HPSX6mqIV0|jj&oEk)6Brj1#YL(v>_o>pku%TaOa<` zbG=@M@zTtXL0V}M6IRm^@t+&dgV{06r@TwMZ2UQF?CpjmN!nSQ8-c$^AA8u8FPf0wZlw z2{{KHD&%TI0tR@`N_DzAC|s7n!KoS$tCr3W3%9*(8MB9JZqHtTjAPcNjsv%l19j)US&G=m zFzloNKJNt7?;q~*otgY|TBgy2wMGrvD=;sNDLGTm(y(r=yw>vA7F>dOuCD4eXrDhI zBe2bMKMTAhHC)*2f0KxclMqd%@2@`4( zWLrkq{nBHp9;Uvy)AcAV;zW_c40R`p`L^H2vFV?>g_F!Fw`u4PZ_=`Le~VLLr2(P? z_iDmSfv&DF?#h(;O4XyElRKgsbo`SrL ze0;oH2rZ?!2(I`n-7r6hd*-;$i5@&UmWcOu?q`xW**uQ?xIW#5dNr#&WL3^eaa3i@ zbGrWkjwF`pz@Ng~`DL9aH~d_fFn$Vp4cPVBEw!l~PM zg`kILov;Ig-oAcVX*Xnj9vc|qw>`T~@za>`JB3_)tJ6O9B;F%^Nm?kbJgxx*^%de) z{wZ15qp4O=!cGw?|usah=;7W1fcwvi41_ z1?~BsjP>iE6s5zd4Y`o<>G;q*NeogeN5=92!Rwypyo%q%%XCz%g;)%@9A>n%`}rm4 zGRO$tI30K=v5r0ls%J$QTm&&*5EGnc@hCltCo1I~Wnu*1S^RRVb;r zj1z_kJNniAA6$`OaEHrX$7VmfQz&bAob#Ht$6fJ{;Gc$mIO*}DIMO^=0sA(tploB= zCG`i_wR|PxPX_CM4?IP2p!kx-tTh{VDh7U8hi60lN_}hYD7;hV!xH(8s}tc@Yb<)rfLz0uAq>i>=<@o^{TN94tWEwdh{ixgK|^V5`xTt zZpR>xdRVgTS=VR{v}Tq+kTF*vx3+Op>{W~?8-eF1wF%iA?G)x^MI}M{9<`r5c~>j) zYE^n15%;T3ZEj`CtWmXv*bRh+ZvDk{I$RAKs^>WV@T}{5W08j2*;ou8t!&xA zRDi_gvCq=FY1y208&^6=5OJ2?03Uan+KBmcec{01f!e5P=&p(plLX`{j{WOWDE|O< z10D(EBbw}tQ#hwS#PLL}mv+j5kf*IZBl!xT01?Os2A7b)V+4#6M<0j1Iyknr6kr38 zPgC`+(z4Lyl(iim*ZZev#sK8w=~EmCh(-BxgOks;YBHsm0H*_tgH{XyTzuP?`=^T1 z=1N6Ng%V>9!{x?13ZXi*s&YwPxzA&Y)|DE6m2yh;+&fl$;kJMZ?NNi8qLkl5lhwx6 zA1p&w;p0(ZU3>_7Mj!Egr&#iHqlAJJ;VQ#-K02*mK6sjDY;hDHO z2lO=*t%U@x78-eZbR^_uHz~u3YV%uPq$6n%; zmtt0ft=kR@9AngVtr#T;%HP9|qqo+nZ3BV{BehQ-0D+(Oaal(h*ygWCjii!$U{?)G>~_KFp`!)cqbbKh)OM{~SM$^}WPm}(^{g~j zVUT%Vm^Iq?Bg5~i+!9ogox8}vtf|MD*ttzoi)`8O?br4_^_}d5t4APWbC5-Tn0zkP zEj~5;1iH}RmpZ1Jq2Sy;eg@=W_2gIQ&y0LG4wI!UDnncqOPg~`#7lC9ulo7Rp+_Y{7196iLN7f+B=w^47uf7zG-|T z=f0ORF!MfL%5%uT=DkbyqOsC%ykBGDtqORj(ELMcw+SMB-T35j-n`?()?pGzeBH4W zI|o6GbUwB4c`bFK1615Ju1U@Z>sY#^ ztd}8x?ZK;-*Bi(U^LtfoWM^k#(>2!x%RNt?sVl7+6j$>}wMRJQ*GZ+`mnyF=&#P7* zn`paoKKE~W>U8T{8~urK7(zJSI3l`VD@`;yQIp-BPLpw9$Qn!Z0YKoMPqli_g1jN4 zc$Y?+Td*|8a$7js_|7@$Tyvz} zwB~yBlD90N{ZDH6ZQx6rKN_r-^B|Tc`^E8}U@PkAu62tI9b=YeB#=bAw_?5*(R?TI zBIe;_wp)alu|n$DC9}|1(Ov?&_@SgmuwJA$(Z)n;6plBOjqvs{{U{ide2$#&Vh4jFcQOd0gsKxJxzKK#rxe;OYo(|`@9zS zh`wWmeZ^0vYvHYX#9F6}@6q)ehPm3=RLLqU3>_)Ds-u4DJo>Mb+?MS1ui5j(kEiRg zM$(AYn=Hqj-rrjL+rwJBTv)>>-q1Q{4au+0--Vh~_WISU0h~BEBc*-i;3?1cTm+UV zhz7&l*Dfute{-qyIeeOhL<>UcQX-qyUDH*N4# z1v|ELyB$Sp*~R8zh$^Fy0p!+wkMQM`xA9}4_N|*VRA54{CzD>nS{!Pvv}OBcXwG&P zayEc_)~qN%10NwAcfqQ5@ovNX*~UQ~1#CfRp+Hl?&@RP}^GHGPCT#eZ$rZH8@a?Wj@Z*ilP8@BBXer8dOn#{h^ z3hr>lf#9he{uSu+>E(B4x!{r7v9Gl3?l>TW*A=CkvS$>fspXe?Sdqy9ry1MRHO}g~ zF6DIvQ@(MU^((z39$87_8>9iLIkLHz4p&q!7cv~>WIY4VWWj0m($L^mAspO?S2T#rdrBlmmp z)vKeouqP}47ny+`j;I?EqshBD=~pAORwsZFM?=MF$!5|F zk({5_og^WK7y^3%=~+E>HB82PG<0Bu7Gd00MdqHO4i_DF4_fM)J&*E%j)3F&)^?)^ zb|8*}0FHSzHtJo^1iZchN}}fJlfsdDpmNz4iSw~e| zcCJLV)cc2IX}gpFte8?WjDD2KRdQ8^SMOwZ_Nv>-ae;tFIX&qa+{y_f<@KcWVv~wP z*<<8#tOr6X15}Kab0ZMV$jRtED{|sNxG8*r)Q+`{tV$Uif*C+6GC8Sw9(`9GQON2X zfJZ70Gm4JHm74p{X9swq+qN!^NzbU5V8kQ8IMeB=&AX-6w0V`v*& z^A(?Ehh+h>d+}PaL(F`Z{3&U$t)A-!+GNtW=Os2O!q<;}S*z!RTui`GZH2F^;@fTy#ey-OhI9-;hWmv*eSK za0sjS*bdu!)^U#;h|Whitrn=`l!%%4V;qA_kLGN5^r<|P+wsLo3_D3~K^*cb*Hb$- z$HET$u;#R3Y_=G4k6Nu93?FbjH)B_$flG1MKYF3NmZHhx8@BcXAc~zhAhvP_DX~T6 zft)iF&0JJaLmkIC91bd5MzMp~M2xZi_pqx7g#f~2wm1~^iM+65z2}IBg)~M-vhlzINd#qFbp&)!6E#Ex@}0nRHf z=FqQU$m-S7%?M$%fX6(X)_ue|AQ8atR;QrcTGbr1)@2y8jlhiNqI*B$Dg$FGJ!^92 z-)RFOf&T9~s13LX$^Zq1ao)9azT;H%IrgI`LQS)W$NH<%V|ED?|^F>%6xBuNeF){kn#6`u4}xxtq49Qe6!Q2bPL{ zQU-DlU#P2A#~Vu$OLfgujx3$3GoR9~MmG{%kjIafvv$8Rs=Lt_V~Rc7iOvpjo_?Q7 zwPpLik(lo5(*TN~vA*w>fma;!^HuvWBiX!^AMG6A)~SMe*yt@|Q-&BDdB>%1S)p}8 zq@g1pUwX>ZSygcRN#h)ITb7o}lB|K*SPWws2Q}3V45aU=ZXCwT&gDNY5A&&*7T8xf z3xJh1x+{>ONj-Dx?N$;KwSXAKbj4~b+jdfd7Av@N$DtLX(U{a7#4sR|De+kp$^h9U zeFaW!%0@~MM^o=xrFKh_ute^KPDW2VNXa6c#9%g9s094oGHMAC;4IQB?!el>@IIoe zYMP9CJ&Rn;nIj)6#GP5Y3{q~|mXcetwC=iwr){So2hCyTg+DTZQ`=Z z8+6zc8~_e0hVW{sf?Y~b@ai+M4I7MYjC2FiscYKnM{y?BK4H$?rvtAw^LWX+^W|G3 zfz-aJ>h-@I{gV^#T=t=_ zyeM~l_2>cbSkt1>&gATOmtHM+6lr&=2Xg1qvakGFF5_ldofWxcC+0Qh&*G+* zRz{O{>|`J1Syy-KBk;vWPtBf1cfu`cYa^OX8P|_hV2UW#!s;0(uqxlsbDgxp9?_g>}4=&cL%VImaW~v@ASaQC1_e z6(A|w&wqO2-t02%L5EebT&R{{XFCMPYp~X#2mfQcZVSJ^b@cmv5AE0X-|stsY5S_Av8vR=Kwd z$qmK=k&}=+AE~aYIEAr|7tE3e1Lkw<)DOm~>CnY<8njWBz#Gf=KDE@trIxnd7kFF^ zsx!CVxi1#2-1MVPT?H1RGNLxvMgSenZrdu`B7|jApW_?>ilZIBm2tfglp_JU^!BY7 z^!V?so6crz9zZxF6h&LzGOf&u)@ZZChZq?J$QT&*t5Mp_XJZ>-c`B#x3X@B-S#IIC z$(^K+a)2^G9YLuuYww>EcrtshsRzKpTK@PyV53BOL>^j zXgyT&Mr+WaRm$|_wk8!hUY$w^-~d?gI^wL2iohKD@l@d;$e<5h)oB>8A1HIvX&5!p zd+2th)N{kUh51yjI3028QT?6M8H}E|Bcb=JgUjEv42*TfH5U$d6Uv_1#UzR|Wwgu4 zGtY7=Hnt22EJ-5@4?~K-BCvENvVV&_W{`p9jmkp~fdUpmwc0 zC>Brqpn@^f|XBR#ZGL{(6Iq98}B>(%aQrZ$rt?)}f6G z2PKcpGm(mL0yQLtagIUiD%7>HHW129x}!G(oaen|UCh!WVL{2j=lRv!jFkm?W0f6E zXI(P5Dn=OdkZSHLYI9n+#!C`0`3mE8MFcs{bAerz_%FN>kVnmv&2tx)-eVMX+qaB& z_NwM7D;%Vjrbb|(mM0YGY>KJMZO0rA^}4nN!tx75E9+ZXqq1_%m%Or zr#rjVX0jV_2JU|DIr`UJnpd2woS_4t6sJo7AD1BTGAggp3Bj|7I(8&{t|SAU-j!Bc zCQvYNap_&t=!_d{50v_hel=3(N#08NWCi1=S}xHJHrAaEQ2HUs)F1`04i96#wkiJ&c#PVI7PH{B(UH&K~pYsfyrL|jc!=@Uhh+o{hbRi z843kr$!R5|j*@~f*dqrco@(qrH1~g|x)bBv+FjLXX}y zaseFInfMP;(d_lhFte$d*a5g>2fl04ek|(IX?`2Jx3rrVGa-*?Z@ZEAt7>%lb$gsu z>q<`D4>s|C#UVe4u6&Dyk~MO?p0&(sI^XumHz>-JDacXCz!l`z@jjn&+vna}Il;|s z=(jSgd*W@_80%YgIjqll@F&L${X4_M!n&M`E|`*e){;tcZP@iup4IAF2a5#KKuIiw z0!MGHd~2v*7A=v;uR!>B@d!9J^GL}u4of!&74!L?7uqE_pI4LPRVVCa)am~KX)74+ zwZnA~VzMrJa1JZy$!#P?B>r{oemwCkI{l&-i>-09`IvM^xQFD&2Wo<;uv zWlvh|yeaWlSkZ1Hk|sD_17`;v>b;kOqQAIU%ZR}1xb^&N-#!_B&6*ys_E){RW?2T} z6l5-Y*ELyFdPwV`3{+j<^cnEm<7!W-!xiKX31U|b-`c*8(7aD?s-j68O#=na{-FC- zex3U&>7F07No=JOstw`ZFi$n9r}#_kb2JEyo?bFYU!~CD|lz& zrT+kj<-3aRH;tpr3IJ$CGla|KUfV3T#(Js}>kV|uc_}9H_zBZ4- z9u{fAU$kur7(>y>z&P};Kk%NLCZBKitA+B>er9vn59eHei2O${iS=0}az(lwkfNM! z9;8;-Oh49V%=!^5L#I!cJC%~+Y3`<)=2JY=67J`zbJW(ogulE1$(F`I$4bhyw^VF` z*atlGirKPN$Rp(nI%AS+?V)6>j}fk?L28P?lAtN%@yR~b&{@R$)v=TVCnK-5VOftb z0M1wribqd+=`7X3`J502M;vClW2!T|>TT(ff^tYiJb=fF+O>rzMc{1*A&IQ5FamHJ zEWa_wTIj76q*IWqoMRnxTI?L}p!BmSp-Ehye${dtVIL?_(bu3IDp;W|QTBvU&Iso< zWe_3U+dv%gS`Evw1Q3NQ@}4@MTF{QsB-+6cp2Ix-X<>u+oQ??@0#OBUy~bc<{!i{)H^a1BQGN_V+H-!=nQvC!pmIIGPyaB?@O z91iu5b*6$jBy?_i{x#m-*bFxV*bXZq>d57HF7DYqt6N#+c=f0AbVLG_7ri%z5SBmJ>J!7~|fk&7y-M z;45t#w=aMH09ASo#*_%#oveB3S<>ljnIHlI#~ju(O2bUz{{XUflaL5JhNgdKsEoR( z+IR=jyBTazL*aIgr`**;rkL<{s0@Cf8s}=}iCbe5(!{DU0Z!%U#~jx~W2fy2g#a$eOLpcQCFyl3Y?tQDMoNmkD z%0l2_y9PWOf1aogGF!OLYA8YTH>+fkwNQ(R>T!+$HK}til*b!gMt45?1me8LxD}&b_ z^zj=u*3L&9`c|c#vgaiO6WI3rYZmMgT-uJdwM1n^Qb6aP^{oZD3^F)8jQY`OYRbg( zjB}2)sF9-$gkUi1+)`H5e5EqG0UI-(;8Xm+F+eC=9B^^@)SQM3#y536eT7FLc@{!c z9jET&HC(!A9L>E-Q%0bU!16_B+R^U;B!=N}k zR9xN=B1o_}8LQAh_(DKX4nZEi^(1krAPj9@q*cX$kho)xGg(SjH&kC?k=l{Y0XfMv zVmov!dgB~suEPmu8%NORtx7GvT$8lty<+8ct2T`%bVUq*FvswZ^O}E@rO=$Q>UN$7 zwLC(_v4FVlJJpEbEOIu1>+4+ZO>B1iHGPQI1t&i?SEy={A_~~shCM;6NSg$B%N~T| zIjNR4B0a~4?s}XW=W~5dn9}OX=>mi#FX8?*VIXpTVaOT4^s7r28wa-_RfsLw$t471 z9YrTS4OKl#(Z&}ej4xdFs|jzl*K2QK$pBQQ1y|bKwh0;O#ap%mW3d5T9jl(T9PZZ% zQ@(@k;6T6wl2?qJl$3xcPBJ<0W}@Tp)FDuOe(p0yI4p5g4+ zkwO%O+mKHgsYSaZGdh(GlA@%JEud}(@aY_f)6bCJOuR!zhvLz98%GgqNDXvdN;#OzR|zerDq9wGk~B1diJQS&@sj_j@ZsCLfQgcWaI#PclW51v(Xb3ec3&tmwc7M>6+Jq z6*(+as}2u5nynh5;~=w-m;udb+b&BkR1D+{S3T~o&bX^e!`9sGP@gVwobiF)tt2d} zRY?juXQ?$ZD((4sVmVSXx~#3cJ7;c31LhqoK4#8@mGmHrCT9+KY;--VQqln+1dQ{X z;+Y+|bqIF0e>&2&v@!+xh|7J`+O2D1T5GAJX&SNIIw#A=sr9aw&MczmB#t?(TU&7- z&4Og-f&9g8>8Z6Ju3aZw&q=-j#sGY)g`QW zh>T<&K5q4=V{GxpsQ4SON$4@{U66~|<#Q@SYbMf8G5kPus|#$YkG&&u4{T<&A+-l` zjua8s(yXSM9oRU|akWSDu8V3)n|d7lwxQ!HkK`dCy(lab9jVqNfXYDy2sK!#7@) zVzDHMpqn8&9f@^aAIHF}a-VY@9$F*WzTr7?9g>OTiMN4U!>V@AszXvGT$2``poEJ=|A2`n$ z9jhwV;wLyQht6<##dKDeq!l4T#(nFSw7F6`Df`cI<&58Da(+=ZF@`(}>a;uN3=83< zI2Z%o`x??Kf=&K28o z#!lW(=}Hr`)Vb1AJ=a3ftuJmNySI!b!$uUcovNe171qb$Ssva6)1 z&UkUu;~for`TJS?6_dnT47!e$3ftY-ui6nvC73A50QwsGr(e^p^ziQLBOv7M;5QYI zb8~HfXe7FfT*ChVxY|xf%m~S?9Zp&EUt`aW7T1yaI~Bt`F}ua(GpNr!4PK3iT&m&J z0C#ogy>@>ZJQCgt_=$gI}>7;a(WS3O9G;NsvQ079%_}o7y+HQ z>JLL&aGyO@SOK@VI3!ld^*H3WOcb`%EW~7ixO5+dSdKioNj~u>80bKzhzxC&Y!kt5 z*rkdle0371-5Kl6U6R=6YT6ef3Jc*>0l){GdeRaZ%YaVaFnST|N-^duY{%4erWtnt zI<|RH#yvT#oz}%$LcG{mNfm`pZgN*GJtP{*90B66cs+SO~L><^pgnQPD zL_S;eIm^31t0@4MAH|MMV_n0vlDJ}i_o({URe2Yb!9m^c^{PK&-60`B2d)QGO+9R1 zv%bb1o|MuIwsJWDe-&(5Xre`3h|U}DXN-H+{-37q3ZyplBLFY+u9DA3BqL!+&vDLc zq6<>JowPXkv=uO}W8e<`NzDhtT1QF zwPE3?vuK}c)OO%bqYD}7#a)tGoE2&)>{?43Ux&Ia%19LIvw$53{IgF(?_MY3eNiR3 znb0RafbU(8h-JRj?hKNRN(oBGY+!4}-l(b-^;_i;l^H80%=;k9f?~k9}vMsfVB9#5zr>7O! z{0Y^xeMb6E3V4DvO*d4;DR%z=dfrlgRvy)r@t@%)vG8NX(&>7cGulXe%UiS`B*=OY zYl?VC*~$p-hBjAR%&q;VCyA65B=tD{RW6%!5>XyO7~qkbcl!sCuBntfX!A>;L=O+=#Oq0UD=Y9mdx zW2f82v$eLrib<80O_B!Z(AUx52>fOCc&#l<2g9k{Ivn<|ku=>&q@U+u^Bm_D?SBn5 zCx%OZEr$#TCmed$4k9frPg@6CGj_T9G2yQeL8@3ayv&ROf`>Rb^sLLYf-OPGLFG3) zn}83cc$b7cRPZ{MkjW9q!nO$>wcP3cB$HFNKfN-G5hB;CR4TTOpWZBzYaVOyL&Y&WrjushEJHFdai8~it~S9FX=wosk;m4l zOLE$U?31=P$WC%QX0&ZlBz7`4m(U9R6A4Z4l%<6(ATRQGtX?@($$w{REFDvU5Dvi4WVQhT&n^y zPaultZKgqj12FCm20oRq1>yxDXCb{fu9)qigK3>MoZF>R%be$N&j90!+puPm1aq{r zcRtn2TV4QmmH-@JcP6(j?v#PF{oYTZtrNJlM^9}bQH9zEOoP-_xEMPfpl=7}4b5TO z-4B;1p1jr6oJPh)BWMFBJk{(?-*ajf+JG#BEZm;8rEj&>fD8#IXdT62L2$&4r7%AC z9E#Ypx=>Vu!8J(EPockQwc}oJz;dIws}d;M3aA)40CQBWY-0FRHe3%{lqP}@}%?>-oj5) zl%CC9oD@GMGnL?qsd1->$pSU!sV9-`iU7_)F~8}Y1>P~8y^ngUbpUafUzqcfYo2ED zXDM^1#;VGL3_H~s;oQ4-Hy9g8u7dV3cNI8c>D*Ra+fbY_a0fWzxtXT>6>aUlRSSUK z*!9h9+FR`kNXNE+oo5Ldl}ds(@#sx#Y2{>G4eOs#!8IC)zh`VhsM;5EU?1;Rg5f|M zug!y=DPU01hTFRs1DqOHk8@?V6~{R}4L+Kgx7eK~9%c?Uj;g$7hmaWx;{!d8MNwOj ztW=}!St4gULEY-j-k~~pWn#xC9Xe*TZKGyYG5`VOkZM_NcLpG29+^EW zHjNSa^AzGjr+3YafNK03Qv`BPIjo2QR!z#I9DP0OQrwa<4l~I5*B>$TvZ&q5sUd<@ z&In`l=hCEiX;*NScXRTws0_>r+rzF2`ufyUOSEs`u{j6R)|{kdqjH_x$7|t>HvU2F z?O8VtPVPt_P6cUPmUqZwm(PAGkxm#dB#ivs39R|KV-{h7!6O8pN=I;mb_WNu%=yne)F5Gj z11m$4%v9$C@}&hv(YNubDl4Nbq&eC^3ObSNR$z@!A27(r9jdf01CH6~D*dkHIl&`8 zjb|ow(~XNzpow;ZGmh1M2+7)6K*`1c{c5~8`@Ij+wJoAx-zX=HEG*Cp=DVY0CXJ(TBx#~H?YS*1zfjqO5h$2Ngb(N zu0?%|(MF(+s?v|;R^TCBcBwe_s?hBx3fsBPSG8Y5<)ZF2>!GxjoHWy5A>ngQI66}40Yn4!ez^G&f6F{RXhQ-eeC?!g-+_GSjIV5D7go4Me2YtgdYBH+_@_@jNf%$+LNxnnVKQ|}&RxPWtm%{BMLxMB*oUjG1LzXxZ0-&9QCMkgN9J0eqqIEuY0ow z$n9ZhcWnR$U%Gk&T2SIL2j1W*Va-&6G-hT1DC%n6v|(KGcl17lR&^E4De7q;5wT-} zNB|6DHG18EU@*Dup7lZ{EOxgG(S1j~Xrw#9&ngHfJOFXpxn1qC(;dwSLrfO~C5{gq zX0%ZVIXr(+-l;>CQgebjW2S3D!Ib2DtPcR5)GJ=rI=Wo7ZxXY7tHI}RJYu!&qK%kH z1~I_EP4y1RjOLR^_$=|}ZFhx|eDJ!V} zN~4Sc!Odw~MofGVN8PP$pm`#msirQX_LlfNM1n^_&hR~J8^+N~bA2l$ahPIJx3LGX zuCo4bBHeQ6hB(Ux+m=yWKB;vz)vz<7u$jPST#g5N`CP_N+gqVAZcgW&Yua;O-$`(v zC0WYt zt3vr-G0bZ4J;1kxMo||fv5cPl)YvNCc2B(L7C$Vw@BoOL4EYgCkn`O2cYf&(6F8sM2H0S@ju0aPZwd1P(Qc8)+a zkm@D~!31YLd)D#MSm%;YQPMMsm9hip1of*kYY!4K2FidrJY(rqJmi$CtAWo$)ccxS z>r|33+kwxZ{&fv0C3Z|xT8_8ZXJfcXRfkL-D`!sEC0Ue6yMqI{nDnk)V=%4)0tY!1 zqfwGEz&sueb54X_sU1+N;?>#eb{hWxl>sj>A1*#c+()HuXxiiCN*M~04(Sg=>5B3D zZDgFGU8Ba@T(*QM!NWL_kdL*4Ch_e^cU9<}CFR(I%q4LY9E zH;%`ocpJs`mYS+Y>P`*|5s(M1eRJV0YfU^4MR=w28VANZHDqOw8TaT{{ZXLg;h&M)aS1#)s30zUL}xUgB%5AjAZ1u z0=(x~GTX^-Yvq(I%0N(d*F`RgHk~Z98^&Z)jh#WyTDNUy4yProjDOX?&;HW-Qk{Q) z-ZZ&V8?d=b-*UR!u`2z%>EF9 zQy)=$S^b?JMjDys%>r-1M=55Cy_5T10uN#?h(1YL7 zu0a`)smN97(~8y!qX{xwNQA8Hpp%{lrE9?o?ral}am`kSA%)K;2aTj0`&OO3qb}Yu zJxzB-XmZQYlG~LT^~h7fts7<`#s+=QCZM-X%bnRHJxA+Ol42-9${yIQY+)|N?Zkqh zhEu}|9YK%)`+@8SdaRP7F)Q-($v(9pkc@O-2h3Nws1xvLcQ_Yx|6cdmmbCx7AncPM+|*2pT?g)o{&erP!$t{Avnce z)x1Ne={BrceAiKupl~b7wf_Jb-D;AQP-Rv)NYG#b??P_&NaC+v6qTDb^?wstTQrFh zHO!#r3cbB^Tz0kMR=bsBnO<1f5|BVL0rV!F=195wRI6?3iOD|IN^Nr2PF>dO7EVBX zb6Uk7HC#;PqAo`)^PqES4Wt?W09W@*4*9Ix*{tCiQ65{6$DO#Yb5`+`jvYor0DgYi zt}jsWMa9IKFP`0cRovrCyE;vCTuG&MN8OGXpXupbb^XxfIZeT8J)J2Amv4?s;V%1q_8 z0`Zg1Dx2ZKUB5qwF%_*>(P9b-?pc*-~u1^zY9D@{qo-4NoQgS=9% zZz9P9kXsxdaad7481Ds0;=e+^B=|o^_~+njFBWMjADu0zl5K;dcL&^8#hxYje{rQ< zy2}>U4S-Z}_}4vpNv>$E3#VU|@=tS(W4jIaPCFWKV89mRjNi?xLOx-;IL~U~tSp)CPylw4I#pGSo_`Tu75hf`s%<{v*TddSwvt>ZxozrInE}Q})YsJ?0Q?B|I>n{Nq`NK1 z<~JPuDu?_Oli^HP{sz%Jccqs}T)&pfxk>0ftBPHzfsk;dlaPCg+0*aciZcem94mTy)uj53OGCV|)I8D2TxTCz zzXih;%VQk|*6Un_o#E=l4Y}G$$ib~EeL)7+3^p$$uOhVAm#Npobff@;;PpApYFb^Q z73ITZa0W-#xT`%)T0&K@n|Zu?HgtubCN2Iv27JJPyErTB~|OpdgM11Rm8F+dQyPkU_xR zRqn5YE1)W%{GrEso?)oc%J&;P zYfFqN2up2jVxJ_3Vg-HbPds!!^???o!AQtX`-Lo;(&T)_bH@UkeF@OemiIDb3=mEU z?d?=$)D%cd9uGn9>sgXsl-!}Vvz&~cqt>$LzsJkw{D5^QB-BbnOH-kJMLMb3jCZS0 z-kD2{!*SqawQzH4*l(1L$BY5T)~xAT+cyMmlyQ?;w{hE3)2$~_7tG4aIKv)IST}xP z#tP(Q4%N#+sttToeW1dZ?W{o&*Sp0)3 zIU~}o`PmpFA$>bmG#49`eB3Qsg6+3%UzZ~s_pMh}%-*)KaIj_FoM7^C{V9s!Q00i{ zimUd5kDb60zcDoo*FI>-g7r4+phhg&X_83{_-| z#&AY*axgnpu%rwDyPn;v8Cf0F;%HhG>NAd<~FxS`IS12*@C9 z82tU}R8W}R_s%+siBmFXx1sf`NTuAn0lT0b!0(ES*wrhsh>Af9q-d^TiVRHlOg5L1ad_+_5{Fxn4SsgQN*Q)AbgoooO8ewsR$BA%;j0L&Nw_(hKuAWg=3A`^!BNvg=BISF|~N^F++M-#!>E8+FmiW zLcVfCb;UeH+f?VCfC2i|u@G`b?Bf{r9+a}C=geS+=zGzk(HKb?#?TZ}PT~$vwN$yc z+UF~{k9yj8ixoVm-Sr=-Q1sTsyJ!?bAE3|?0 zo}QTZrMHQrRVjiAAfCWfyFEe|S`pg75g#O=QHs>HKmw@w_YQDrw&{_AIX?V@>sPJD z*vhc?+&k7S+1%`ebtJYVjlg3kspv&(*+dzD%7Vj;{82eq~i& zgDNmZW}V&9*p-&0=)*bt%m_Fwj+L)yf+hfSj(2y@xT+TIm5g*G*9b51%N7RbR%h0aIoSMAFAWlIJmjIXUyw2%ZVwUjZyJw`=cw>w=!e8)NIz^_J{ zIHv5jHLL4k@cM zV#0Dm0NYT4Mh7F(wb6h@YRN;fOsTr6b$7pblu z*Y>;kd#QMe+d%OB#;M}HTIs}a+i5n1mPS5>u}R9x=5BCHR(js4BWjaOBAMh!Wb(F{ z_Nw4=Gv2um7T>L%v`M%lX`P_r;^{)^3 zpM7_w>en;*MN}rn1o9i&zDpB7Z8>s^8b4C5sp4B`r!qW_N#R!<=DEwi90~1YgClCN z+<52luQt`ZOLeK;k1=)t#X##<^h;wk)KKg!@?jAba$UOsJ!^^(x5n*b(xVqWj`qeu zcWX0TFc7#KS&29wO5wHn?IH7$*5FEtalt<;cLu(=u+)AAd=JoUYD0n}`%`6&chBRO7sap}pXAa2}dY0;I8nr1>N=e@%61azkPBWNJw2Z2#W-e}5z za60F$Iyt3cV<0MLA*-@WO%Ee|StKe`InO;Rc<{>tAa`Sw*$s0=;-^Dsri42 zd)6&XcrA?{*#2R6oCp&gzAE`C;ufo3+>lSwq z{h|%F`*6VEZEnLA;Nqm(c1O|Ro7(5A$KyX0XqORvufwoQX(EWl!E`bN9Z6dCUj^!# zM!N&t8&9#{0GQkI^PaVHz+M>Az9Z;u4dgcW^Mw1Q@ae~Fbj5MLDDdKZE!Cou8Ffu2 z*`!nrcs|o^+3C+!0=eT!$~@87+ErIOr!CKDOW3Vol`fTco_7qIj?VhnF%hyV<0O;T zsOa5 z9QXrKj?UmlEMgY{k#e#!oZ`L~f;N#C1cxiTYY$&a{WJK%;ilHU8|$`~HtiezfF#f-%%rxtIJhT@N=GD7PL|Wi`1fati02q*O_7`^77`HV+lM zb8F@&$_~-X9N^V4ZdHa`1rOfFe_HhxnzMHseK2eMiYB|6=Fg{WT1J212krj=tilqdy|r z&eD*0Z<~>v3bb4_JGugR$gIS64gf$`8<_VMJIIRyZf&^E257dH6_At8Wn}=a!ZXc8 z*CtmA#JR=-j+v@2aKVdo{_=yGvuf;5DVFqJzgo$rcT$PdX(MXnVoYOwW0sa@YZS0rK#0Ln@wh8-n#t6B zV4_n8ny|&m2R(k3=bkC}ffk!#Ec4yR2vi|_(~^G*@oiJ%8c96iz1h2sNz^#(M zyw3*}jd$gx_nyJy9~XVKr_;2MZYNdC9Dg%jb9b!YTdbeDAc2-*bIxmmm*R!hv@^wG zOfpZ8GEhEaS+`y~mN-LACUe&v>vXJ-5#CzJ?QS*ez*rT0H_l6co|Tz@;w|xvxeh^W z4_e^wejvT%hiTZOjyuv@c$F^X%)o8#4|-Q@YAn~EUJ^4SHtZbj^v!2Ybgs>o+s|WF zu9^vyG;%8Qou;v-n4@hOBzN_z8qu0_UX8_%Ur|+Lyeu1&yBNtms)v}tNo*Q75JCIg zXK13sxipsm42*p<)|g1(jy(k%S9URz)EY)3%^-~N)OV(=E8j*_nliu+q~@|OqF~$% z9PmYNk+hOd4{TLSn_}fhcK7K*)J84cvRy9aAe9*)de&r-x{w)2HK%o9C9|0$9S%tK zt2S}LsxH&=&FSwz%tdBW1`pPl?#keUf<5b@o5R}yY?a4)&6`ai|NpwP6fNOwrRsD+N;M(?Ct&n-1sj^jkOzjEu=sRV!Uqj z&qH6SnlFYljV9PfV{066{{XYjX`UqTR*mA#9^*^Y^!RQp=3T6Xa7Va3jYVuVD$(U< zeF~MS%bMu??eLF>2ieh*;dYX78EoL!sQ6dG)|T=at`~n91J<@YS^G2G{4&-J=71&C zbZJ|5*$HQdo`=`#Q`n&p1(W1u2dT!Jev@u%Q-iZnrkCDxlPTN{#(u$9l^1$?V%e0*Q8T}^2IbO|Sv zyLt_$2lcP4J`w0I;g1LE7dMKqLa;NFpWzBgCpGjLTqO%w;Lg39i%)a$w#nKkHb_sM z8!|8--cnSL%vPPWggTZuEIafSmuY2yBy#)*7)<^ADAJFzhPYT(ES=%CN^?-K(M%NH0KXAzEpaZ$&53bPVhur<=OMlQ!s9q!4AE;`_2 zn$WYYEfe zP&P{=X{u%sl1V3_Jt=c^3p3DS)RdJ+ZgZbnvu~)97z30! z1XmSlf3cQLXg?|vpZDO1d$Qu4w*b;8jkApe5ZP;7~Sn$Yg-}Q*oSfG5Hvq-Md4s;*GxV(P)vP~fWp;RC+wJ*+ zkj!{2F;4q2W4L|u$0x06$!uAlfZvZwQ4UTqli!Y&<8#>ho=cSy(_%OWpy|-(>rpg8 z!6kO@>sreLNx>(cea%$4w*-!GdK^=XwA9nt-qtakZ6|K!40%4LnYFO&OJla<(yv`a z2p?RKPkPV0k#}_=xWF8ad)Gu~bPko4g=@3EGswW_1FdCCHY6t`WMoy_#xcZqXCtWx zt!7N%<7otAn(L+5;;RP5SK6Q+PIK0*LvqRj^b5!YX0ln%?Zf3imTE{v6_MI20l}s-DAdDy^@(5gFt;GsPiU=fk&uYR_y1Cm9qQ*l7Y_`GB z_o+p}ErM~J9z{j9V|MwD!g$Y6F-Oc4f_HI`#<^|WZcCv#k-WwPa7Q>jO*vJL0O~m; zo+_r|j56K20oJBfSyyuqqXRiTY8vVaSJ1MHi2?H#G281_@Cq&%1}6=UwN@nE9__;! z;~lEV$2e`kcckQR5_Cin_1<%znx=_M4uxDEq~rlmD~THo$s7agRulzX1_l2B+V-TG zs=Im?Ae2bC01c-J>f*H2)$6Jyk0LP4TRA;FD%30c zjF-+jo|&dwL`-UU&qAJtr-yO_xrWkmR~4*ao`x<~V5_STV&V=1Y|o3u>@zU>{Z>$Amf}?HgY=S%MpnY z^1{0k19Z(@5xlq=CZkXHxs+BwR&{vO>{VFV=(JS zE1I@QR#2tOWDl3Q_pOUITVNt~-%6a01UI%Gg>k$afz*eFtj2CZ`>oFC;OP za=0@Qx!%1MPF+t5wqqStG;dGhYb_St?x!!YcwoZ-ekY}Twegqs-PSE$_eAhZL1z$s zuc-wN`0fyq$JV(|+FRpw)RVV{beW0(Gk0*OmOtDdO8KDkTBuy}-D;B6vrk>|9;xD; zNw0N%Lg!PvbC#OsLhbt3=#Tsoi{V5c+M4ggO-9{Y`#pZerxs87YUAf{eTD^odU#jE z8i$H?m^AG+)?3TTm@IKbyyWyX`@Qfx;Z#2dehc_(!j{pjH#YW*J>0+Z(&5*0=jeF+ zE6L3(`x;!a-JMt}UfQm^BHxLutf0NRdDjw7-eFLAC)E4b&Yu>%4JG-xo)Sz+E=K%i z)w`V6uz1^ExV3p2_T3_5_tkP$$4<57TF;Mk`w7F}OoldH&acS|y)j=K3XUr0vFy~P zW5l1~MZKiL>7s4Kmyj_#MtUFqwRAoR(d@hzr0ZTWzKlsGiyIp?kp|N{0wdu1b*%3f zd{mC}P?{(TvN7 z(2Uc1*z$Sx9ZTXbiSF+%WowD&k))ndGJct=H*I&}$1MZ81vm_(f)Ay4{s7W{u)H)0 zM0@MvTXOXZJ63;>wf#F!wHC{547YA_z;aM}n&{?=lBX4PcjEC9NzJmy!~Xz`R|#~m zPR1EgNGt~usp_V=kBGWur}mm%S|W(2JOpfo&nCF#)LKnINRg~c5AfCPPf&Y{WL?>e zbB{{hQI#sHRP;}De4Zy0?c&t+B|1p6OceyM&pm}!lH7TQG3k&y(QSM7t5ST@tq|Zb z&hKiuWiXN%EhC%YHPyQ<=>x|}jf_b$h%Upui|rnR>W4#XG?dsfby<9q!kDG**u zD>2Slx)JGF77^}*EX(_#Hy%1w-}p^DBjLCeH9)SyP*no%W?uEilAO6zZ|GEIDt`Ah zjmsa6`u>2?-Pmi=Sl#D1fkJKJ9PBx*kZu&THMm(&df&o*hXpdv+zt$Uy|LJ;COyG7aPdf*U1G z2<1YSCnuf0l_M3EIsn}de0>deB%8I&TkB$Y;YLtc6jRVIHK%J68L$D`2PL|J>sj{E z?Njp(2ORbK*5#;Kn63z5ypvrq@-dMjtFAVNE6}!i9+f|kQ3xyZDd%-bZ3C%Xl~4iT ze-&9sN0dMX%AT3sU9s54*E85GYqXu*@LHg_&i6eZ(>$!rwB z8P7`SNXhG?D9R4tNl>722V+&G3-aWL2aY>c`P>k}v&hKE=~UeUC}zVFbMl|(TS6(A zw`~YZ6x)DLW0O^$W?*`dJ072fO?5KFbDio2cCQ51E!66x01_3rIpkKnx71wr*qNqg zk+!zbPXyJgNdg=@X=T{)S0!0enF*p^JqkT=3^xWZEb_BVx#>JpR`QHn3J4~wbeyEAPhk&gXqQYnSXZX^yk=aXDaQXRncA6~V$q~2wK zW5+?B-qk7gBh=nh-}8)|oM6(3@DE)5YDj@4Pw<}Jl{3Y$f`ENNswZto0_|Mjjos4E&}Nt8Z&(?T-hL7oSvYUsQ<;z@in zs6%j!-)D9SEMVgU*0BU2k~7Ub##$pA5P0wW>Q=0oM(Fzs!CLe)TElk)!IXehp0)IM z!(WIk8G$_CEW`%{<0ij9{s{QYEX|}(H}4naP3PsseJA1j+4RZL!;+^51B%j7)7?Dg z+Lf92MuqW;`@@n(%&m2&%0XO?*UqhP<-7jq zj57E7R%W5B>OKVVAbUw1+dJ9cdGnNO$WjHE%n&(1hN~s zjkBu=;@h+Wzuv}iUL*0x<2{e;Wv5&Ca@R+<)jU^nrdX`*vV6Rs^|IrhPeEKbj5pS8 z&9XiG207teq_=4L!{BGbjSs^b1Kw)aDQl-$jLUgD1VY{NO>>{LzwK>lru;7Pou`Hh zLt!3^JV|`x5hyd~3M_+2dv)7_vkj&%FP)uwaf)8rZmV7AWa57GFQ5_D)D9v4L zSQjiGEW}~D=k%vZd%-G49C~qB*Ea#3TX|dz=hmq{rO(QH{MEXV$kol&zmg^+j=g;; z<4`VDnStklo-3Re${eoa^9%xd(;HCv!3qkf?r}qs+?1?!E2%U{cay13>9#t|PFIb{Pf~vEbozd@BwyWyGBG#i-^@+qQ*nLE~`m zTt%Lu@AGdC-Ff2`cxvq7TOd9;!Qj#DIvOf%>UIl%u&{{d2a!!}O;jvDOxqd2L!6Sl z@l$=VvUagioZ}o;POhf@%KDl&S7Q^3mFM4HVo|?nq;?b_-+V4Fg=A) zxtOsh%p;yqb{@4p(Y=av(nF<{LqX)}j8=AY^17 z=QU4PemPcl z-MxtD0-RSBWpG4C=If8W(zfkq+5+IH0DRqzWSUxB*+a52O`9nr0PXLdzmIwfh~z31 zGh>sUm7>udpfTuix4jZZdZ+XqDy$CX?#HEN+=%|}c9Y(;CEo9Y(4K;*+#RE7!8z^TyJ1sfj;HQrmT2%XgIRcAY|v0?aej0 z+@yU4KHWUCfWeo)y-v_|Dh>y?O3qR_IUCWG9y|bdbDo2}ShBV_IOE$DVmm+?SRcm~ zcG}$}1OiKrPZ+6k^Eu@aBC-I3!+X`ZEP-qy1@$=VR&A_#+~au913fEB*4VfJ4ZNHW zK~r?}Gq=5FQ-;L`NIkF{(ya&!NUGTiah4US6^a78kYTvTsi~gZZUV$SymYLkC9&BC z(U#W#03cEO!ZUz-)mbeFkd#0WADchYw~ec^kgmN7(e7Z$ z5Ov%!l6uq(ODeZa^&K%<(m)v)?D=~h^_wNTdMR=v?FUrQI}F)$d0Y%v+nQYuJ} z@)Y9)WFDC`c{xS>-q#(+y-Klc+mpG^N?IaQWU!;NAm9vVj@4~ZlmsAc9FhlmY%mg3 z@z1Xnptj_$0-WcAo`cq~aysMOfIuB{ij{F8SjkW`%O2De&T+$Z$mo8xJWvd)s38gH z!>v-ZtZNoY0{o?v0(xyc3{}`il1s+9>ON7NA4)A^Rsg9W9?UC77!)$F0eR&5QFd;n z%dtiU`S?Hx=Q+(>F_Dd=v0?Xs;2z?h@PTr`1I8QW9jb?owZY-P56Ly5GG4jD06s{^ za((N%l{r*u$380+3^hJ!J$oHBhKTbj`LaiBt;lIxX~sP&*zm!U7h7#c)_Ig=BWA$% zubUIb`i;%|%O%vT@?bvZ$-wPe_8vLDigkGI8C_0r8`8aK(2RN?8;H#7)s^Dxk45pG ztFL%z40l>}xu_q_$d)E&yilvevGF^eR^M>i|T^yb<9wpE&Z1n@Vq`^X`ED!$x zUc572ntcycAsnKv;L#3*S3-j5k2Te)CwWPo?Z?Dh`)iCP>wbM|B>10tNZoPrXCY5o z;qGoN7AWOIy^3eljOL##dks{my0h04;$6ZhB7UO@f9E($hn+HDJ^r?TeAwiw(oYMAidbA$aqw0GN&%`UHQRT|2af}ZB=ANGr z?#d#G2;-BUbL(C?CaSxGdBpLIdQld*MnlM~x#FKBdLeVoFNpS|@d|frHNy}SfO)~G zzu`ZX*b^eYc^Uq-<7upu+#kk+tjajw#ZR(Zq#n|L%+F?r;u|3$=g#lpUWTkiX4j+o@SQdsl`xh_HE(;Bm(os9(jb2GJ%#9<^MFYhqNUx}Qba`1aC3Rkb4z z00SxHA57PGq)Bam6mr9L71l-zBkuZF#WU;vXa{iw9_PJm_*=y~t)k3CPIs5L(s>0StdE+;YG>Tab{*o>ljXWqVikKz^V zo?Me%8Ks3bq4hn@S8Z-hYf%j>gr27-H7T~q<+9xO&kXDL-YwLA(SAhml{kcOtK4F$ zd|~*Qo*3}8!&_XRJjJ#!2N~(t+PU2aS6fXXWQtZ!2e&ohpBy}0F1M=-VpK;E+p&M0 zTUOX2`JK-$x{`fTc`m1sq?%SP?%a}1cHam73HX!aAB*igKjG~@YwOwFSNCa^jY;L3 z&y2eVmVnH zVknp2BxbdYZ60ZC6P@I#$A8H3KiRYPW$@0E@o~HhqO#s6g_UnSQ>>hBxw_z`x*p08 zTzXgA`m#eWoC65TbRhSynEwFanl`MT2sK|4>ZOje;!DXd?q`(bxGr(erF{Y7eL^_Q zk|`3#!0mP9d)L8cmHn+(2q*zZ37QC zDA6eeP|xB^KRA0sb4E57k|vkVryB%ke$ z*#rUi7M=VOKZ!C8T6#8+}y5%X#JjnI1a!{z0jlTz*Yp)S%7KS`E zy_0zaaszGP*UmcJvP~n&GwztPZgP6_(!PiAeY~xIs_J%fmV0|vkRT&1k<%UP<~?Up zXP!ZTs#%WS;OCJ~1s6HVnPv?rSZo&-_L3w?^UEu3+1T*Va z#%@rb!y~0__^pI-f}m~PjMuMGT5*SSJOh5=b16J(BkS6jQVeOiOMnYm! z#u)dlnPCi>+!PLj6y{$^{z!$?)?s5V)yEKKAqtCyl>)|^cb!m zOM-M8CYECYEw41Fu^Pm7U3 z9^u(a9iyG6wS5ur`^6q5@LrDh(pt#`?-`B=*|GBxjGX#ZpRsm_rKYy=YVdi=()ndu zk1UbY*T4SQKM$ht*M|H#tZ4CW(_?0~yD{{Pw<_)IdRGF9ty`4!qY2^Zsb6#DnA6A- zd9rzvNX;a&6OztONi?2dkdOvPKuGF+Y5XSa5E0Mb1FyY87Ys)%Nd6l4+)`FJ)3dmN znB%EnFh(lGJNA`G%PRH)ltQc(80788PIFVo3t;jvcq~s}rE{%UnX9<_FP(!IP!A+y zk?matsOCM`IV7AK$8@QF$vsO}$R1Z5Z6mLG zv>#+_yoMYA02%eoHW9Kdz%B{E#YSh_87{2AoNXBSYq}S)7fW_kyTgUslpy5s)K+w} zF=j@>=LJe&kH)laz;Jf9&wO>Pxue26?&x{rt#v^qxtr5eR%y30so<(C`$PVIiz9+e_)W8Va4)~spzOp^{5aAS(u zIxAbM?p$LWRajR>5~oa@@x?(qkjFSt)1LLKdJ-a3eeSr&O3S$c`haoQH7upMpDtLQ z!1t&?Juov?if4W}%|o+=;~DC6N1P@UoPRnt5#4B$ zHQK74SnW#)3}kbPX^|y#P@c%ZRSp}T z-Kk=;e1gN|=NYXA8|67tp;+YOsi_tuBw!Jb){V6@PQ@!&_DlD74D_v-0oVY-r>Lre z91t;(4{r5jHF}~ zVp`pcxZW&+4`v+&7PBoj(N{y%#pD}H{`p5kTOmsZ9eBwCxI3{XhpRTwRA)H$uHQ+y zTj*9`dUY73{t3t9Jq~$R?^4n&*7iJden$O*09!NoOQ zp?{PX$;K<1@NTLEi5cKvaNRktR`B1$tyji(DISq%_mfNaVug1%N{2NSc{5iRvN`r~ zEIW{{;qTJ6^bKfU#pFAU=y*Nb+OVe6r_!$GwY!Q)CXJaHB~q*yb;zu5ww*IpUD*i4 zVEfeQ)zMVs)3-y=uJ*cy13aEH(z0g+w(P46=N;=-$wl>|&Yegk?Z*bH$!5F4sU>+V zao^U5D`<6cC#jvhmnC-r+i}fBxtHbY3CSeYwz1om8DKdEqL$NWW+%&RbMH{t)H{`o zgf}@Lcg;m4rcKxYWAv*A+>ww8=N!{g$%Kah?L6ldle$+lQM(x&vJ}G(I}uuz&H~1y z6VD`ds_@&L%yG3q{LNd2ClUo8QV*qJChT-ZDA~7|u)xS3o$*_Cm!B+UNnCy5-nrXU zZ!BPPeesIhvH8aJ1gEYrc&=GblT)rVlRA4#{K=ioyQWCywPU#`dJw2M+%gSwG0JyF z+qhAR(v8e<77B~k^D!u;Zhrs9$GeM?di=>iT1hPcN5>z ztlOw^j80erPDd0#sR6B!)_6>*<${{USv(1%^4_(;nR zewCyq<%vhRX~7314#uiWY_cI@4_}+N z71>T!TC;|5zK1VymB0v~`0j*o!A<23`X&PdKrQVng(rYfo3^0s)-b5&-5+l|WEU;^Zf=iaXO zGxl!!8EF7O02{Y)&0K&502Kj;%bfJ58*{ymbCZl>rGgFu0rK?(jMg(Yj4bR}idRPxz%lKC&{YeYbqnS(3b^Zz_4EBtu!LQ-F{Jt91eGLrJPOIX z`IxXc0Of|>Ptexo>c$Qgat}Ce^_z6A>yi#Y$Uu0lW1;BMrE{O2d~h2qynq(7=gB7+ z1pC&l+&?P?%H(o$(y3f3upz>d00oCy(lX`H^=aYnxruWV7X%PldJ3a*hap&nBy{gv zS0XeU_Z_*&P%;H&&a3x!lEnS&bgr0A;;V*MGhPA|jDW!4W|)z)9fV|pImKFF5Mw+q zKBqM3pxO=r$nJaBd}n2;^BBB6t~%Zv#yICccQt-HH<3o(amLo~Ot*{$O~WnM*0kdr z*>iwMBacs7Ii2)AcBDB`1U99%khoPOW4NsgXj{vC5IG!TrH(Q|Q_2o^4&tuc4fqO6 zMg~dSiqFlQ6Sq);*Uc z9JRUC2&Jkq&9eeAfd2r*bIA1Km&*A}OaneM$6?;IUeVkJ11E2;4_X$`hx@xgAd}Pq z?^53~vo?ogE6MW8y7b2&R97(W+$`#+n(T_1y%bjVG8Zb79ch`rS9JS z3u;YR#1L=bsZcs&oY57p-pUe3W7id`ZB&OUO5>@=T8=ndc-xG7j=t2JR|J_HQ8ANg zCj$+G&086E01Tdbe=1asqU0x2lkO_VlrpSN;16H%td5kLAxm|L;Ejiz_vuqXAMbZ6 z^cf!crZh!lFR->r00W%W>m>6^WNrWuPI>#y=7(?Dgr>pYf>FNy0(7k z9sARC_*F;E(DWqs?N;?s_G=kkNn_bp)}1uxH56+)k);PYJrT(Gm&R~u7j0)8izVX_ zLT%CT<9*@(02CcU@mLujED1cH{{U5X{v+`e zQ`qTRl1P$8B6I-ZJ7T_~l_f?tTc4lfF%repb8k+EmTNkDY1)jjf>-9?`%=Rv?!g~; za2wXQylWJ5?7e`3vMzWT3Vkb@hC?6+g2S-ly1dRC=xynD>>Ma;9OoqTu7BbhE69?e zIc}phZ6+4vuml`>^sJu|NG_qtegMZOnwO~7r!xi9pb@#;y|YkE0c@!t9%)?pP6!zk z$z%+;3*V`yYiMYl#u*A{4T>7&f{f&hcJENI1Y`r&n8zca$n~eP4N3gMq-P`n-i5l! z8R^tjQ!W7quc)RMBm>_$q&ryA8i3DSFdZq6d`KZ&k4mAu^(UWSS`>qVa&b?fSh}}T z44ucGywkqg!AH%z9nDc^IW&2MCu2xfWyZdFSmfj$)eN!&fwe_eW+VfgQf}aIPADCT zizZZbu7=WY?5#91w&`vGP=C6;&2yH~?wNCw>zcPc$-0zrgi*CddsBN^irSq8mGck@ zIRmA4UJ+llTn)n@^!Kkd(~)I$8Nl|hOYr6`?Z(~+9Fy9fiCoS{ajR=74xUhg5;Ver z$3Nj;J?pnGsOz%Ho!gtBuTAluu(rBeD#*aFBOcY}dK67_sTK-djyS2+$ZFA7XWk$1 zOOM(9A0PZnJ}T8MeEl~>Rhdz_Up)QK%NehwKj5+cHj_d9o-_{*Lgfy>tfEL2vHQ4` zz6kWrE7<=4;Ge$=7vZPCJsw$L@^7_e7jteOF$eC~$zSkVyYkxK!ksy?(QUi9E896y zTh?~>S43xqPD#!6W~b~|@m?(#;NFpCa;#D$5s1)Ws08Dl=Dmww@!hVYr!>~Li7DhX zQT(NSh_4U*ko*s6sC+Q7wYemTHps5Y_X#}nUb*8h+1l$$)2=lQCsZ-X{Y##hxsJ;iO+IERIQ&a`eZcuOrv>Cs`S7=2=xTSSxY`T=7)8 z-kGahTU}Ys0Uvz^QGxCop{_D(3(LnP$&YyIBP9Mb>C)2Xoarw!nehjS8dhcVS7Mxu zb@~%f=uIW$>`9PAD&j>0KQFkhxO_KTh7(9uRdK*5S2a@OMY_>iNZ|xR20@Gy)7H9d zqJ*6HISD4!v|HmXmBYkiKPJ}3MPOWLHXb39$h?g$FWrgrmwlV>Xr!DMK0Xi{RBRmy9N{VuH+vU*Swd$>PI88^vx|WlyeT?8b5LQh5 zq#lNk6gt~O6b$Io0y0qNHDg)TZZ1%nC1+(E5_?xOVp~zX4D1Ri&&(^wej!7d%MqwzBbikRQvgjZHEOh+m8L~>)(oAC%4e_nWfq%kr#Ol z(;r&+S5o+iZ>IRJ^Ix}UlWQ}w?(2@6`ikgrjk#B;rXDFTX*H`iJQ3kx;a?Bl-E_RP zm=;_B{C6YTzLV0m%_rk8!i{UfGPj=hmJE?dtGJ0k%8~RUzA%sCB)&4%HGA7v5&fg* zniK>A{WJ8%d(Z62;&$-Qi6dtN?1os-D)YFXPtLm{r3opfIYm{cD7ES8e7yyv@S(H0 zS2uR<(8!^1Lm?pkJXC>q9l?usBN)Yd5A9vxRPaZ~oqh{{lTOv6O-|e|8S^-N9^BWI z+u4L&qyV7c9Q76SmcH!yXw6#b4T=HI(~NVT{{Wp_hS+&yKfG>mame-*9s9Qz7R02a51md+L zMG@zc263J~xvtuANosR8Vax0 z1Kyk3$mN=AQl_P+yc&F)a3qm?4nXc}$^1T|{bB9%Wmn*WP1ho=rjHIz_=i9ASl#Co?pQSc71YnYWohmtrT=o>wFL92< zjN=*YQIWwtM^jH2b_bv|fdY~-l6q!~_7V`T^xd}tog878fuF*aRObVydYUXOc>YwH z9>^`XMIaz#gIT~UkzE)Ft4#c*(-BRC+Qwb1x}_v}|Gl6H6FCXSjC)y}}Va$7k$ z&H?F93^wlEc{!~4ro`D_x_RcLX+&U>cZgZY9^4i04kGe2%Tn?9YvfzyO#dp3i8??3natY*& zX1vDUU6cO+_32-sY2QO{#nvBcv`dK8VHc9UPu8+LEv+?#{!wGGjtD&~sI}XuY3&OT zaywTCs9QFjdh4`|9x9sbR9^8tT|k~QBJw*CRp)@I9CYdJR6G%@t@Xc=BZ9+#4R6gB zeqsounKJBVK)YNL0UfG&Wt!6B9YV?cyQrKxu;Anzii%|Nm~O}6S{hZD)nyj!-dSbg z&p>I%MrTsaC+ZjMuIcg9!5U_*3`;H4UnR~utco$uv9GJ4_*0=jfi*7;YPL~LX{bcW zIyeV#BEKyE0ALS^n*RX9kBfisjYV)phNgVKEzS_Q=ttJSSY*`Z@kW<5x|3~T0l$PtCG1G+&#hPHTV(n{{Z1F@4+97o+a>Y;xQU^(}_ejtiJ={kn% zi%m~Sp4uoexmrdTQTo@(b~o*3IOlFS_O9l`Q%n6krwrkGn&eI09de4b(Z<}{NucUM zMgkBSPCD1nUjhCv>b?TjtaTkKb%IUXB_#pM5rd5M;=BXIQoYI|l6ORKze?5cR;(hL zH8NuzaoVQ6DAbFLwqvC!MJBgL=?BLjgFY7c!SFj(_=n-EV{76K8WC}G8gjDd0OuU^ z=D$4k*j9Tqn>i{ok=)nNz5)HEej)rh@SdTeczaN@{>aoKD=n?e0q{u1?Dgqh6{*1T z+e#;p%Pa0@Yyp-hCb7fDRH>?R>9g_^SDTu9@9K1Z0lf3A3N8s?AM5X3J;jPMWFu#; zYtOzK%JAIArBnrNg-Op9>knl9X55BupFv7V#h$GnekVIEwmEN=M;NFjvE$`d1GafJ z)IFZ$A(S3)8z0W0HZB#uP}l%u)EmkPY;zB$1!Wxd$>$lNEV9jyr~}fvfu|Q=nKxtm z-_nvTF%XfBw6#cT2Rbaut{HKT#-&SNoR#HI1Y@Oa{{UugEP8XuIp&45DS^tbOjJo5 zL0GE<3m6Hv=57uu?X+24gpi}KJ?YWgC~`5L>%1T5SM8u1*p*TPbnjk$M4is+cj!qJ zTn2SG0|YOvYuiD%s67gR3v*Q9M2(Z?{o4NkGg@#acmp9wBe50c&2rkurB7YS64)We zSg|LMZ(6@;Y_4(}ws`4HfP$pp0(czutJZe>3@#UvKDEx?wr5gHp<>qEBXu8mvEwV= ztV<9i`D^n6cMy70lX8W@RpW!yVx@hhf|G`BNl}50)i%4h-ip>Uni9xLrqBj??@upGjtyveDcOSq#!2XECV@bFWOm#JKPv2m&UwQ^izz5jjFIza9Zght46cE|8R&Ws z#$)z?;BZpVp*^c?K3@!fhN``p4Y_~^9Ff|#OhA#8 zF{+$n)~npaLiu}wgWU8U)zKO0V?Sh`tjKFL+kuVG*Ko~IxwlqK00GAvc*pqG+?Ko* z#!tEVt3G(bpd%oIzy~?}D*2u4jg=VRV+P_D`BbX{cv8F?8W9@>bpz&lR<0Y+evT<8x;xjw=phPh+7wks^XNhCB>&lTj4&l&MTGW&o;; z_3K#oa14^l0V5~d>07A1k7kRp&0Jf8aItSSn|6&HA1*0wJoV5gxYo_p3w?lI62_|lEy zv^wfWE*B}!7$`CV`qNZx#@<2cR$c<@wv<3BhY0($*x zM$z&H6nnO79vhLpLn#290&8B*U6KHJLGuHOnW-dsGJ4qAu?1Lk13aGfy>AZo4y5h+ zb6Iw11|-}u-`cjJVxcw_2r+^ilj~fS9=AFvQj^@CHB*7G4mcokYGDX@4n9>pfKNe5 zXrz<#9ofOjtH~OY!GOR6k%N<1%3CCNs#esdw&N@B?^W*N2qY@GJ@%d}VMre_10-M! za%!Z-BvuMB*_YO;EKMlvW6B(>Yzw=&TB^}tm&gs!@x^J)Xi)A&41c@Qt1vqOz#I@V zGen|_gRw?fcY+Dy+tb#f^1vmE?QiybRk;z7a(3sYN2Nw^qjxIU!9KaEOQI{ET}=6v zSY!syae>Avx0b8BX(55g=Cr3qUzS2hT=f1~st*andY|EA(vwLQdyP4ek`4|yQaSq6 z<8Wf2oDq^p?^Vex${Rd#K&v+LsSXB9V2}q;D;Al&TDf5&!-3RihUi6TS;7!6%DD%y z>rq)@`T>k`Ta0^((})5@VO5nuIUeGeAqlgrbY$bvw4mXaNu>#dt1c$gxWm+0K+$Fs#?pX&B~U4l~79-NzUrtV@7M9=^3%N6b_V zdehvrAx!h0I*@8P2e0W?UNRI8SRT~`quY^4Wmd^8j?~6EH6*;8W15Sc9QL4SGI8u_ z;5Ko`N{=}Lo~_Tl0u+!q<2dU~-cJ;aFlbT8=9ug+QN=W5W1MEB;d;>!-s(TjDIi3p z{J3&H@!(_9mM(e4MJWbDj(sWduH*NxPo*1)o0?$S!IuEnbEU~E$M=9xD~8j@k`b7L zkJh?)bv7`pai-tKuSATRxd)4Ga{&y&i0N8?0Q@ZeGWd<5>8L+>cYPFU4p$m6|!aeu*Pygu5e!|wy@=($}p zQFvuwGl3xEC-A76ZjPIOEd0>+c%0vJmbso(%sK6v7|ZC>L;izoLk>fm4iGx%2zsrZbDB-J@{cf_7GlJJ-joRN@BZ?~_m zD=%E}EVnB0szo9{-dQ^Rb6m*2K4VE52-A)f*JU_K@3G4&%V?gH;Yn>Y_@KSi3~Hz{ zHxftF70rA^)-3&CRP}rQgYIard~$#!Y&^gEjoh zspgXGbC7>Z@s)kDWDMuIuCK!PY`01?z#vy9B9fB1-Gjt9%_j6d$oOTgE4F2klx-{( zyPEnZz}ov`bI8WgesYcSh4tpXO894bR4NeC@Gu7eoOQ25)&3=0>DFmy47;L24b+63 z9{69Sc(A*khuK1^sRGySJMp^v#FCE+_-<=h;!KN~nXw~h_(!HI;eQpTs%Lm7^Ca@v zpvFKXb^6u@t?>fp@8Tx6sZVml?HAHAC^#E4^8SLoGT+A%)5ZQtW})B#2K1PqlsH`#4CxD)C;KZyZ|$$0~N= z1`h|@iui-YkEUMSM|Eo&tct_u9mh;p)nBq-j1G?_oy;;U_cq|5UZqBJfNE;gge^N8 zVO2)uO8WFW{{Y*r?g#k6pxQ#p#@|b8ec0=@hJPbpKEPH{7#V&PeN@N}k>TyaBb3I3quauISUU(B{kyV=oSw=`b#a|;wY!bbU%Y%Y)2tAmNFq$`^N)!yr12K3G^@G%)+Isfj@hE^o<1q^ z=yErglU>SKoS*Ac>H0kPk`0PmJxJ!gFTmdjFFZ>+m4KAaa(NZ)x);GcKS9|g#>OZ& zHhX)0YPnKco>QpyNb^sIUkuek-(LKYAl`5}`c=P(-ZG9CYiK0eh`Rt@K3eI1BKYPl zBG3Ci$VC#SLlonu^{<@u&1PHsmUd!Ksqb00%v4piBh>X-Zlz`{dEgG%tce`4C#N)x zHW=q5dr}9>bvQn>BxtP3k%&Tf9Qt;qvE0CPJ#$IB0QEnG1cuHJ&DbAGHry=GV6eg9 zgTbLj*4jR9;F?l52L5!N#~l9vo+^Y*w~@|ERI3IP8NuugM+0$?xC6CZRt19*oO@E( zG-OcjQ-Ux$eQRe*h)E*kfJyFaILDlvbK4bSAkavJatA+3O$*TKZ&9sajlm@JHCydq z<1PvHroGe8r!tapf%(=$Ugk6;VD$ZHZCc^8(Cjo@lrA#Eis&51;I`e^I3)4KaJntX zNw$tbARc(G(WEj)2nqqzNU9PEU*ZOLVZph6wakpo4 z&OAS;ww->BlY&66WkN4>_@+z%!;Dvl-az(OOr&6`;=Nnp?xrua-z~ml9u7Td>JnF1 zC!P|W2VT_68Ja7lAYnDcLO6MA;!rRf~0;``p4iuivv# z=s`*<&t9jWQcf+aX!bs^Td1!sEj&SEh6O>CDErEK*X0lV76-zRXnz8YC0p26Wu^}n$~$Ee0f6&|T%P|2J&dI8(sxW55wB_|O#3QtdZ^qb}T zO}e<}jz9f%jh)9WH8K1rVAi^;GICgk;Pf@_G3ooDhJ1+}sL323dh!hpPqS(XyK3$u zCAwF(>6XhatR^fra1PVJtm^kYEF-dMl)l2O&OTr{Y~#H~`c^pFr_`U~9<|XWpDe_2 zo(TT5^s;3jfx$fceih1|w>#}~pBi$3xFN@(=bDCHHH)3Aka;JjE2@smouoEFKQQf4 zn~V~_sNc6Z&ougi|r3+bdC&b))1XQg&iY2`^QrC4K(b*LJ{db5Gg zCqDj_R=%T+(aNo_2Mi5K4Z9ggW*;^<$0D~TvoNmWgRV#;JXL3bLmXfO-1Qa1QZ1d- zwkt;-?Xm4amMxYU;;zC%o?GP{61{4+)58M5e@c-ZjCmvH=L8z_Y9^0DosCE&210X{ z9Anb7>U2>_f9f0bZ)WSoJ?0I~F|miMU~KQ>7Xj1kUjf=jDA6E{*_q$?F5U}rh} zO-~%V@VO#FPdVx5wUd%SIfi8RskQU9pl&LzYzfl+liJwOMh_I6s9l;Uh4CFBs=N$*WSJ zDnUgBhw*k5S)&eA9zg#0YT7Mc;!4P&ELaCBT%3H}2&xW%J4jg^x1#i|YluSO9d|GB zDd~>YNt5Qv0;AP0HlX~-C)TYP{O4nX9Sf(2VKVp6#zDB!T?@u@^) zNQa`gQAxWM%iWRMKnp2#7#}FlrEl9s9^_7gsLOC`E)b?Y!4+691Rg$>uVm52Mpx!; zoOG;`yCufT#VA0JcE;S<$UoMtAS|JsK-w~MR8@g~2-*e#&N!_|6c$c*2IB|Q-m&&p zTAIQ|1uBxB!(Ux*F>4ZQBEP1Q2iqXU$}SJ*yam7CnDji9@OF z)2nSxCi=(BjN=15;F{)c;#lQ5P%^j$dRKpa0;pi1G6S?T3|Bd*+F|g68++q}ShZ(+ z9gezdQ=5`7Mlpfhj`dPT%V2`Lxd8XAE2xN&Alkm9)fr=0$v{aedCz*P22cht4r(Q)h7@TzGqSS0C08sk!@u}e)wR-; z3K(u|4CDD%8*OlQL;)p$t($qyF(+OEkz5k9-1Xr?QQaL!o>{(NPdt0pg}GTUxxfp@ z*15Q^3}rVGL14J+k8@h_U8vo)caefQ#y#shTcObCw2j62R}9Jl$36c53X0-Ask?v= zL*AhmD(i#8u%Jvf=OYd2OX-XoL8tNhfJO+&cYTWAeQH!TCUR= zm=*!>KH>YnTF%1ea@2FVlV}HaIU_wPyB)~Vau^(Bj)s(6G^_v_J(PE|i|!k+Ko}cMbIn_0dQCYR zuw@9r+>AiM{RL>nGQzx{Eb zx7DNE+lk(yzRHwUIHeQw+SMIdIZLU#;yrz1)MpTrgZB$_kEJ?IakV`#rSt>!89LUPEs++N0|H#h0Unv9@a(1F zmeOnza#N0TT+XlJS+BKS(#XxU{JA_2YQJTDiPGmdf8qesc5G_nwHqh2W(4OXpQp8Q z`hx=GW1m{@b(rJ5MV=$c8D*{mTSks@56kqV(k*OKkBJ9ElSu3Yt(knm(2hymY1msCa;_UB zV~k?2TtXD{j2@Lyi392DNfd~H#V<}rQA^da)KX+n09faqIixtc!H~K#~e|xir1GOb;j<4Kb3u({{RHi zv5s$!I$gx7cEcOUA3YClE8>ij!yJnHpZ*EKXvy&&D}Znk6M_yhbNs7HOW(1}iP8t` z^`*#?%rds%MnM(wpZpX9!D;aa;0KCy;*TZFFq^0pbMvPl*U^?2&mEvHyGB0>^WTYg z7TS)Pbrk5XB!%O{bU7HV$y2hC=SF<8O#F2HpL|awejj@_c1KPCJq9}0v02=`$pD?( zc**KBUp`&w*Ix>LGWdtXjJr{8*`ik+QI0_e@vlm?)pZ9T?Z7`cu~ZM=Tn7Xas_$rr*U&| zU4bpTC#`eBoRZYqtyHxp)NfK?TY|vyLB>U7T;HPS3%EBOYL?opg)sLfIV5@3C<6tYei-W0YNQ{V-=&djJ>0J z8fkp9C`5!3I4z#FpLYkA#J4#es$&$8;RZn*mOUxAjpW2bvX�n#M^%*&1_Q8RYE% zfPHXHA-EFbBN*cq6t^B^U?~7+9MoQFfI#Ppt2r%9T)Gl8n8@dEQAv_ta^$b2HXvCK zT;rjjkf&~PJqJowGriG%(r%!s&vWZnT5PE~7&zcoQi6-McsQ%k5g}94j=18nlosfv zC$VnQ0P=tVHR#?1@SNJT=&KRkk=viZ*CpZI6Z=Bx2lu*j-n~ZS#nAX)O*+7q=Tl8%!mATt`9qxb z&qH0_$wmq>*&I==I+a|e?%C>IN~mB4;-@TqD2g{AV4R*Y?N&UNk&vkhRP8w6Vw6Q9 z6%QaT4)1EHK^<;P5*J1%sNfvc!EAi8TL9zkf!2mbbOp#?W7vKbWk5ii%Rb|`0CCp3 z)OJRYnnieFbA!2X1_lQs>+4c9M222lBRM0bXj|FD7&Kvn5?P0(NcQE?g8l6Lze+h$ zXQ55I+~@A?2GYSYHt|?ez{oPYombRn71v)|D=A zP@`vddJ9x@lfk?x^8WyMM+e%fJONfhc^z?{mC{LUk`a-i={!Gi0y>76s;&vi8RS-H!_R^(aPVH=6@uRdhnRW%X?`Slq74H708WIe zu_dG+4tVLB=4bfdb#~EN!zY&#fU3FTsx8GIA&pK|nz{5z)il2jSf#as#EhdGm-^S9 z_}}6L%LK5YP!C2^_r1+?k$8u7iq9p+SdNU`{uSmvHt~s^kd2)2k3mVc&M3u1#_?vk zCFRSLDLmI3Ex_J>ZhP}gy}xN{-xUx%40oxa zjIsRnYhMOJx2nNBp@_nj5H!3^&nttz; z^aP5H13O`uVvjO)Xtfb8M`51zY9WO?vJQBv(6`IObM&hrvbpX(XkzMGiwvN3tEj5D z3~F?-=qt-+e8}CoqQ7?9Q5!)RI>t0!-60!rHE7p8COqNJF10S6XmU-um^pgyWDy9_oqv7tRH2qmOabf9>4EFb=rms@Lj|{1eOef&SU>oE{I=Zj0Pn zE+8($?)nO~IYLTNCzdk2C$l<_*(2j%*Zf7Q=$e|PuBmN1*<8Xi#E8UkUr|pCI)mS7 z7qKjs*AW#c+=fONV1GLJ6U8GRG23+gfN+jMQ@bpc2Y-1lYpc!=h zvTN}o?k$CuRp6XwxHW}E%#}|2`khoK$o-q!rsFov=CddO ztQ-MOJ`PJ`z9X1ROe*~0G5SPrJSsv~+)v(sXWj9`Jtfs< zWP}H+9z{lkNEkQHcmtlMyxFFDF->YtCL|CDI5-$RO)PT|9gB528&4x1wMC`?=WjW| z9{#l=$r_9^1>~MO*96s(yjiVnI;m5;aXjastq3kbauu>iLDLn_T1dO!Bw={>tIIH0 z(;qO&+B=HLMH*b#Nt{MDF4K}r6M$;Kln8?DCHiAMs{~AOfWZcN4cqBbq=A8E1wieb zf-6|QQ9X7yKGWrg1fC8;ukh4uD#;@a*%->M-lshCS>ja)HpV_~diALmV zAH~|MJW|N**q?sB^>A&F9mTPp3mRs^SZ_Hz@=a;XnY3eafsX|`h=CxUywf8al?#U6 zy-!-a@vLA48$jd$I2BdgFIEama(#WPH&dafC7syhRR9t;16D2WL3R0o0Ps3ysXWSY zmI|xa2Lhvx?jTN1Tkj0?#Y|6nxlkP)}Mi zScTpX891nMR?zw?71Bjp$cw^8(pdBIo|Qu8)nrrmfl+;*(Hn=)G{7_ZX1twPO7J#aqh?rRp}$gpjr35<_XSjJ0Z zixX3w9E)3IvoIi!T7${n<97?5)zRGPs#Fos^#s)!Y}t8L2OTS>HC3#9>!jR_!yeOx zz$2|#hC-{7yc}eIT9M)V$AD@zjDeClJ$)%UnjW4T31@a`T1xMQLd5m?dwpwa#~xyZ z3{;Gttzg@sP#K6`K)^ilTXu5s;8I-jje6-+uv=wqEsD(XorPs)1J+H<(>KwE-}JwL{% zTig<31=|d}S(N?JR#rwvS0m-g&#CmPgK6q`wW)H6D-r=GB=e7IsVb=X!x-B+l;fO314J05%jF>!#9IK6bj)sdNK|P}yK|N3Lp)UNqVQmSc`;{E>N*24V>4 zaBym{iMBT5haCYKrf{WqWL6(5kcSPDGQ8%iN{4m$} zm;hAu+&z5_M4gS}sZkUm0o1yKz|R9UeliHhZ2c;!Lfd&Ef#J}8o0*1q9OIzpgIrgL?`70)i^#=<1cXkj^{(RX{w*`Z(@Nz<%!6uy)km#) zq?dUz6NV@6de^yx!+RyoEzg?A;@x@jN8v=0UauPuAamH)E2v&HO@Fpuouem_iqMkr zeAeB@Iu;nsbNZdZX&mErL0n_CdRIKO?qq3tnV#9BCmT*hL#bc1aI&e!Mh#%t-1%3s zp&$(Nj@32g(R`3jI#ljmx|ugEiHm0=(yiI*CM~Bude(j1j7jwUY7~&8W1#h?sLsS| zZ1E@!+ZYwiYPV)dNiDkus(qk>GRLoKhFLO86T6yS2D_5N!Q2jlr{!cPILYbjRW0TK zpOkg)RuVw+Imq464#JuSSqZ-*k~-vZ%`At0K?gl4VSMloJ5;#y2Htt;LlH)Fz#J2f zl~!fL6O0r2Rn}sKhJOmJDxBo>J*c*Tilugpk5AT^K;ZgQ++q1AibK%SMU6Zkts;QK znp_RMo-sfp9FCOz20k?dCY;I$2ACfW?@tT~IN)}tu_SDH1Cdwlaf(|3V`~ApoF4h8Bww4Zev~|=m=ba+fw=&5qWXdsCEUQ1^{>A_;GK3j@f5PK z`?078J$bKy;02`sjx%4O{{Zk#h!wRv&_GbggDiR;z*h7=Ba0V)O%K!boynRc)QV+8({-dkSc&Nki~jB(z%9~53As|hlyK^Oy*UGr(H9#*XU#r?WG z5hjnXXnr%cT#(;nx+*cZ`>aptUU%XD02NJrVH{~F-;ABSSK42+kA%DRlN>MQfF!1{w&NcLAMW4MLPZ@Z`?f-B{*^pw@tL+bL}Ltf)&ZLQz5Gc;@i z5(9gG`s1k$rH79*YnoxdW_>)k> zD{dr{wwUTpLXLCJO(RHEK*vGOD@3k&@^>|@ZsXneK^f%M)xFy81xFahYmMGPBpy2T zts4p2LAQA$&>C(nt8^hWZ0dT{N{XR!e(pykn!>xB$}m9Oy-jIJYN_|OjP%D^qjhS7 z_;MS8iiDAJli1G;RP63KsS}VoZ$7mlg8_Kw>DHQKY-O?3f(2HL$diMf#BokVCxE%< zze@6SAUtjIx=NlEU zig+V(dW?($E6I)Y_j3Zrocy3yaiHI8R<`D2D?PfV3rap!A6oV6VdSG^Wg4@Kyjk=Y zf&Tz&{YT*Tqis0Q*6gSsypCYnQ(T#PdPRD zxvFbzeQ;ZFlOw3<_*EY{q&(vX(v&jAI9g4ncSkyvFODiq_0QV7$66P}KaRS#t>KG< zbE2iZHqgpf0F_>G>MO~HbJrr6pyV8n!k-!f%rbG-z4}o0aEv=2GghOmQdJtY*|(wG z7^MS_M}F1x5A5q=ZN{$wLW-v;&jWWtE9XloRyoj-mE(b5TYL}GL>d&=U?^GDS2*jA z*{$!Tj3M!((4mLR0Jd32d}QXOjyTn`xNYbLIr`O!Vpt*#8EgT`6y}G_eZZW7oOaDP ztq%K+wzM1AZvjs#2U@oh29V$!DDFAMHq-8a4UwmwX&A-~<6@97IQ10< zr)9PhLljffo+|V^_pal?{VF|j9NCsq8A9$Eu6}QIo${k(d{^w8KRm5> zQeLM20E~6-B55t)lQBj%e5Z_x^ZV=OxMIYTX>}XpaXUF|?&hcyxRadH-*Y<&xa}nM zry$%1bLm2%Ve9NF^`)7Rj491l<79AKATCAu$4Qa%!A|EuTuzy_672oDe!zQ?oM@k&1@KmB{3BNWdrD(!vNN5$;VP zZw(xhbYPT`Evu=*Em$CyCK0D6khHtb@`@=sdMx04Tye7((VLdmlkz%=eAQu0{{ z$ioBq)!RZugQo|PM5?@P=bRpTQY$V--VI9BR$WfI*2doM-Ei0-V~X?t02o`FJxn1a zarLg6&O)(~86b><&OxqY#!l}%EW{JH1kxH>l(fk=ONKr3UZJ8sbnr3=7~;H&!eWwx z!LLs6+}>`TA>?csBz3CT9riQ)TX0RQH&r?O4SBm|oRf}gv+Lo;_=eUepy8 zWpiIg8IcBd=c%t*@XHpn3Zp!8it?=;OF9wLJ?qwdHyW+GWA5$a-ktgxy~k@4D=FcD zQMU)Zd8deU!y$Q*4neO%)Ya}Bmkc?^O;9>foNO_c?i&;<5v}e^@Z$K~Y0m+WZ;u2i z^sb)f468UfJ-{6+gwT}e9w&_$ksQYxk3ehDOs5#nd~@7V+d`e5r#Ennk}xN7k;k=n zKL~y$Y99>zO>3%IyGP}?E{cBg=aF26`~f6#FzUpU{?D~T5ZfR*!RdqPPo<3YKVx;D z4`_e3XTWP4yD9c-cCq=^cVLalC>;(uWQu?M<8qL%(mFcuda0Py-Yw^$UYsVJa^r$U#w^^<%OP`oT z5aT}e`8R*8$*FkDTDzIcokH$u*pZG|$^NwKr%q3^D-R_&tM0C1Xu<5Rz+srfbJ*g$ zy-Lw9^hrokSb$WHYm_jE?ptsqAH>&ZqsV5nNJ$$=#t&-cTb*)x9&qt3-L#Q7A(ta2 zv21;P)PJfBJ)x)%oTbJ}$2#-XRA zgy0--Kb=>&Lb5wAVcXKVuK{ZY;?zW2cd~s5Ljz!6j#N-Tn>ur0Y5% zw2|{IvD>$)D~>DBc9Xbclhb{C)$qSwJ|6K+yplG{X?{r#dS|9<>Bhv5kjn8eA2N;` z(z&Lr?0c9=^Fqv9RAavbBNZ${LJrZ$4M85ZS=R3sbw9=!%lW zt$shTS8BPesK_gl124?xrays+grh`!gpOk{D^0_z|u2J<5F;r*g zE6(0CP5#kj$0f7C=QKFCpenq&ot#>eWDXQAa(#Z3_^#QD65|Ku$mHg@==EhFBOm^? zUH;QCzX3?jK2y}zysxN6qHOGK?$pQ>63U|#ua-taVUw2}$wn&TtYiZgA=7|&nO z)w^9x?%W#~cTb{J6CU0!Nn-lw#Q^vrkvceoMd$CR4!c^(jvi+IO7=i&1X-j7Ril($>ifTn?AL; zaL461Bc(*8t4YaexnM%evKCOn^aPKJIahnxgui z6pi~-au2<2PG=QZ*zEM{q{1K|w_mMv7I%FVuuw-NbIoviW%`rLbnWR~9ktd_fJY}a z^iDcl_$bwBo6+4i@)?+O^KKkt)~+jw88OqY7mU^B zcJ|){WR~<8toO4L0NC0wf=5GJ)2J*X86P%IYA+!MKsm`gj`hhWV2t_BUh2THEPg@N ziR5Opr?SEmj2@U3)ZE6(tfXWCz$2;8TF$$*ZN#tuAp4+n%^cbq(~55AGj(PoZbk+gq_* z?a2oK6OL&kgA!w489$w0v$x&CoO8%y(v|Fc5J@2|;Abnih9F~}m2TQLQh6jE26|M< zZM6|qG2cJqQwTTaBF7oQ+t=`k08`T$>5r{%%zVxv*hyd*m6@}Sa0OP}KvJu_cgv75_?pm^?k5495rgb~ zDwN@mEJ*_cX*fBqh&5|i`MgH=MJaOgG7K>X<-x(KpDQyrA(xOr<&9os1gi`ZPBD(3 zjaL!wlz^p4I0X7tw{@ZC+o;kEZ{{-;+kwFQ-&$nKlBmw)$546`R^AX$V;i%K;F>`L zZO075pOj;0u30{ON4H`4RMuAJTZKkq>ANkS;-_fT?k95d)b%v&0wb*5 z>aj$@zR3}1!RUPly>;QK^VHX&#f+)!so8Zo4-@N#YpGz50V5;2&@&kM7`Mg0~yVI5?re1%U!p2L95*uow9MxGJAb1I^yf?Hu=vu z#zCsu_1K!=kO|Itta$DkEcafWdQ_*-*K-q7Z?ozyS2#UKy(3(+q3!L8Zm@GWAaZ#= z)m|wuxhFMfcM?YV+pz}*p^?<^M@m^3l1_cOs2t=9T}Fp9ow*%7X^cqcy$E?eo|L%b zC$Fsxt%ZfM;;dS!XK+1pRbgyN_|>a7+c9CDyyA}`GU_fEj+o}ARy~^;JbKh_y)nj@@ZHB6kvG401Eae&(w?ZpU-!il2KPaq@xG_o$c-cJ(yCt0In188uv~9rIT9 zWSsOJ>IYMdfk|jl67V{H6cd7Y{3+eY&(foFj(XEf2MwfVmvI~p2ct2TH=anUG4;t|e zq>@cBbSkBe2;#pz{{U!T2oD4NP-uKA%JEh^u(cgIrhLAa&36!Oz>+!&&$(didyLj@nRdx; zH+;Nw?^o7iBxG(FJb{i)eBD@b%INy1E@X%$OcBQ&aZorLhdB1D6J>$#7~9gS4E z&z^Y!dkplfy*}yVO^v`j0h+sqEyx5`Qd;O~eTI+$I6MlgFEV4){&hQ!af}01<^_u5 zgN$%$64W{mW>&xe4)oPUX*1J~c>O9M%N_t-$D!t@f;XAK3UYc?B!hDPl><+7s4#tr z=DPH=Hqaal^Mj7{Q$W$=g#_}UPcRTT=sV!n#;1QBm6J~UTWaA{@y&2mpS@_Soph<+ zR$cggpdS!;iD8oxTieGa)X9wEIMA^_rF`pua{At_cY1kg=4Zei26!KpeV6+tcvkns znum(6blXP0)J~Z!FvI~rGx32|zisb;I!D8Q4*W6HZ8Up*6I{DVt@Rjfhyp;02x7ju zBxbv?mElfqo86v{BMT=QT)j`5ei8gTo5lVhw}|vDz#}6yq4CSb7FT{EZx3iu#|D+9 z0SwYY*%woR^S96*g1sY2u#e%t!5wqObAc7}sIsssO0kwsGQUBSUo^v(kCIChmUGme z=j&dk8ueuAOJ%V0ig3Wvr-zTm)!*oYX;#M=6*?dt4m0X)g z3=47J>raoKG1tC&RC!V{w}5(zv2Pyq;0hZOx*o0NI&CLCYt?=l{8iLE39V?m?y=$f zr!ikf<035eap{WklAt&}N#d{SF}(gBaTIZfj??% zXnZLoJ|%bL%^K5m&HjOw_ad%YR$0t20 z%%xud518}2fVCV9$gSKrIVXTSijyk;05CZKaxt9Mrgu}lv@2Z1rHDBT(<7d13sQt8 z?65rX>s=+n4<)$yNX|RgH>t>90~^5Tr;o<4oS#G6!#lGs*bgv(ps_jNb*TJ9Z~p)i z@P2HC#`QSB#aq*3CPJYw}zZA8_lm2^|ayyFSH4BflTsvcTHKN%3+OP2>yvptK zvCRmepK4Wzv8|mqNs{5$YHsxyrj^WFNQ+Of^9bXL>~x(F1P;mo?eAA~KMxp`7Xh=+ zBc)BMYKf<>mI&UX0FlzP+e16KJ0d%`iU{`Y1FsixG zWKr1E-IZ-LBe|HgjK`e%R0Z%cO7G4w&q`*}r<#?f1PANF7B~ySRB+ocHURPsokV+HVv>;PIBmahizuq2&JnR0M13isI2^cO0GvKq~w| zzz8RhGeW-Oww52_4Mo~$C31d7+yeHmO42nGf26@S1D&7}1$m9i8@L=}01TchW58PP zlMJ>}wn`t9*ieOS4#M(WKgK7Af*AQOYf|Eu^&Cu8DZ~PPY_Lhg@_rufSeQM=Sx2DD-B1ZY7L%=opHL1XE zEu>S<7a$Kx=)Y(G0E_yU!M}(aX1ifARY?NOqmAjFL8jMWx%#2;kKxq*Fz_deJUX(p zz9O`krKmZ1rHu4dAKbzHEAf}c`X&DWg1l{cr0Ouez15||LdfGBgNpsd@iw&o01!MC zqkKZ}j-e&H-`a~i$F?JQ4bz|DAousL&u`n8_G-J*ZgfpX&_{W%-G+OM8+S8K6l3Rn zWaN`x)#Hf5-L-YMe>8Zoz8@3Hle7E0kCk?;s}zR?i0l3pwWP`-lY(0vtJZ!K{=`2R zejsTU_ZI#Uy0*4+v`=}67oPZTarjr$o(ul~f>vMKEE4#)#&%W>fLaR}T31%|EO0Br zl;Z_<&wW#r-J*}rZx?C)eXpFk+=RAqT)oA))GTwfpO=tDe$QyX@J+9S+C`(wtDR$8 zz}((s{#yI|sw%zD>|y&b$QMqR!-mdTNZTUb$gPvp1fS_#uNy@@S=9_JX|Hi?elB=c z@wDw#RfYjL=Dl$b+EOwXB1Fy?n*CDnZ~POx_HdHsOY1KW-D-;yylgJ7o#V>&I2f&I z^k3P-;SADQP2oKj)<{?gp)eWce@>*-&lg4(lX^2+cpOD*!ZB^HBl34o(Qob@TF%nq zR#w0;M9fFv1${yA_San3{5fW1(seJj#GX?FXC_>Z_{DwQKgS;gcqd8Iu11MrqrKvn zw$q}uL36h_1(Xk$@vLtPe%IDF+Qy+}cc;f1z~G~y87uf#H1iq>tNur;hsrULPEn5D zX!*ZO_;KR>RWk>MbxBkvShcGS-%9GVui5*@dXfOX8oPiJ2I-2f2jXk#h`(w-4ZKCR zxn%_GUJvD5_lo}jY_AV!78A5~@=HEHcDe8GNoE*(T|dCLfX6q>$#0rGtv_V{03W*K zTG*=NJ7v2Y$@~T?o}v3Gd_xeQ_?Fqszub!Pay|XcR+Ih;1$6dEO|7#sGX^UBo_#Bt z*MH!ldb?UqmUi&K<^EZ7mQn5vWb-JsbryB3vlSO@su7ExN;bU#Ic{p zt#Z1r?6vVm62-g|dng?m7e@UvQd)n(LOeldt$EKX%^8WJ+H?1hMyNFp+rPznq+;GH zr?JZdXN`t`$*kp$O=|{rLockJl_PJN<-RQVdE(CrZ~ofwzOknuY;T&$_JiM$2*r8* z(+PL9Wn?)kk<~qoeO;jb+kPvzaF-f{T8tn9%WwnacLzLIORfAp@Q>|j;Z0Xi)AZ{< z4*2fI+3sMxut5^skTd1TqlV5p^%a`03CSy2owLGLqfOLylveh7{l~<{;Q2}e93FCN zoVU1Nl;gHd3n7vmwo*Zql05y_2Mj*~D>n78K?HHwno^Ft98;Cm(8+ZyU@DW?d(i5NeUOBJ%|1B}N|UtYbv4Z> zH>v4Tskd{hwbk6XOatk+oK>-?JE#&3{)0T`xj3D{Eg0uG1Rg5%S5LI%xIAMy{415K zGufuineUhh2hADVxRKhZ+}<Y^tT{mD#tt?dR1mS7%mPm&fe9m71`z1scW0|H*6U0+k!H2 z-!;+c_e`J?LvlJEab9_8e9)`z1p6B3^y}=z2_;D!`+8T>#pr&1&i2mtdjqr$r=uQz zm7`}nImmLG&rY}^xcdv$3$=P2Wc26Ox(f@U3mhqJ-91eWqiS<|Go*q~Bq+&dURj6n zeib^+muFn>as~nR^sL3XW``KVoRWKWsY{}kEAuG9HODyhJvvoiLdDG6NXW_QkaJ6l`T$E?1dj343a%-$*7{wViT2-0FQeHRUJq?(?azqJ+_}( zxFBPUji3zUrb(&AzwXo!H+8N_q_szNH{9ngEsAs-hy?ST;#zxQroL;E0k|j&T5s! z0Y|CGImH*$_UP3zz?_mFlT54?aAdy;N#M+_Bio1 zRkSC818&0<$;r)56+-0>2Ve(Eh%OPBx?pw#y-K@*B@_U7t4*y>nypiv*ayk8f8IG@ zDhVOV`Jco%2en>~GFR>b%aE;{nrxAS5|XZ<@^>D;g=nX3Pa33Lp2l3U;4pFWw;+m% zZG#>GBLf3G3cDl)gYZ`b^4wIm)>}!{sUGz1K9qh~rtrcTuZU#A0de=4J{b_9V zFB7p-2rh#uf(}Lp(AMvX?v16wLZE^ax5|Ah$h6&1+iIG`Y#f>6Cu+Cx1I2wV1`>@v zchvc8!fDmyT?l+rH+N2WZJuU^7O!y6bV$Gib*hsxh>JEj=nr~LMnD@ReAuju+p@{GkbQk> zx|BOn8}P(nA4<(v*h=$SF}I&&GtSdgnR3})v}_Aejq>%+)|%fk^yAi_Ajs*+?NL8a zGm+>hWm@cr@lU}Z8e!zpj+o6CMOY93Pc=QO6@I;YReF3!4l~k$G;Bu4W;+h_j0%&E zbJm#)mP`y&v$FsUue};>K%IE#eN9CvDmdWcoL?j6Jawk9EDy{GrM7^`h}-}N4~mXJ zLUWp)LBf?kH9+K*Jdy<)kEWzddFKY0Mowwx7|uDSIH$2lYIVkNY4`+Yn;6R&Lx$kHRky4NE|^ZO0`P?j7-7kvyMiRtnC< z42|8smF#>`O&(44vpo02DIzq0?b=BrBavQb;%z-HEu=?M!PJanJXf%5aEAnOzc>f! zTpqn>+gr@rg!`D}n)B;wtc{@T_dhzmCiqGG5#x*7bsy-LE$6_&a158u?z<>vG_OkU1R)J@Z}_eRV#I3sc%c&!K3Y z!vj2e@l4oqJLkPUn&D zTFkRh-1P6qTJ+zCnqrwm%@JeNC46M+OcJ>v*Yx=w2 zHa8I(o-tdVJ-PcWzp0<38a>>t8eH zJYy%)yMK!R023wn!SO%EJ}|q8LN%M$o+yX>vyY-L;4xfOmfl&({447y!Bc{X`G!9e zEM_8AruAVNBy-ojO3nAj8%IiwSLMMtKD90nY#uSqZ)9)?F_D0J@=a)4L|F*H$81(} zKS09+KZSHUJ)oQcj!$X?^e9aOB4_V5Ju2ao$?)y-@wrncv8@~XBBSLQ?bfJkN6PS( z?iZ7^9QNj=jwcGYURoI$gUEU*t2XSw7d_7Gn#+PFmB9ldwGL4T{U|SErSR^*bEMtH zacgrWwZu#c$rE7b{443-h99*tFCCYRArhe>S5km{z#azW_OFlNx?~`Mf-7%N)L0O}w&Pdp0xJL9K=uRLw>en@TN z^AbB`8&GE_*0%f^`!mmKW1m!Cv&7v;QU!EIJd*}PRhNKn!sHAN zm7k}K$%B$lp{=OlE6``qgGG}Ep;7`>vPlc*Pg=VSg=OTDr>C_<@B+Y$xjjdv1~M2N z9CpPU30g3bXs2a-VMzLNM0{{RHS z_^snhXzn~g3V2sdkcJI>_?K_{gOOif_#^%aOW-XIINMtB7OARE7i)+%#H9YD(o#q{ zUqka6cve`6-MM=BW10!#{!^2>3T`BwiQMV6d7+3oYXf)5r(+ob&ipN}`Ir#;LpNc$e%? z`#$*F;eW#$?H9+l5Ve+|fTonKx=1oP+{1uDO-xaif2u6O@aG~>oUp#zN{jOsE&6?XumMJCN!y&mVj+phYDyCshDX8ju7<{^= zTRT}EiR16uV%x)d&E$ce zMTr+AlEn6|{{Z55#{FkOvbTd!nO%`YqSi$$9OEiT_NNSFqokr`{f`vy3XJ-W&&M7g zx6y8VJ)*}hi)DCOQdn5D*<;iL(>3w$#E*x5B={HcPJMsGH<9>@Pt+qn**b;VVk=|c z%wPHF2SHk21wJ-Gu4=~O=0atXOztC<`g&Knd`h_R@5J8&Y91EU7_2uRyn^O@JHs&L zU-Ru;v9EU3N9p8ywCFl5Iy*i5{%6Uz{w~t*?(Q{P>&Kc~f$~dnA!+CO58+>G;$iU{ zf5J-~`d0A^lBA9S72o(1_DnS*~v;5yU{)hk`-wO?e;0ZxdZc zZWcR$%n2D&k_C4+I-a58yC@d-Q=9Cs5Md3l3xG!))eAoWcyVM$mfCBk94We#wokYn zYpGyrJ#l!Bm_x5%K~{eqHM1XIj1h!>33lV+A}%t*EP#7-HTCn&B(e@&*+;GoM)TP#WTs`n>m$avDB^q0NL^w^Pi=CDf>74Rn>eMZGC;HCC#eG8ze!~X_2;^ zS6O|$i6mDzKYN^$_}3e%+nL%%EOw5^t5+>-H94M+9}!-a)0OUdz0@94&oSmnWoev) zk-do|ezl8p%k#Gc_O6>bvgCt+2m3~(78oQh7x;63U&zo=tBXLRSny?NvU=Tred-0OGZSN1IxUz1hme zbG}2nf_WZ-wCwLU5DcUa)q*lxB<1tQ2U;SzE6zqM=o9RIe$}@-yKDZVa3F!5#~BsV zY5IZ;Fk(=i2(L1*yun?+H>qRS+PZ5URE(Y2R_X!hYG+aGIIgFmTVIJ-97x2RZBvYU z)}_3|1UMy$?cTVJKKU4ZLL zT+OiqWlwhe>n`qP0hyQ*&4I^$!iDB^#!4j_NlZTBk~z-=4|=6=%<*vBlhUtE<&Dff zX&Zyie+tQ+nU#PHG5jK<(Cvj@RzWK$`F9*Atr4>ri>Vx*Gt(5=bpEPaS2{^#3@hZuaoRE0I=sjyH+5E;;Jg6AKttfs;7D90t z>}nc!vCTYJIxaxP0nf|<;B*94!2ojHkXUi``c-IMWE+&@10)gpRhX5SV36U585|G( zy({UfI9#sDBaE)x5s}}w<3YC%g;3ZB029Fc>dY3;8YsZv`t=ogAxRWA7jYf?R#BN2 zze6G&BM9FwKXjk#Q*`W`DML%M z99uV%10CZyBOPfcg>9iE;S>VQ{dmP|3t{srfXaFTdRHa!M^^sA@a@4;8RP;rQR;vC z)zt*&H4_@GQ>NznBZ7xg{@vHL39tlFMpixVk&Jb(Bi61O!(J{J0PIwiIO4h;4_X#C zm(eH)vLFn@lC{V9y@bELPP$O1Re_XtvPicwAx@} zaOqhRu3IBJedyJKEm_JG6?2s%pzTmJfgkT4J5*|XvJQBr$TyO4iV7#f$@2rpr9&cs zPSJpAisc&>H=3*u9MiaLSI0D?tt)Uyrqva=9+c(hZ8Y?!50tN^O@Wu~$QaH?sTDH? z`G!x|wLxrn#&epUR@{Vi0CuTu>_(R&W(PRyP2LIqp0tePZwDvZnlXdUDS+(brxeK) z7AGHGX-+YYp7iVuq!al1(+;E#es4}cI)wD8oqC>_s5?-!DIw7uII9m6Gj;1#;6J<^ zX0`-j;|CaQU{gtn*vgO)*lvQODo0M0eprUp7{hlJS7}^&7U9^YC}p#BuBF^^gR z0;^^hI=_c@ziZbKNC#?`{Bd5PaUH}8kF-ds&(gS0g*L)#`x=I2|kM9W~6>(e59-JQZHm`Tg)$#N)#{ z;#!t;ibvcRr#<SGl^>FDBG3-Ol@qmoD%1CZcopj>mNj zGQH+|*NVSsyRR8sO0YyUy*LmOFTE>|R&VKCrmc0RYxbh#i0yC3K4d&AkM@VHc?M28 z;-wcdr~pXV`qvb&tD5PZ>Kv}|JBxI^yhHX|Nv=B^k`WB@P^rgJy=WcrXgH++B=2DtV zN2&B@!GDGF>e{`rSUlK+bA!n~)uH=i+$6sc^j`+rZZ=*Pum(Y#gL5~`bM6-czK{Ku zEOeQCCf+QZBszVayWGy$;RI?~x%?~SJyQN}8hD3W@ipNsb*$dWdd0ehWAZQHxT}ig zh1usuTDh~&CeWIDe5fF^2@$1A5wu}8ZMy}D+-@8J%cYpK{gXD=fFM;Onwcuu$B>+cNB6nAQN zNIpg%bdT|`Ncbh<(PQFEIOJt}V8?Q|7~}a@&Sn&67`UG14y+|4-o@XFKMA!RM_BW; zn^`VN2QfwoKZRj&djjib8x!?j^t;peQBj##dfPBE0$r_5BXB`HRmYA>Ke zCI|rW^s7aSD9PO2u~d*0eCL9B2CPRQT(RMQhZWOp4o1hu*ngM}!@g-=UJ2V(u8Brj zI4jWOwL~Pw0b+ecD~!JrZfqQe^fl<74X__&Y20Ihk9zZa8AOhuyB?M5ULM;XmK%0k z)YG-dY_F_B5>~)2qd8D{u4~2T$?)69z-Dp4uBTIN%1$u8zV*iV+EBW3FFD=@PebcN zzd3_IT`v?3nov}r&CqAVHgljFgw-QCC)at9e$OV zvUddIHEARsT8!d=*5^TCeDTOI!~@1Frql0~U;qiu2t3ybZ76rf2pIdvpsU~83vD9# zK(mkeX`VArKs^@UR(}uNCE8ufV=6ZFk|!$PfE{ao>-MDayIHNRrllFc+4iYn4K6xY zi5thcxOrxko>G4E8q?FF^PC>0r;;t*l?s!m@Vgyd-^4!_>TYJb@gAh^f4eb3pP;Wo z@E`3B;{N~&N9EjJKD%eTEopq&;~4o{sQhcmF6BrBXs!uvpjQ2lhS~I6jao~{E$yLT zjEx}k8IM2~mm;-|KCw|gWhQ!jpBe0-@ssP8c5Z@YR4~ZA#(Vv1U&Wua7Ay)ym6g=w zi#~D3-wN`l|z6Z@s;GBHN68jpLTQe zaa?S(yzjV@Pipijyd|dIMZ+`%3}tz&seB7@Zs5d4%xCU}7&OwHC;-(z;7O2~G$>1CySBbTxk4!Z$`jf+7#CYsybIbDLXbiY@@aHB#xo91ttClfb&H z;}JkeKAyEe{50_dtO3)QsXV^mFgRfuCDc~TF(dOfd&JSsp5CC$u* zfl;^sSHB*$+xUOsHKw~d!>4$1X$+0DOFNPW9Fy1&tx}E5ol97Lc~x_mtF#8Q+x4YEXp9DsKoIR?1z7Wfmx`<_c3 zEfoR7Ff{o}bGDW~a=3s4xG6aFt~*_Q^JB2Dr>uS%Lw&f55R1tV$Of*<@KeK(LK^DI zIj1XJT&*AY<6#zM!=D38CFxM#E1_c$j~DsN>$Z^zVXx z6}0mfZ7pwGIauUkM^Dz8OSn4Cy%F)QnQ7)qrG^w9eXFXH2@E1QQb`yN1Zqxe>xg_G zp=pSs*TdRd=AkeU zX&3Vb>9XC9e@gbxgjzs)+r@0#n}>!TINFB<=iKI|rZL7XsHNz?&fy#!Z+XeTpZQq%6T#o`N^gbU8`R^v z)b&kMT#7aXTgX-*-|rF0uc-V9@Snk73+)o$!yX&cw0O{C%z_cT{ao|<*I$!a_p!9x z0Cx2?^HuS+9m#8+nA0?gV}dUWf5s zOLh2oE@UZpqsG9i?#bkFniQnfq;{$?j?qWUA0K`sB?w+-I6E5|WqVlc}2AP$2*)#G=1(=?<9$Q?6Xk);-{iuRI= zw2p&Y*KRdkpJM~tlO51>D$Rnl)rF%U800sX5XN^AGe`$P7@fV1E zqC<-a5*VN9oRBM>*ZdLiBTiJg@Z#!`+yH#J<-06*b|ebtYYzS?YIROk)KZLWyvz>? z>i614s;1Fc_B#eT*Vf+zFX4ju6U7k!0BBvf-5spQ3l@IsA6`3G!B<`&o5NbpTIKGc z1eoN{QJyni<>6nBHae~SmY^8h>LL|}e=)#T3OFLT>e7nUA=1Qqyu9A0=?{h?x6*Zc zYlEiRUFlEqJ-aUB7(c`_)~|d@_(P?5lGy6{OmTVeDKBX_#B(7ZGV$$R0q}$NgV1Nw z#Fv+;bRh(?@4^5&2`9l5wvxsKu5)*NljMUDXV$4c6@nE+Mql0hfEaUbweZ9dK) zkM_x-ytaqz5nVN=IzhAgDjCi7+l0^neg>TB!iIm(i9>U>O`VOCIG z8R+S9+7aNYTDmBulhpmiR;)qdJnX>*e5a56g|dZw+Z z)fxShcHbpR?LiowRxPVv;;B7%nv8}RMA_yeAwJDGBfW11{EPrEsm7nQ1}Y>2p$DyBu(1&{2HM9N1JbnZ^qr_xp#@KD;~n!!%IfBk zi|leUY5Ny(2Ml@v$g0M{*|2$RbtkB<+T%#%5(xt*o`$M6k_#ySk<$Ql^`@rYjF$&{ zBy$4A!5Hcfdax~oWCmQQ7|QZ$5?FW)ZG2GP9>BIor&OqdmS+=)lWjMRo!&@d~ zDggsO#5!iE%WE2zCxAO1xUQb+)GtMhS+207JA%ES0h7H#Z*i&8*gZN25 zF!!d1M?ry*qXRUXmwyMK!0t_4cRB0Fsl?q|m?U634!FUr=VEsO!T#-b`ks}z2EZNo z>sIl#sM{OW?+GLf7DMQK_< z%F2o{(38{aM5bd1+3Ig;?5YE?Jn?{fS4nAehTxXS>)yFrSPCg1`JKgX+fC)B?BSQF z70E4K?s^oLE~igrJ41|seNAuJ-U$HvsxjY+;$XQ}RSCEpgUPKZZ`%O5O}W9(3{Pt2 zlpAMsDp>4=+BzMff@gvT25FMr0Q|PqJqYy`gKMZURwSLh0bWLGJG=87x!=!Xr;ny9 zHBI+ARa>o1Te-k*nDTfBu4;ty7T~cb(RyaHrPR+m@OuH?t9`v6aRjjkjC2&-9lhfH zv{TTxH2k`^P%ol=^1C!k}T^Tj`9q8r4b(UM6) zK*OGQ5Go*~kg5ha3_U5aO@(p0APk>cnhcg4246$POHFBGCtsUZUbP-r0CL3gPds+5 zI3f8?;6f?LInGUI+K>}F!oT-VQ&*9cI;ka!9(^)vk;&YpR_exur_9W~9dJ(mhc%@R zBm`|Lky{J*Rl5_@-mf7?IbWGnW7O9) zmthGzv+KrB-IMj^!V5IN6k z_3NETTg5ZUC{h%#J9rqcmb_c6NvvyEx2(kZ(jL8l;18vHnRPB|yzimmX4U24$Q56!%CE9x!O`0b8K?mo{W zD~G7(v^uPG}3Z?{b>wxw?IFYDk6|+p^EqEPTqD5el=c#d)RE43Q6Ls6H`vv z9P#Z?9%*h|M@Pf2*18De4H!7b70&~L52bWgj!Y;+XCtAeDUeGZM<1P2X!_xDYg$Ah za#s~fKyq*~L7v-`N?>4sdE%N>1JLo>tU8cEG=Xy7p7kYmJxAdEnQI+FeS>IaRb$uI zy=z3ax0Ws2eA#-A+iQyWL7<~}HaVT04aLJ?s-Ktv>t2a#XqIDQ#Ljl-E&Q`wl_1=% zdDFX_vCivPZ)2&079wT&fd{2C;t!9tKO5QD*j`JwO|n2%(SGZ=>h0=kPZ8M^MV@HN znF;DEl)aoNRg7ajMP(X#TIEF}lhzV6ib+BR+A=UPUSTsAm6)95lbZB@6&H(7lP$}2 z8Lv2G+t#}2t0T6Zxki-$3Q(n~Pk(BBXi3g7{cBTOk?c7Gv97zpz7CUIoI5Gj@4yHU@~#)G1Sz%wPbVFsN!-~ z`UG~$V6cG$V{1q{sq__1tPh>#D8N?QTRAz;HSJ#%JSV37EYft_RguoEbi6>z^Lw#A zmB9F0!&iFM=#CgzDI+qT0Qan67tN{1&vkR%d^h6@j|u+Bn%04AS#JD8sV16-aLU1r z(=hhH1M#mdYea(AYJnI~6?TroyMGw{mcq+L(rrO#^ynLWglEi^y0`SMHbs*8wv{KI zoib}gx!!@z(d=aXiKm&c3l<<(Pp0^G9WKl5my6|-_jh{wn%MB~h2w$-ySdCrI6?HT zYsY>pE?Uk4268z4Xq`n$Hd~3=`W#=4^+eXDc$GZJSLO#F%Dn1pnY=%$FWFn?1919R zZyYp&A6Gr=2gDX)dw1HWa`H3Z>0FXIhD>J|tD%I$ z!Slj4XQHdrsT($Sf>8T7C)^Gyq|kzgBbv{#zHD+%4;if`kT@&bu4~b`gSuv9pvE?k z4{=Ou$^$ZjBTNcl#3b+0|~<88EsGq>l-O6(U?A`m;U&P93OjPE44xj^4^5=AB|Yf}Yg zVx$vHX6FW*M_P9_u;C#^#%gqsvhvtpYO1{{#k+b?I~PsK=s1rZ1u95e03Hu+^+XoQjyPhs5E+f8-k-nQ1@0l_tZ>QkGUFvIu($)~2PapW`+ zH`C}$3;ZB)f8P^ zH{7t#0Q$N9lx$6n%_`F7><6}o$%AlTwA{W2)K^b&K7--I98=#uqo@qCf%3%|^aJ>- z%(WjCU+Whr(#eQpBg%m|NWka-tcmXCl6H9{EbMYkE3&uT_9(s}Y0foe{41QF z$6CGnZ8NyfTp!YxMDQl5txS=_Y~;2v#e0v#kJ)7WvuvyWr8J|a>p24oaVi&;qUCN;y)CXmg?49NJck#s&Z@UF9>`z@UOyM7HvxF z?6E@%ZfUMCta<7)UN!NH_RbL{#+Ts3_H=xs+!MQZy3u&YViz>^y^g}y;Mama7`!K1 zxVN~zWXjtH1-kSnwRJZdgfi*E+Q}i)?WFmYZC9$1(4O`2Plr53d8>GWZCg{glIrF8 zFDn+WQ$G`~gb_s}NUIXzSRN{GB6&4E@iZ+0(#!2v*T}x9jJf=44^#M!;cXc}vsq`8 zA#!o~{VV407W`4UxMxVD&lpxG`d2p=yLGEvD3U^VbCc^)@3vnmvGm#apW*wN8+=nk z)CBo{Zv0fbhs2);+*ve2-c$ib@Kk5;_2#~Mj@-_Fix}z+U7FS{0MtM$zdF>r73qPv z_p4mVU9nj=W6|dLo284{W3aTE8zGRbHw=FY8{%D>EMn3YGNHpeu7Aj{9ygzCpKJ~7 zpIYA1JU4l$JZbZ7C!RXz-i}uFxI0H<(!b$8n(1~%p-9eEN8ZV;Ehghq)XStwM2s_* z$qUUh!+#Gx%@Fe<+Q0@UJb_-RpxD{k!?GofS8D$7_03X>irmf1n`B}5f5EpCt6bcj zfG`D{lHRzlD*dQFD;oelA9#JlQrv};Ql1~b6n&%mo~FHj;-AEuUjTSE_ga?WM6nE8 zz={60L&B*3wegpLEN(ni<5a%X{&ekT2-^}mR-8RfN2KF#yx!5>rUT)wO0R!upPErs0dzy*Gw`u=s}`tQa4PWtf% zo1w#D0}&dhR#VU)%DVjl*7XfS`t5Jg zBS00mxWcE<)(^*D+8MN+Iu8%{Ys4o0C{cdd47@3PV~@kVarZthx!13?tv24~Nz39w zs@&l5fmggm`!$VIOSF4?X)X0XFMOtF+mJ``6I7uYKO=s39k)D7!ny{n@r%bAXNUCI z8TB1j0I@c<9BqC6r$=&-A?9zieFCf_lNDR>?4Cpg%@m&bC*83 z`qkfrUlo?yTeiQ{ZXr!SOdmUlIIh@rH}zZ;HMmz3|nwwWo)~g4We7qw?dCQ-dI1bDUQ$3X;Xtl{@X! z*1j&T4$^;dtMPBg&GFCT{-@%GO^`~l+gheDOfm2Y_a`;kX>yBeXq6WTsk8;{it{f9 zd2vi#uw+s4Te7Ht> zq6X;{EL(faq|)D*i}|jh5!O` zTR7-w7a$e{V?QuZc>2(n8SmJVae<2FljVAyaCcf2FKoQ{fCC`llU)7wo+TIx7aS9u z*HUB%2?T(8sUOO!Tw9SY#bDeGqzrx)=TzTQ(5CdU$IGL}&Pc&O)7F}{2OE?z8OKhw z*QS~bhvf%$GRL(k)0tcpQZen;xaCsr&7CR+6_%EP_b4(R;I9I+G{CQdjzQ@dF0wz1c4eB{uV#bRaq?;E@Wk`15Z9HW5tsBK< zh?2XB;A4SVB0{Ge9!>~371OI)9)upmoCPXK9{9#7=;hQfQME@co`#@p&OUAwaB<$J zk0loYdiJVU>P=LdX65FdHn}Rx_-7~5urD;FZKI9Hj8}1aY!2iO$~R;Haw{V8$&`_| zdZ#}x^{t^c_dY76G;wli%ES@K2Mv+wQ`^R>gBS#!3F<{`O>WXRIbx#?^5UY76-O#i zV>t)CU$oP64`$S~BZdMG%DaK*6>=yWBP)=4;P5IyyCDD!73Yrm^{2+|<)_G7w@OM^ zRz)b=az(-&tB;uabf7w+0Jj;=eJYvUt+iO5dgrAQOY^Y=GWW+!Vy-KyWH}>IZp!0m z+wZ%ADy_})jFw_YY}Q}e81TTekOv2kY9^Cm1Ak6BR*5B{4o=&ajpIy7XXCKWa%t;w zSTmOkx1c@iAl!v>f+@dh05Z1LOa=>7<8@;mcA2KvBn{?4$@x{W&$ULbC1!SFSardw zFEEjnoNiqFyk`cbNkXwX0eBpCsFkcrsM^%hh>V3@%00Roxfmft!wt{Mq+klqhC#p_ zoaExHq^jRD5=KDzMRCox?-Qmj`#l+e5wposs|tZq=_X z#dmT3HrI`YfpF^~M`YLv@_DbGZ)GUQlFQegwdj8kwWheeTSYq}jdIxQ^{+d-A&Kq= zex!TX)L?1*I(L0fjm+zNYSvpL4_03?A~CxorD9p$U>KN?I%Cqduf}uL#@uJ7E0@$2 zqnNs$dz$o?r{{;H2l; zpK%zJ0)PYUR2(`#JReGYGwdfPudN#n!x>sOBR^Vj!>=3>RvIxl$RnjwWihr7QA=`% zGHv3RYE>cAo}!{4gU=M~2udj3nkkJ9UZSObX@FpgbL6+RD+4UsuGU;*p0#0y0mom^ z)ktzjO1LscTc@QYNfWWbBy&)1=YdkrI6Rzs6H$ORjz2nnV%%^eraDsY135oIQLY2^ z?M`yI&ssppWjP|4PSHyL0D77Cg)~ywQ(Xp@!|dn?9Xgulfz58|BpYx6`Mv4fBwcsh zPdpwfv!>#E4wYfRSLd8m4tHY*IL%AbLp>Q)qXf22b4?0E^ffCSAL~*+-$`$Q_MTApRpoD{gtOjfmQ@u-aX-f!J3Qbr=BSjGF7dAjjqSnp#cE@6fT-Ju2VI}nO#}czUovJ-gpr~NAZNLygJvgtW ze`gPbuXq0d6)p5Y#T)FkiI+KV@7AgpBCSqDU-(t>bRe>%C^lDgdTT%Ap1ct^w;?YwE?87`jAId+|bNd{Oa z)~cie(%;E~B#JVC5}*v@-mJ%G6|`>3$1E`H+~?Y`wVVF{GD8i#t@8|$M*_HG7WHQ8 zIvCBm;$yjFXKblGc+GVl8}PdEadI}G;R7CjDv!hZIc6^TB~CDS{JlkA@fNopk7#zV zX$tTcsHZxSp&NrTz9jgDEjHFGSRAB;?O?Uzvpe0p?gX*0Q*@i#nj~smnnzjN{U^AiCT(_3xb4Jd5SI9)_oaPTA(WdL0JSO0AL9_X46U zPBYi8MM37C#BKwwDtNe1Nd>-@M6CuDGH^j2hJ@UD^Vt0=a=dlq9@LEvBrTE;wJ#uV zsSUIy3&5k0d0OeX}%Pc0$M&YI^w$+nSbFQA%HkxgXn4MOEaaN zqf9qUR}JC^FMLLN3hQpCA30-@o-^9GokBw`beXxz{Mz@?<6rUQb3 z)Z^GxqT%tfJfJ?64(=HBKJ=?A0mU1N!^9FZWP|?kr;w-!b|=|tCYjr-5v|HlV=o?@f+Jk(Eb0)N!GafFFsgNLCcuy?eDVvM)sIWQ3~*pB%UMew9t8GBIXs zvE-6^)Md%{WxzZNWKccK!B>`J(x+q#XBq~>cNN)uInk}9 z)RJ4ks}g*~vB1SNglKpN;J&Zo`NKnJBSb#)?KmFhy$1hbGp7alYr+9v|`7 zg}gekv?!CxTP=)W3cIWR)E1r{vu3h|Qd5#*Uz)&-s%!gz_T%L zT=uNma@~T8Nc5kH{{S5I?}(RDeY;SeXm=6ivtdB&Fe}gPL(eGV6`qN_gRw2~f(?KCow+GjW;uN`T14O&p8QJ5!8=ZsV~1ndZ>b~SKyNquJpowMMmDli5p zx+eG7=Oak%5IJ0`jFX;8_O2^l)g{$$yts3-C*`ew66*eA`Dua9Fs7^ce!|u%8s_QO z4ai=+)bm_Q4Qh424}2cFfQ_dK5Ko=F%Jyv@*GkeYQF5PX$=#pQyc%yFMW>cn+#+<( zx20v>d`h^LfHw!q2RY6SM>j(GF10?f8pWA>j3PGxZt2HgrDbZKB7)qr#W^JiAO_%Z z*1lA@_^WdpW;=%^NB65TZ;CpV!niXw)1BDpX)C=Llq_!iY4JSXF!5X(dd_9EvHjyH zCuxztz*gVE_#?LQ9gMqqN#n>PJZF<$W1y3K?>msvJC`}HPViCkE>=kRWMv}<86vTi zxnzv0c0XnQ68uGq=fk!U$mNR02K@CHuSK86_He|m#~|y;*n{cD$ z4bWG!MdE2Llt<>0n|NWyYtoBSS7_q9j+Z)b+9$;k_^06Kk96Ben{1jjzzj!M3!cA) zd>h~o9Y+2Q(ym@qQT901fH`77$2In3(4ALL)U2kHd{&Ry93+Ya-_^NpeZgCp~K77g}x5xwVy`nK&mW<{pN% zPm1NeR%?ecI~*MJs&SV!*2h*C7fvy8p65lZ__Fs_Y#Cgr9Byvjm6aZYrr&5@zy1mf;SU)2H$m{;r?21HJ8RdH zTTN$T4#*J`Dx3SC=U*0CAVO6})X2h))%Oql6z}5}pWtz-;@(*2wF)5>9e`E2Y!5 zagH)DI+MXQ!dqMzfdF6;)OM>{X0e4_mMjlZ*jJ@3j!0>t>sDH-bs;RO19CCJHPTt@O|Xr}BZHd7T-`RA=&))EH*Lr)GQ=DLYVuyJ zak29oEB8^GBDhUARSg+d`G4B%YhF9S^8DGz6`Ng6>CEb2zIfb{-3NZTHE3L`pj?no zSjTGTBE4>`M&sAor(4h743o5j$m(%il_SixJLyl9=sfmsnTOruVxHdBGP*F|xq}hk zll81vFAx%{aks7;zok!Yd=HovmpSN9f3117cK44)l$zBWu*N)+6}t2Ft1A;O58pfj zgzzgm3&9WqPMGA5d94Vp2kyeQ% zjos@Sdy+G_;Qsl>{?JuXsx3{6iA;eKLIbL(8@r2`al%zj=;uBPQ!zDLa6zs9+( zO_i0GZ#n0lwcQVYV>Il}OHd`ZV3p4xb*#pXkTUuV=Nx9X?@%eaoz^Y=#{ho`+DVu%Pr7vGR--Jh<0bLb zR!P0a>57{hutJ0#Fb`Vf zQg-)F^`b;=8zgyB6}x|pOACY1xC5pzGfZRzw?x~JK|IrzB9W<9IP4pmiAl9?$3$e- zgW2D1?hkR!e=2-e1{C84De42;Jt z*~kO2sU>p^Oy)oWsbiDS)j==^f@mYQ=)RxlA$W{Y*KK0c~mpf5LRjVp+a%TnN zJLSB;x}DVg!>R9FeD@+YVi^8)tE)L7lzhMwh92z1x8Oav14h;j0r#khCyv9t zNXh~0k8D&7ahx1>=972ORvb<`5ubWO!vqo2r8Nl21ZIX|{{Ysa$YgELIHVM`l=>Ec z*4~OYTaY<9tV6)9jT7gIlirJpTN-kWtMZY}Rh2Q%w`#R=1ZWs%t}%+L?X-Fm+KZVo zzaw`&`4zR`>ksWO5$Vt{1sdT2By<_gW3rFE$u-*iD6?CS6xxDV1O`$Ee02S2xTLf` zhq}2%X-M0*JOk8Xxs7$c(J)3}NcpnIrD;j`814DJ+o?6qBvWe|nxJgPI6HaGaz;0H zJpN{#%bp(a!|RvGQ!g29pkSQ){VVCOhhG7e{5gNAUw-bw(%M^@WXB4srHId_1ycQ= zz7xx?Us}c?ml81CF^#KVN&ePPYw(lyMe*I$qZp-=Pg@l|8+@l=$KXwRv8JI;<_q_g zQa=`TOW~^Om)Bv73GO73Nk76^ zE^K%o!k0cN)~z(lXqm1q)uy>%sO-oN z;t2KYUq5RQwAYU!0kriay<;`Z?#Npl=BIq@-db)apCt#vG>DPsF(Fj)LG4&g#**e`b1jk2UfwQ^5or{8-BQ?NhxLzD-w6(fIU97$7)u|4)X7eVB~kNBBOO3-otMD7?Z|W za&g-SzrAz%uByP4hdyHDVR{O?s_G(nogmIS7U^94=ddhE_3K>G?Dj4#Eg73Tkt-9H zVmf+Ob;i&MIAi*cdegc&QUVni<-Kc~*R=?&p=CpdJxzAOJDWv0Xi~q`7Tr!5H@#q7 z>Wujj=N_i1++4lOlE;q1q9+yT#tCkA#vG!OSdw`(|M{IL>`4;>K75)cT&(miEDzagsS16`;{=3d%+Z!4&VI z4}C(yAtF3y4Y(1N?OII{hd^lGc2$Q?yj4vZlwNE&T#o%STRPsuq&$iS=5BC$Qbe08vF0SOS%KGzG}! zldrWjCQPygTmjE|YPZTqV^QatDIl>T^~Zl))TqH_9XP7{jVN+6%>ruXo8A)R+lsWl zAh>}%xZJKqV1^^TF7kSwDO%xTY40Ssc4uwJJxx9-L|}T=Q;J>Wccu*%0HO!Sx*djscokxJEU?w z0ih9QceXR;QMN@~9`)#c8q{=c0et;R_8DX!FhxCU#Erxe#97)=#|yt4)>X~g%6c;} z=u2jyr_gp?Pop)zjk;O}hSor+2Lruv+Q-Fz?Eu?lI3u9qynSR1_m@7kAu)hR0251> zKuI03@kP0t%z1J#p1or!-^!V^b;n%f#M~5Zy&UrgVUKL#OU1Uf zaylscM7hQghjV{{iIUaeX{{UGOXbU&CNZ?m}466-^(oW3taT%5)4QB;(%0IKe#2CC6;)cK1 z?PjqH2ElbY6haBkK{f3E02RM(O((-X9l5vg)yx{_h-Zxbjyr#v8*}~n8{4q;ubjMV z;@wxqwo<`&JXY||n?xed5|;(w)RMRbbe?uIxdZ(8 zy>}TTD~>QdY6`x=KDyBn8b=0a#)^DE6}9}a~hQ`Vk`Gu z?5nrs1E&N30Iyl>zGRFul6n!=p}4(SSdovHjiBU;ulCXkbMh0FE9+hAozk&CShl8` zI7s@EpniW!zil8`Kw*>}7t*m(QnC<8!*icXy%gcFNp5)NslwY6EK-xy^tmBrW+w-q zuQg8U*h?P^#^0M9Vzfa%V{F|RIZ@AQdPlvOZOV`V&rH`FZE8YkocH!*VDAN+k4JHu;xs)yX}o<*k-U&&tCb5yMd03u$C#$EJloyS+y_%~)+W$vdzx zPdTlmgbk-4ob!{?p4ROeI}}pS!3uH(GG!YawZ)dy#?8E(isv;eCIR-0FFX(NuHyIi zY86xhM?!O-dgiqhF03~4dE=q0QfpI^X=|b9cky{-0tY z=S{geZgoM$EeBd*4$uiC@Ns}C7B4RgjFFrGI{j$#6!5LJk8E>NCfM==Hc1CKti762 zI^!j4G$^)b!5}kZAc0ZJxj~jJgmMQMtATc_s&(M5Cnsvr(Ef+jq$K#ZbmATZi%o#27&H3ka?n%oYP6gJCrCKQArwIUD>4`bewGi0p9AJU?52+ljwyI}BhNLxK|_)rL9_olZU`K3KLrkVyN6j4mD zJu73w>=mTQ{A(Qat#1#o61GJ&OW54Fk8)%j_No%bcel4aD_!9U7{)p7Duk#>$tN_o zGAZmlV*@>fdRM`~w$o}CZZZ`FN>5yq+Pv~O%MsXm*P?h9L2IPAP#4UTu6ZC-%S1&_ zg!F&6JnmVS^{2H~Sm> zD2MwR%Kj~)GjBYsR74ZlB6{F%R z-gp~F@cy3ew%1ykESq2Ce=HNyzeD~d_`=)5o*cSGogH3! zjeraCf;v~oU+_&26() zk?@yCjTEGstn0Q5afXnbb#H1_B^ay7i)*7eT~fnO_R1t`aanXE^><;(E@v zb*kzzUfgXnIXF}N1Jb3}{6&A_Jx)u@>DpNspOwc5I{{c%R|^u!^JH{9^Ij!dj#q4@ zwb;pm#tXZmjAt3*ywk-R$Der18*NNrAUPSWpAu`pM9@O}n-~KeW7q3kYTGpOvyge` zBDhrDO^Z(Zl;$EvIDBO1r!CD&+#U&94f&W0C_m*E6sH)=f0hZVpJY8R`tzc@1|u|I{~bf>0YHMa;vk_ggH=v zW}F5n0~79ME*iZewRq=0xl{VPvSl$HbDvUA$4S)brbkjM5X^wMMDEsR`n{AB5UXhP&gfGf&iF zX{~2+uBvgK4JkV-o27jb#~+ui74#p8e`h;esU(+0yK7Qhg@WUh1M6O4to$SKE{<5f zp7TZzF9*yq>sZ3HC$lt_DZ_Meba$&)8g0$16_yu`hfvB#^Qb(h#&(~|w@k|8DQR{Q z)MA^`fViadNktS4QfQ>3y&(!HJW^7gY3Kn(B`$cR7Zbe)Gzy*H@GYC4al z9e5pR1UPp0tB?T_XN&{QRa6J-QcExkwNE|i+?FNG{&boC5!=$CCz=?Xnr#CojFktI zO_hfj^r*q~?Lj%CaImkJvRt8!+pSBKu#h=+j6* z!aLzt9)ud;{6YI$c*j|lTTMtbX%)9eHMu54J%~MxYfcrU({VJ7I*F+*534>0{?5AU zc!J{h#;{!YZ%NVB=hv_AlgozE1m)Bd!!u(8>t8DV)Sn(Me`p_!_r4^xhwPpmutn3f zODteRWgp6%dbi#`TASjZ?N{-0;~$3Yd^h6{9N+4iJP5vJl7)pQBcYV29$_EtSC*oJ zoTwaQ73w6Dx<_r=Tk2uzyxcPmgw-0+)l?*y#sz0pmljyCQrB^T>s{8PFt@P`&9{Is zD~r*g^KYXB;|Ddj;(M%f3mY(fFPMpb`Zy=b6V4{>?EgSlMzMNk1BRP0MtGzXk2L zh5rC*4ShUQrX|ON=D(6;x`hM4V|fnWpP=B3b6)5Bdwey~e`jxn+MkI02Wcgvc!mY= z{lsy$VTL(*GG{$V1OY}P6+BGR=Gf+;hlK^qqrUgt{JgLO?67?9z^OaEM{0%c3hGg& zKJxR5w-u_~C^`+BYYqzXj-xcIO20d*-lkHRn5PThxPw`j@5Z}6A@+4 zPH?B1^J`sQhI6}<(477iq`n{#$kLbdSd3)xiu0#N$E{AXX*(V7_K@0mWGR8P^O5UN zTa^P4;H|h7QG=eitQ8w-^*VjJ=yg{bvD(hg z6k|Jk{V`ehdWkBeu^{qD;MX%Yu2q5|9P!hwQlD0kkCr)1bjN(vYrRN7MlKc|?j%KIWD3W%X4~^Q&T-POi5;>!ocHvoloCaIXM3~Q65z8b8QKr0QC5+Ia^PV| zInD)hF=`B`3Eav~(8T=0v~Beb&+_D;<1Bigaa=Q^&q|$W*yv!C`J1+*Gyebx;{)2X zB4HUPIXL8FIQKQr+upf9d;q7Z?N=bVK*6wA8Sjjm%~96RV`W>bH&Q_=mkO-Few~d% z?8LUif9MV+vb`z_nrENzj56l>q{w_BAS1EYg zBFSycM%}};Z(U4G>c?)y?0<@~uFx_#Q@jE=Tng!hI~;SUn{4IpB36`~pP6{if5Ne^ zEymIn2~(0!Z1)w@+{1>!#s?cuU&6Adw_w-<<|Ln+rFJMLaLx2*F+4+WJQ2=18n5O3 z(4c1-+&X5quWkI|aH_-}xa(BKubBS;GV-UK9AMWce#tGiB%VQ11snw5MSE2Z$Q%LXa?YAn0$tAm* z<2*vB@`(dvbs=l2j8svn3N#Z~1<;+EitGRG_)RQ0YUZVo;f49rU9R7RRf-XN(p3y5Jv7n=}axaJYs@QF-<3$014)h^ra)- zjAns>>q$iifk48wGzP`XbBqk;vVqNN_80L{LW33Tl6zhU|)oJ4j`Qrn% zTxnHe+l|DY2c=M)lh=|>GbPxmIT7-|rFsvCqBgo@3__q~+@l=i*PX6syM{g+a{xg4 z*QZ3kyF?)EMG6#ltSUB^h7~2LyW!BXYA_e%?#?lumHN~BCD`A$z%yl10?rCUK zNgrf>!Z%KQQLNjl-E=1h&jS$SM^HS(wY6U0QH5j7~Kl}Zg3 z;Ox&&m>T!*iT?l`#+&wyE5n;)xKEN>r5`ML9glNbx{4_Csx3=@5dJ3VUkt4p+QJz1 z9eL9W7I3it0JQb;=DDoiYr2%WhM{oR*RqyposL51_*c+X?NUi5xRzfnr+z$Cw&Uy*Ubzid?3jzhdkALxzaT| zNJR=p-sYC}{6sPrgPbmZTF1MUnlzSR2s{IvVyS7Ns#hs$x97@WPSnP5PbR#_#MbCMJY{O>)T8bMwwDZMo8^eBY~nvVB3eF9<|n7 zXvEGwRRbNmesxjw_~Hcu661v(MSSe9D<4TtNa3|F4{5rR?3%*b<27;=PmQ=Ix$!nD=WdjzOcmaQH$sr(7i zE`D2U%Zq34xCRx&>pueZOI^)m(cT~&fN;d;j0*ZD;@C#S5IxB6QT>Si>;)>=&hK2- z@WoV)t>|&f15LkoL*yM>!aAmnF^=xWXixh{aanC_=y1aaIIpyx%~gq!JluiAsvG&% z9-;96072FFi=846Pt4LrHZeK%73f157PUC#fuGrD$jupCoOGt*zKqsB0_gW`BUoKc zXxaOc4azwD>&^U4@QYE=BlAR%%Lisw=kl(&R)tT6oO7j8pTv(k#yikPaaN|Xk|$(k zJCxw!qG%md5~uO4Oj9DFoKS`H<_=}T#F_T*)-2fawzO!O$_R)j8bk^V(NJW)!G89PTo>sh*9qKz7d zAfH;*kZ<{iJC7&UksBD=m&896M{6F6z9iL`OGgg2e{6p&Fz7(fYW3@%j2=Dsm*Fi3 z!(KP>74EO${W4!SPqUEmyzm*w<*}YK(!5W^c4cHsl)x%^uh9Ph*vr6P0QlwbLs-yf zwu&DR+FZwU`je?Hw#&K4EjW1E{X6$381h7kH@ML8e=1wwCtkV+NmUNmW?#rOO^mF$9zE zS@w4S0J>tp0#8n~*4JlUCw`|q)*D-s^JmhSZ88ZN&(gYyEId$Kuta!%b zk~3D7nR4uBM5P8;_pJz=?$5tkYQPh!o<{?%NgNq&bA{=e(#f0>c9Cr^FfdLBw`!PQ z>~aqT4oy=$Bsuz&XKam1mU7XtYr2z*HvZ!6$B1>s}X8BkTy@LtN|(Ah0i>e6y%-9&5{5(BXx61ns+** zQKO}gQwrG(Ksg!BNSe6YjlkgJHP0p31$ke(delpFLXMuiS1hQ-^g5wgtqz>)3|J6J zUAR2be`<0t+#kxhq@0!`IVPn`l12*<0plXGR5^4;s_TFD)b3g9oV9M!n!LX$OZ z!!FUtepPU!9On#rd)B;H zLPb{HyKx|a)K?t@>x9b#(Brq#wzT~~DddMk$OE7?o1(9=vaI`^TFRgyC6$jj9+e!# zHze&K@srxJBD+~PD@e@WPKVNo@5xXUly7lIvNUy7iKZ=b+jnDV+A?@GR%rIOLWijI zB9yTl7svaEc*DqB=T#tMcTd8#jXh3xJMLtQZQ(vpGENS7G>c{#$r0rD=k=|~>_eO+ zd@e!br@dN^?Oi*Q>Q;<{83mpOqQn~#-A1RgLkTDFmx2JD@|zVG*jD@Hv$6Y}7y zwlF&rR^ro07%RUAoQ}UrUd%nBGgi!%1Z4(taoVKV6L1XOdY*Aw^IEgH3WbOrbC1)# zGFx4%A;4BoQ;L^OmZjLmWCgN6A9L?jqzjpLj(cwP9+jze5apL?Q`C;W^_Oy`k%kqA z0dQ%1F18msC|piX0ALvBe%{8YNiWJiTpz7kx^Pgcfq}~r$I`PUm@e{8=D+|Rf~ir{ zLS0g5qzJ?nQX8C+)7qpAVMmatZHt_KHBuXdBq2gL0E5!1c%JobbSsjik*z%@o$QS=CyesZbS9Y@~oEEjnZRwYa3A?eYo6DC$4Kg8>U7)RHb7* z^)=^U@)PUV-l?}F1KYJaIX5Rfd-~HR8QR3)&`pvW_~?3`)lz852iG+*U|eSz9VsPz z@)x}xg)Is!0%wjhOB{qaY+{_`g4n<$)F&Wy=9^=&vl^|`JvkkPW=j~0DnSReTGB4W zQiWr|HE&t4Xz?gtmAT@Ac6Ko~m0`|)wWDblnuH%ps8GlTMhK~GSjQ(lDFYjE6C#nG zJ5@z(r=@FN0VH_5@90F)J9RTFl2%yc3!EMgY*kX1y<-<=6HCRAUic%z zUM#oN!o`1YXC!!G!E0uWQFSYuxNQ8z z9J-vd1N6mwW1&J8*)n$LsPA1eQBU1BrF8|(xx1Xiyx^bKq_ZSku{aD*9-S!=10FWu zbPL{{5(tzr8_wr!eSV?mhHuqUYYtifv1IRS7EK7U_oe3Hi$d*{M|%KGEdp_cfV1IXWJgPbNl zcJ#$@>Xq3FCJwJ^(%(9xkg+K&2Oxem#@}jt%SQ>x2L-WOn&*h7yLPsN{T|o?7yblSz2is+~to`OrnbQwmSlfD5iiYqKX>=#kwg{ zNX<4#3iRfmX)?wfIqkxVmfWs>asc$rX7prgnMC}(c%wP1w@V91cDQePtnif`N8wF7 z?p4H%@ZgSXs?aXhNmLW_WMhitf!cVjEiU8Am~)RqRbpB_TH0~s<%q%SS(=^Kn2rpA zcJvs@HCES9GFY+51Eqe9{3`HA{30I-{4u3@ zTV*z@bu%dIB%d)S^A+&-!M}$y{8sp%;r{>(+YyVqm1)W9u&2!D@)i1fsp+V$*6@O` z#-ccebG1}peuBLGQ_{@qpzSNPd8Njayd39_$E{~cXhswO7@i0`*Li(pV|)M?wgKdt zz`L-(`?>Gan(?OEdLG;+w?`L$Z@HK@5S}pI>zvdr)R0Ed!=~R)O6>J3Sp#EW5%lM& zu5$V$0pnId+XJn0FMCy?=tie9If)@}J0k=(PB^3xef*XLFdJC)t5;EDXt;F z_|$GdRSXGnkC;{SChk*LS0OO;Qo)EnFBA>Qy;*QOlgB>QYBmL<8*oP%=}OTT&iMlc z&H+4TsY=_4uEt#PGcO2pgOk?0{{Z4DCBB&Ddf=RU*Qh1COBld7>OmL=xbKL%QruWD zC{Y+>W2e1$(6=&Z;ly8-S3Yagbh&SDTg*~&-*|Vf^3%dLLeZMyC3V}7aniT^JKXKqDo*;?fEF1^iOm1BxLDWp3IXAR;$IOI#)J8MV3g5iCSpgUXRS4rv{-pL4pODmy;-`U16p{E=s@g_+jqsaxLgy52BcY_Cnm|2IbN>MAr>Vin;-eWHcc=1k)97gklkc!O z`qato$p)ZM2VB(A?j!-h9rH@$HL0g*@{y6(sI6EqBFhZ=fmwE}9IqMeT8zRwa!9}f zjBrgc9LBRI`?5OveF10S#w;vO!2Y7 zHQHU=CA4wILB=`}#be!RDR2yt923qdEl65!RI}BZFn64a+qAq{QAy-`*D*bmw(LU# zO@ih~7;4cKgSJ?e*mb*VffRICv*X zmp-AgY9$pYv?p`*t$rwcKJj11Ek945!sAbaMY}qEq&HKKEismsLcNJN0=S=q+AaM4 zD%Z5jUlrZ>cShH5X11BE)CQ8=xxfVUBp$WrUMKMGnBF_md|3viX{krA!upSs1A^AO zRAjL2laX5AvxkXoJbm#UEv~OTn^tEw;dn0NILYiYT=f%Hmud1tHK>JwPnJ*B9)kzs;KH@cuW0Jp7lDRV~W zRvouKR#Ng6)Zutuerl<8Y6&4np}?)vYzbWagaC8anJfZ84#G(8D%0q6&$-LBvnkFo z&kK%~EKuV+j^6dBJ*zB-joBQL(xVp2jCUy@4x+b5n<;B_WengK+mXjTF;0;LL>!Ed zTD~9+k=G-wSCZKGBL^Q$(q?N#c4W^MbKDc0)k3>QGJmCOO&MTL7XzuPQpQ5yV}aE7 zq-_~AQDRd$2R*5!vJNqk)~tDMiZXG8dpGYMt2V9ySDn2otolyS%tDR5 z0|I$HswA(mqQ6nsW>Vzs7$cub%X!<)Jd9(~wdYnS<+3tygW9n!QDZ!v+4QejQdaXh z)mL&vyYk8mH$3j^&1u`;?Tlca4sd-d060K;t8f7|IF*P1bsXeVcy#&}tF@`q{{U(h zISjji1Ofeh>UGqM?;-bKec$U`s!qTXcrVcPsXXixlZOOg4@x72x^^4HX>+AbU+xT& zJMBE5YH7GE3EYDxqZ~Ka-mxRN^Kk4oG2ny!D)p!e-#!RD1sy=E`n8Dm_cR28cfx=h zpeMaa9BvsmIAC}v2WpBqRBXXG2OytX(TWk~gko{joK<~Z#$MY}&72I*oP&-)&JKO+ zW-Ck(q^WMa9D$mS!q4O^LX{m3ImL7qmf|Jaz?5tpw^~Y#_6J(aQ!#YWw35Fvw+C-u zV^$MSV81XwcR5~w`c}V^xY~Jb!)ZN*SQcEdP&aeQrlDjnX>$qfo>>8GDbH5)s<)Py zIbL@Ty+&((TTQCXj1b+fZOJ^jG^wljcFKwyIUPFZb>JUNiShFVN`EoKg zl1J-ZPN1JCzyldL70z6#Op-}+*X8M2MxSBIrH+2~JhAz9sNj-%`c_QnR|K4n0Rp$L zV#}4=xiQWcPf(pGuOO zwxu*X8$HE!+BNJtt(2Ctf2`YuK9$SM11Q`%6t;IyV>pp=4I%Day0m*MhmFA)Jx_Xt z(_sX12(9_mHOnzN@iL4P(y+)eGbk4iAPQPaDBJ>&jMC9XfGYNm3A|h! zRT{gb=j}rSBzB-rW22CP5a;h_pGwNSY-b&PI@XlTLJ`mk%GlR8aYm|1W&?^(xtZMf zKG;X5+oXs6V;5pOS922NmLce6J6_9VUosK z=W-e-<6Lw7=JuPM6W)tSZfDqXs9AsjepTZ(N&=`~nMN=VTIsf@Tx9uUb51S0 za?D8Uno!`NTyylRI%b(JrF8|Y3i8b>g;>>dl6wlZ;$IAEJ`>b_&wmG#CQcV2v-pnm z)3ExTm%^Wt>gklI{oXRekF9zzneOCefO4k=++g;tI>%AQF*t3y=O z$PEHE2SbYUUm0twZeG$BCT8K=upX5(sk)BF^x)?7F?>y}oB8~da?qzATEQJDNX-CZ zy{JV+Sst2Ps=@(^S`Pw%b62rJMHC8X0*WZ61C@6$9<&rVCm931HkDYeK7H zN-r;qzfLnv=x1ZxKVtoN`yck2@KbKuF65TY{*HtX*1t?I>;xWcL3vgxL-p%llAp8Y zpmhHLj-CR!EJ=@9x8!@cAJ)HM^{7C*yWFH7F^?zSy!!CFHH@r|8&b25P}^0bBZHnu zu6FGeUO+}d@r}KB>s^kexJE%@3GeCbE1J58C|M3h-gkAcHgL9$9k8oy&KpsMm6>+n zH}N-0<}acMv7ameTJ zHDVa|1wBaTrAHmVn9R}u9G)>(&=C*HxyJ>4Dy*ACH*G`gv z7YZ_IKG35%0xpmKv0H;4>aaI0mlE z;wPO7A>0m58`6{4R2FW>G4USdB%UNj-Hzj-u4cwQGq*YISA0!7++U6S06J85Y9j}s zIO&=T*=k99If(+PI3tl*(?0Xlk6N{PIdV98(bRXgrD!N?eRki(*kl z8OPVP0~|R#@lV_jBPSlT&;cFuP9LQN4^Rq&&U@7G=XUHIXQe@q7dgr4#aXo{jAS2r zEGBB(0@%+^xB{isE`nMS(Bp%iY4%a_3E)62OiYcg!M3+rEB|Rt7N_{awq_!j0F0*h~cHpBOdQsR-}!Q z(aRiHF%m#OTFkuAH-b=L_8F-yu8V&T-M+o+YSM8i`=Fk~(w)fjM-F0^)2y?+eibYH{BJ@bFQ_$OQE!!j<_DQ`-?oNbinP3 z!En|HuNWx#w#yd zw`syi;PJ@T=&?9)#+*&~`tc?lzpQ!<>GxmJcBenQ}Y}ZdDR%YSj zFvE~rEw>zUFPS79e_KHPwFCI@gN4L*xAyQ}HIVb*c$077HOJ6(WQHc7-1O z>z(jKMfA4~f;Wym!0TBkIIEjcij0jz5G)HEWOAdY(xQ7%7^3{Z^}wx9FKOg)*QY~J zuraeXbByC154BWP*|gn}i13oeGNXm+I28c50B#?5xfQ24+A+KD4oDxRM%KX`k;uk9 ztEZ)z<;mVxMp%hY0Asl)wO5|me9ycy?OKvA497c(03T|vMZOe(!8l5?f%e z13krb^XY9LBMiKN4QI<@q%H?j&S`0}tX{jAhBg@r4=1OkI9r^r9Czox(zIU8=Wgr_ z@wcE9!E8&FINh9&dc~u%oV2k`ZO$@AIp|L{Ab_aA!98(T5dxeA1bWh<6(A6H`jJ_* zbws2fEs~&)I(yZqpe__H2X1@Q;{mc2PbaArr)z2$5}cl#)V{-HmNuJX$?55e>GaJn zcRLi1)~jijIbEy>+%cSF`kpN2J812S0iibXie2C@0;UC-zY;DcTvh&SQzR=KQ zD9?YTdK^}fxIFFv;A7UR+-R``We9PeeJD=*gNft!5cz61A;vli#J+%qU{?fmuA=(b zyA=zN2{;~=&t1j=!j4Ed$?acV?DRY+&dkI)G61X_cNjgfP6)^XkU}0#4=0*g0cK^v zAaywPr^bplq9MYJeATRJ*O=$|2-MF^noL04?0PV==2;-0|8W8M9IUVvVUf8sQAO^wYjtw_O zuZW*%eGO|>#Gf%3BY}?9t!|*mjl7eMrzW#4NK&L7{CvTewDdX+ z9uldr;NX#;QT48Z)&+D>Lv3HW8nE=jsoAj0%8s?sTDB2OWPyRn=kcy-D?6=P0MalRAi?>7m#X_=*4vPIeYk#;a~v{*-#H3{;KEi zCGtX?a<~OR&o$IvL<=w{4aoqGb66KKc_m38t~vwx*71n%)aS1X%+AE+w{kiYSQjd# ziNhf|&U;r$bh|hO%X%M5!?=XZH*5eC_ksN@u9Dp{FR7kVFvZS$53jv*o-c(=C<2hg zpO~L+E30J7o(h67gIFFfw=wCfDFK!#75@MX@^f7H+I-X{ZidunYKb$zH3+<$rzG+i z=A8){$OPx5YIx61YkfvZ6;GDOa{VhF)a0BK^skLxXScaKw;qGJ^s18g9DPBlrf{I= zBi^dJZztdNqh<{ujCTQm0QRcV4cN~F)bG#?;Coag;jw~g+d-lR`@~Qkm%a^HHrx_= zR1Fyb`+Cy3cTBSd*=7pduhOhdY_j~r)6%mC%UhZgM{=eo#{6(i43@hUH2c@ExW;mX z^Y2<0YaPn@u;VBAYKEg_A#*C$!bg}Qrw6&>yAKL{I=t~VrE6hh47!Eo z$NH;#W!%NPXVSi%(m!OAYpPj2!^Vi9eXZ`7<$!$$Qhh6=`!5kQKqM@vhBb~`;dFO#YWJ_NN-NQGFZdG+RQ7O~lJLHLp;4$f%_dO@W z(0FT3c&EQ!zi}oR8pKdt=vFTi&iRIL9+hmMlg|ZAW zDR0L>3t3?yZ?_HmsU$*7n+I~M-+V5veT`m%&obtqzUSV$$2XX!{TJ1k+$VvU9 zwWuU48Vh((?>RZ!pX*$NDZPL$2Gd-xbJM4Y&N6*-SpE<3Ww(Z|`vFC8(f<1nfYvMnTY`j-3 zlv-uGL1hdmuErpqV-A=dKop*`u?qW_{u;@3<<#3&@cx!V{D z9F;ly)%bbg`-5w(>C=OThUzB=9*zxu-)k2FNn)1+62TOoQe#tG^<#T7Tehbys6JGJ zG8FTJ(z)GGEO|_N|glLi!QfxaSM#YZ*4qTKQPdo{X#b z8s;yfJDB7yN6ZQ28tSz-1h&vNo;M2SH4KCsPnbv{f%F~gj!{r|JqmJY>Pe3lw)aVIZDdp_Ua;rNY(MQ@JY!B&{MDG z`#rPoZz|h4rEHeT#wtS;k`aJ^D!*r;O*v2ywtM25LQKexHg1YdU$xTYl&{KpJuzL5 zhu|6J8{H1x_!ZE}q1xz{!RJtjI0W^pvWwJ?ZFD%h9}T|U(t;PiAJVV1R@!_$2I6=< z@lfjiD~2~IY^MW^;djkq5ZVLpFnLG-cs@W419>w#2iH z)j!g&gJ9qf&YLR~LC|{o;;i0bZ(DCfJ9AmLcL2L`eL7V2(6XGbxsN1(YARm2q@#00 z9cTi8F-khq1_Wa?&@eMk>PV)JDT?Ei@&KnvmB{NzDZr31*Xu|JA>0pR>sKuhHaW?v zaf6n~&1hIC-lNynnJn9lSz`_x01Rgtu47cXH&;yA$Q)+6JNtoaVowTjz|C?yixN*< z)T@mOgL?own(QrN7TRhIeg12ev5(B1N4<2{cLVn@j4sd)XzCJu3Kv$q?mQ7xO6?1b z4sp$I21%pH1+Y(1S@$pl{5a1_d!s$IBDav|2d`Suh02h69-{`cRT?~Gez~jA@^hLH zT-Snr04F#el>3WBxI#b-aniHmVls2|=C4C8(mHWeMVW(lro0Y%u6d{0%jL3;K^4_p zTBYnfD9>UmpS-a!z856-z^Soovp1l-^DR&lyN-jkMS48g{S?-2n?IC7oM3dVONJ6c z#1K1yO+MsFHhl;C7G2ME;y)31YAH4iV@aAdE5R(hfm8ngV*#ptx-hFGM`W4K@+KI3hko_XyVL&+GU{B$Azxpk>rxz$Rmbq;fo*2v3?&R8Xopt>%WnXhOK--jB1u! zw=b3xDvB~e^ri3}x0B&pa`{$fW+S-gp{%5=j;JThY)B5x1%f^a(asf0xd!xGu-I3lEl74X1r9dJD;OdXbne3tvj3ZNgoxc;?T z;>iv&gFP!Px#5=m~E z9Ar~ON~+oAy5t;IO&#!BmEnV9U=xAehu*D?IusZIc9~9Q*42b6Bn^{* z27mh1o7~Z)wntN;TYy6aJBI`??d@KJr0Ma=9A~b3j{! zvnq|MMh1OBsB$xsT}=pWnFa{oi~zW;1&1uVkIc+>5zykF2rz=ookjpBky5HSA1>4_ z(6aRPtW#~=$?3R}!mY87%B18kLQP2a?JB+(1v(DocB=|LMr8^HcHy12|F0stKkT-GJi1KU!ccb z51Cd=Zy95T60IC3v($Hs}&jJu6tIMt8M|f z9FdA+TA0@?u2>xOsn#PC4Z!@z98}JtXcE^$O4>kGK?H(+nXRii{J@~+YWjBetQd!u zNFePEmC5PFXj)tZ1$N-$Ao`l-=SG%&PQOdI0!0`B`;Thgx4cv?qn0cKAnja5z3$`; z+_yvT_|}vft1l$=<2m)LB#T~WVFlti04U({K=1EUo1-%2akw0FT%3A|+wzsl_dV*$ z>Ws%Nf)5#9#-Y11(z(z+pfShZUZn6om7jHXv9l6(x%=BqX346vI~F@a`Y-dWtK0qT zGFxw64O~{YY=)B1n(lceV&f`=5L*?`PjDo8BXE7V@9k1vTsds3I)FzFRb;u!NCM%B z%N6cDt9Ue)qLs8L-ENSa1`-9p2C}YEi3;2hdyYT-b*V1WTW;jXBb?+_3DawZ#~ctr z%GIi=YEy${Bu|v?Dn<`Zrn3A~7@tACKa)&*urqg~REyP|%t!|fUzzxK50I!K%k8al3=3X_G4goAV=CfRU2 zWG!ib_EuxeGIRs$S=zsWF5*)qx~-D$x0n%)|6Yz*j ze*uou?Gd%b;-*;0{OmtLUzk4!d@~=6ek5pGRKR&t-XWB?-3S;K&$cV~cf&si^q+=4 z9@O;PDOPM6&h(uw{{Xs{LU$_3j!kmEv`4~E5?$!SS@4U=afacc9)Y+1$J$RgQGn7DeHPI^>Gzrxyg4=Nz0|wmoA|@uU~G zJ3P!GVTFirM|!Dyt3(x&7cm3d8O{$wTvnTHI!?;5^PCN%kEpH5&{)Q@O8Xl-h>+s| zQ6|xGwuk12?E!w;r|li%Y2=g5D{vyn?S%ssgJ_}f4?r{c()>Zb8mGn|8)^!{)=f%K z8G2wG3{qR30}c*3#dD=|)RxH)m}G_E^VhXT5AhMtKssaIqz(_u>Uk!S+l-zDdeONr z-C7;r!_gdRbCNd|9J6B;wX14{B4tB?{7es8=6oJp#WKVhn1O}Ha&ywJYbq@yf3*;s zMs}L=F-`NFRTsT;CjRH_+wSgP%wW7A^zEAR4;^@c^&6ikIU_k>PI2vBZ;N&IOKC1- zAR_#TqxioH@|mSG%JMhfW0O~hrFYFEvV;>{%w-f(6N>as&;t~-Q9=ODXrK&HOah81 zpasn(6a$(!5V#$v+qlwIi%WL7eXG~4HLC{fM5sx}deqQ7E2?;(RJYS~n=6e*{&vK6 z_R*JiKEP3JVBc#mSN>HJQ1V2og(Z0Mq4WWoRIOi45-&_y@X5Kl;Bi69xzNIG2 zo9Tkc#n1vc2R$p8)TRK)5V~&1&DOTHIaxStD97GB^{#ID_DJBI^ec{l*Cj=%+fG(7 zE?hG)1Qy?%5ztjq;1k0UkXw!vdsU0~Kf*V2oQ^V0Qz{}%zDQA$19U&uvGqG*?zAU_ z9__ar24j!N`Y`&|Y)_q|j5j-g!Tf66G8OB+0UQOv2b#4Np8r}- zGnS0`R{6ZyODh)l^{TPypJ{M=lTqAU6beVuuIU=QFtGCmMtU0dk>tCxb3yPxymdQ% zP?7<6-fVS|AZ^*g9jaX6n8>nZxaTERV zRh?ZHS&F>x&<2eTG3`~MH@?KuHKm&7S>st0=V&!YRPjU*TZpXkWnuFYI0xFXb*%`j z0(dtS1CUNgtXq75K_pa^yO!QuM#LPG$v*X&D;9E1Se{I9K^dxj%>|?elYvV`I~E3t zD4+$!1EHXr2NW`FIR<$qo}g}~0#~+0Dg&J3npP7gjaauIm#u5rE(aqZdscO_Vn9bV zw{8WzZ2f@iLUtj$nY#VJbx4BazH1t+UtBfC%dt5FfmP|6tr4Z5#9r}EGDkdB{Xw_g zGh=BTy($d|2;-Jae86$_spJ__#N?j7l+wAIdXBZ9%NRX6in0|ZQ;g#ko01fBx7^k1 zSpmY3c<)BU?m;9R~TKuY_)xbR-7 zb>e@7UJ2AF8 zqp39{($JjTXFIXaTC)tlI2a&w&14&Z8MC#N`iiv#n?^I&(zNtBS~jic45fN|`qkB# zv5mJdJ+Nyc8OO@JH#~~CW!|~K2N=SdT}azf@Y)$zmuBawz|Xx%ueHDjX&{V_n5uD2 zj!KZKIS0K?vXyU`h9Kkv!KOCRX}IADWdLVBwLU`f6r8R(>MEK@8FB_n>R!0#G zpMH9Q=}yGj`w4UcsUY!;b6N6`k(?&(GnG9vR_2fh+z@feJHMr8$syRdDo;$F^^~r5 zQGW3R&A8woQ_xfjzr6XBw>T9frCGs0HaN)RH4C;|X~$fUJBsIREe}>Bk&Bi9oP&&3 zmYBpo*B}AV6V|eUS8NV2GgfUb3`_~Z!Sv4+oR*pz-p8Y8a+N@2Wh@Bk+wu7JZn%A_sP(Ue$+;Yc^)^7Sc9NHHl#!ln)_M}xP^zv-k6hVWiO25psMCui21_iatT023FMyCf-^jYn4Bof5rdOiziD{~ z4hScwa%)#kjE$~Wp~>2J0Bi15T@MO0w7GiD&nKW^fx)edeK-KFLgWsV+c-nGBb}p` zZZlg}&|yI+3jY9m^{gjdwWB&6EsBFlNj8v#D)ac!CW;12@q@yipr1hmLz1H4mj^o!d0;0y;SqXOhy?`7*)ncN|{!%s;1clilpQYIO=oF za=txL{{RS{q{MF96CMUDt(|u&s`5xbFzH<9jiK{ADXFN*Re_m-?~h8(rEX|i=8&Y6 zu6*luyWkv?kO%~ORvd_8lOxu;OVQ-pBV^!?GxV-z;!W&*b6!4&V%!qxk%0_&t&Km# zcUJDDW^9g!HGl?ADtmazI5`8>f!Q6bJ}$m>3iBBg2j!4%$LBuS!5nPqs&E}l$ zjNd1y@Aajy(|0h4XwN(mPLf|RVTU8VX@64*?eaz}2|T^epxqKYXL zCfZOjMHFlSMHF|YfGDDh04So0U=&eBAPOj=ivXZzwJn{av?OB{mOAszYQUvJ1;#Kq zp`$q-yZbxawYIH$ss$c$#pJ^w>WofF{Dpqk_;bdqQxvb|#xVE>JPcBK}G9zd274)w;*0eZvuNB35wul)+258i8E$dw; z#tEjg)Vw`&0y88`@s>GLGwR7GZQ#>*=95c+#lkQD)TLxE*upD|1t#wCZ(S1ic7EIt_xaOE} zI4T%)`c#r~oPIc}<#D@kBp#Tm+(&=l4N}iV({*dRwO5sVv2F?Hc5-oC?}>GpHG9df z=6#FH70&=wT|I}8LB`SP{cDQx_lXRSY#?9<{Io0ltIeDo>e$Uw=a-Sr_>)>Xwc!YV zaGWUh_pE&8iU7rTMJTHyy(E>8QJhgh8K7ZB6jQhb6j9AE3MitB08yOMiZB2e>q!Vf zz^htx=H6}Nt}#_>LqpS~)b$u`p#i0pf*>aYB;uXzsHLj0dGDmTx>sg7$m`a=i}1`I z7}tC+;Q9P`y7b-~(2R%5)29dbOxv9lH$um_uN}GYL_Qz!Z}v8$t!eSk8z_0L)xWJMC}#3Fo$+_pS1!hwQw&JA-q zh+()PKKbcgb?{NUcF?K@?DhO>mDEG|jDTTONBJFBwNHYP()D^*Q!6UCH2g&lJ}C*z4=ssbaT6!^}HH9z#QPy=LOVQ0d4WYku2aw0Q|@cHGC3 z#-OvZvuOBVoOS1|NhP(vn~2o=mZxsPbGp*3EOm`NnL$IehEtF8n)6Q*X(vX#bsVWb zDX#wj!o{tuZEAH3Lvf7l=tVcFo~hjPnOw-^ns0CGLv{RVDIEbt6i`rO z6j4SvrXfJ*tr+K}B`S^FP%DqlK=!AqN1&%}whFJi9D)-Ofcd zlRXCNo(BYo$=IWh{=F-q$pGMXKD3dOE4*jQ0OOP1k|y%l@_h$1_x|pBcc{)=QG+ z5lH|l#OL&+hv+><2TW$BwV%p3-Lz*k#zDhmjL^{ojxt9)(Yh7vYl#H0{GXwzK~Z6|iLxK=n{Eygl^fu|PJxThP&$oF>E zuI0Sa-uX-@ig6=H%laRsb=FX&*MS!n&uz?-^UzYe=wu{{YHIkmNG-EzNsaiYj!~SbWjt zQmH7**~@;1&6<_j66ZUC80U=EMAsn*1oi3dU023kD0pMyMz^M+EfjNku$%#y+m$%3 z1+JGPIVaM*jgHl-v3agla=frT#Z3jk;Ee7)vsle?OMIkt;8c-b6lOb*Urv;hDW; zfES?0>r;KG7|SpLalt1C=}zq+QP8(>KJmHOa5^5fkuA{PSAm0^0Z`mte7%Eo=eOfo z@?H}rLkylXP}#K=XQ`lWatH%z=f8SB(!8<82><{&&1bB;m>lHw;+)NZMt=7o=cQox zJxJWpl0d3HeWTQWDpj1Taq{P|HAu>13xEzV2fg0elWV@aLMAx}qf2k7?5NTn(GDq0tDFs|uYTPtlzniq7<`{x;8tbdh&HKbV1)G>jYEylT_b|CxZKH*pm$vG zE2GmGDx3vToD^O&?kkP5c-2cYe3{1`PHU{v&y*17rx|Yl09yBFR@)gyn%L>I@09}J zE=NPwx9k{Tf;JwCxUO#1RoFH}nCjW>>04H^6jf*8LCEJP9jl$yZ&ODEbarmU{O#$6 z#wyFi4qb7z^SEdAs#a1bm6c;5UP$d%7BMyha7fP4&{s67jYO$+2?0(w6a?TLgIRZx zmgtNNV~=X+fE!d3^YS(kS@)2Uy%n>Q@~{5@Ua|J+F{)R39G0e`iFCsJ+j2RqONVR~ z93Na)O??WWVYm~LdSqs>?jl|pNiKLiaC7ff?cZ>8wbap#)p zG@F*&w{}?c1azm)9J?J9E)=PhfWD))YW=zL_OIVPM(;|?w|4Uovj!N+z^zCn*mo1p zraRVdc6*v7+|ae!2pw6vr?;(4aKxMf4^i9NtH2sUNx@8Kj@4yI2wVPH4^FxIRxY#Z zX&8-ds<{ib`kZG2+PVJ#6e7vty-zKYU}evtKHinmyoF&TDI9mMYsE!j@W!bqQ0^fo z(`d-%quRDxg^!(lRXW2QQ7AkPeznSz-n6y*(|35x4nLQ3U;?hQ$7HAsi<)=~#J?MYl|Eyy)H2ePhkbIG`9S?kd6rP%#^<9zf{{R;6KeYU3E%GpETm(W7L>K^m z7_XrJ0N|hABacb=uc+C{KEbHzj~NZ~1}btZh$4~lkgq|seLB@IANY@0_>tlZonKqk)xnVcn@@#y#R7H%)cTs^ zTFTxcPuiuGNI3zx8T77t)!h9GOO&%LAj;l#*ld6Z4NRS7T$7Lc{YOYiNhmSOp)}Gk zI;5pLr5j{41C&ry7=zK^=o~4{2&EZ~bc>AcMg;x*_x(NkKit!8ulv5PbDi@(U4Kk{ z+9CG#g=C95PkodrWo~{ zCae0d%;CKs5^GjmTnAR(spG^fp(hrWY{z0Ffe&v|9f=xM(qDgx+4Wq}I8NV;D-`Ev zKtx69i+#c4ikWAr;CX}aYrkH{5;c^~DE{&_o6E3y;54$7_*plhXq@`R7mnN4t$)li zb@=r7Fq;j+L(iMS)q|$Tm`+ymXklD2|HpZtS+7noIF=zROq=25g*&Yp-6}kdz$%mh3SXSToLt) zoXyZY%Q~PRbGNULRmcD^@5@D1chK*r*{=R8an=0J?zT5#_o`yS!3QFKCk!IM9f1Q2 zM-aFO?4Tq5QCvbyQEJyimBUxsbuJ4B9x-5Sn4A=taQd5iS$mY2_>CkAVJr5HmgVOm z&zrs1C2dueJ2JF6#~Swgp6iJV+Ak&Zn2%~1_|!}HZk`afoF=!Moo#--%V{GqAj z{@Bz^Z~LbYc18mC7P7dz`N|R@7eprWtntFeWZ3x=w}^l1jd;8m`Qa^pJ>CEv_`o*z$}z9 zCKPIH@}$%!`b2&)DZl4p1Ro~!dLZHNEQ`v=Re&u@-9{LC0qsodPQ5hQz|x!-eHmKD zRfh5>+31l)VVZS3`0Vj(dRcC9Z-JUZnGzMu^3c6!_nRMBd**0HU8YV;|7~>&4JmCy z)X1X==-VkZsq*gLFsL!2)(DAH$~qS_l!l9f<)B8Ic!^?1%f`90PnZ(`GRMstm5+G* zyEX&Il%H&Kroi4d`Oxvi#I&ROIcv04&bDH3UAFpT!ma?ejHTp_AaXM)uJX%rx;i=e zMt1RP7he$y@4f44nS8DqnK_E&40hZDySQy;k%icuOo83Q#*sw>434q<~n6fa36Y(2Xq)D?=p8xik>HJ~vMeoG$V6}10 zm%;0&6SgV9SRk%Cz=N{<@PA#p1#|+yI37Gqz7o=?4(X5tb4X<^jgLgl)mY~HV+MVA z>#wS17zXZX<4Bw@`N(y~g}^mF8%AQIYo*PBp|4DrmB3uzOYA9A3C;8?|9Ei<`>Z>; zpwuM4@%b(Nbh3h^AB>q|)BfP+@n58@)zqGV)*t1DR=|WK^RaU8O3b6}TY%Z7b?MlW zcrudBzr4?1 zPc*;}pqGsKGDHK4*GC3c7^Oi%_~vY6LA}P#QlfGI$tHQXu3q)48qQY%SYTkkB2v&l z>c9~9AALyq7Zn;GAkTA;ijKOhW_%mw2ek{VI_-|cmO*ozT@v3wW}wp)&BAYtXoL)* zf#M1FzFt>&><)~&Ay>e-a^QoSPF(pZO;PH1Y zc`DC$r^!oei|lt=z3lD}17E=v$S&Q@(h+Y|CUWRqpP_I4@9f;Diaq`@&g9a^hI%s% z?7F_#RtpihqQ5AzA-`N2_^$SSy_gm6({K6Xe{)1nBwG0V;Gc6Ue{}x`V7i}9_8v>t z1+3>bO&AJNcB@(6yzyIXWr4(&iT@lH(YSir@M>W|kLW`l%!#SHoyBx!u4P6dZT)D1 zsL}qL8HS56c*1fk-JxR>b#Um>W3wJRAS6BgkdNg(6GB9YNrKy|Cv&a@}EUFn96fiYn_liqe1qUxNNRh?rZ|QHN*G>Lm;&`rV#2(zQ7u3;*`Rx zCjae3zWPALluJtp5S34Wr#QkZ6nf zZ>=WsMi>U!ow$+qc_tfFN3zTxQ?4aWFW7yB@La_h4`#=0hoS6n^!5Z@R=V;=Rs!h% zU8<%*r+r|N`Tfd#BS^b4Hl2xHtYrOl0p)@VNA-wA` z#RYIvz_9bh(`K@Pk0Adrs7YDgh~codV@htC@~_9Qq;Rj9tUcSX@}kN84+7b0swk^Rh~otR4#yy`Seq-m+>C$Gx+NGGr_nx6ElLv_mnLcE}k*Vk)_|Lk8ZBR z9|SV-zfT*X_V1QY!z`m{!h6MM-F`fFHbPQOwb2N>9W>q=w;Gz)I>N6y-4x@u34(Oj zbwmD`vVI|P1>nW)3Zy-X+!}$=$W<~{>(JKuJ=zjfT)&>s3NbX#>7-nbDA!0G8i{oV zM%`CzvWdikj-&m5?pfEjNxo!oWWH}T^s5?_3p1lSlU`r^Eox!cZ8l6sJcFhGxl-Tp zFnU;z1Co7Sn{cCzyA|wB|6)f$_=_onT_fM+xN6FKUh?Xg-yU<1^>xG@;>amq>u#qa zBLh~XXV&OW6PK)xXvtreDOgUPz4_#Qxg{|-E)v~IUs@*GJ~=Wcu5rlWY%4s-a`3sW zBMc?^Sn5c(^SkGwr0_j&8_K1tXe|4g({q01~mocqHjbo zou7!ZeJHzHffrGF(Oj**#WYqWVl}FV3Eq$Oj_~-+5LI1p2KYPap*b<}78N-(dCp{@ z<+6s6Zp*^^tjzGST{#}a@{`1!r&7x(F;Yo?7>gT7T<%bLZT}@PJ1d1!Tk(n|cVR^r zohz=mj?2)d($c6W<4P?+Zk87$}xN~F)Q$kH>TWJ!DD4^0?xX1@(9n% zVCBC#mlI?mRK4j_9(7i>s_wvyxms!$k?#T%){&6c;D23t#2deB&kj@@I}(bg17^J~ zcOPJupYW0%rT4#mAFg{U{{`#0XWi~T@2i6@`5480ggCNN7Mxc{e9qY#4!^)A##++D z48*_fn6SM75D}wmnEyTE^9)%vb!3WIk0ajZ>pUBLf5k_k8z#{n48z;nJXJig7WkJI z;iBO#$@UBMXP$iJOS_tofwsy+{ppR^SHyBs9F-z`2I#MnLDkGT@;;Jv0xNh3{rD#i z>g-OeR0p;J)85(iG!tpw{ay5!aMM(k^@C=%0aH=YK`s1s>De!kbf6QIEK2EeP#iiSFwK7NPI9Mk4VctY!A68;_|QBgS)npQf82`u zQ@E}ANF=3&Sc-o0KR`WYWVqZZm2P@~afU^E_Ek6Vkl2IvH`x{G&txJ7C@X-?@?wW? zsNg)1-L8y;n3wcQ>ZamY59IM!F(?^KSL;i(^$@<2J5ZQz=)~6oi&zgSnfxK3)gpYd ze8HA6SH7YQWNrU*d}689z6q!?CQrfLgX7E>iW-8B%cROPgr^^R{1~@gmM|}uY3RYi zOvBSbj-98BZg*qbAy<75}J(;9_z>Xiq{l=ew0B!Il3)s6W{eVM^x^- zh?9N^wXNyloCIV#_y=LXtfgN=Fa|5_FS^u=u^qg6WI-k zm1sU0%+3`dUD1=DWwtku7QHGlA%o7>ByxC3t0W2knJTm+=$0yV-3GvBt0T!j3=u>b z`E-d1uHa!%sbLPa)fTTj&b0nsROQS)~5BsJ{Yyw5!BvPRAuIV z+P~}!Vr%%iM0ZI5oc$3bsC=liu!Uy zzbY0(kFj=)j_@4)0KG}%^j7J;5tWeSA!eQ3e3~L{cZu5@Q{;HmE!Cp5ms0WzBgy<= zM()^hqFC)`O6b(+wP_cB1FIAUzt z7aU39Lb(PT=Mw+Y@@srR&e(@+QU<``-WxWa)&(tM0p=;*sZSLGHG&FgrBC7c`%c$x$;YczRrIYu6K@0$6Jq-v>sP-IaF#AF7Jg7Z4!fBj?u2Xtv`owEN_$#^dcXpOty_@Vnx`Za}bH zXC*GDnR;D(Guc|nI(8=%QhhInU3MtZ8kWde>#V%;zt%c)D%+4Ra^-V*c;HJE!u%0> zZ}ax_}}O6s63&u83AMyt}wO5=%+!r&IA*7MA}iNt`h*A&JL{2MKz9N^~ z^XPQ|X}6C86CuJC!lAJq%!oqV3N(tIMd?$Tv!Jp0l_Wtql`**<7R83I>M*xGDN#56 zrgX_MV8;9XRs&c}%!w&sC<_EJSO0?rV#<{;tGJxDz9YLivrA{k-OIXA?qISYomZkm zvKNahK3{5Rm!>}o(B8uBbklHmHEPIw4G*n?5Zmi&o}aDsMoHl9CXvkIoaMOtvX~e{ zwQg$`Ts$rY5$CS_r3FYZd1v$^(=lgZbEm`&u|2ak>M6}I;wQR7R>^zc%N#kORmZyt zvwP73ys5b1Ul=(0`BAIVeBdwnm(Y0W59={E2?5y=rwd9l8atpSh{X8vN3MN7Q^s?p zaUDOdm8-;;(}6k_SrE+8{DF;N1~jUh{0Rs-pT74e|C94q65t~%Q_F4OR>Qc4T)%d^ zC0+H*wSbI~sWG5t5^r2Od#$Ns1rD(JzIM!q@d{TACUW6N50@2GkGy^)Bb|1u8_LyM zD_s%XCP&7@3R{YTeTa2tM~dWXvNq1Qn`Kg?k}{>Z%6`v3dt7fZv_jN_mB^vntUOa(Z*qC9S@mD_t8zB;l>Q zme<$@U4C9d4Wd#a9xLB>uu{p%9b|Si1q)KgG-4I7Hi3P%4oSx4j);+VLrhX?c=pC4o}Y~pC=``PY{ zc6cCm+Pyv7zXGOZVRLI+B?p;ZTShtC(BF|@(qy#}>xoNdUmS4lK zFL&N*mL+H#agbimqUiz^Qh@N#fO#N&UReQu-1m^nnz(_5s^TP4`@9mn zZubImiGASuri8y?;R8*2%3DeR=ek3UCaK`CvG*QgDvMB?)l#WhzU0%mu=Pb8RJ2fV z7~904n^OF(be%fysZ-#al8|8dwr4wINm3}TJ9hD6;>^(_)&3uV`d(woniCbab0%37%j;0vA#?T78Olr!*=kPzhp*rMDUi{*cNI%h`xHmTm2|SRw7hz)i(sc9y^6I zH}46qlG;hhr&75%?tgf@5OfV-t2V}z1PN+7ChKaJK8mBz-{&soB8rMvr88(HA>wf> zO!!}jlpjFEEtc|VPHuoniF55l)1J>CqfUK?kQ(e%y}5J^qw5xx%2ZkSE}%D$=+P{bU(+C!=jI) z26Gn7kk@ycWH)+DzvHH?3`8*rUy~~|7;8Qa`Q<(^-2(N=3wliUS6SUA=84NY>-k*m zNHUEYZG{k04z^5H-M>2HXPWzF;NWaNPqOah7usF(^uWNp-uM)+d+>S3-b{O4{p63< zyl|uthzp)oE_?d=Ze<;4%7J$1|J-dGa?CrU>&zZqq0adiYDQi>8p*;43T~zmt!h?-ulY(#aA-kt`5wf>qVC;Ui}_`8lQiY z&ZR0b1@k3~2RNbaIecCi~gw-cr9I7K>_BtITVHb7Kc^_dI_ zcac;+xxUbWHe6Iuswh`#jF~iQJMPtmDu51+#Vtt$xiV8UYz$!5D^zLMylAn(1>e^M z$Vy7$frSTPz;jN)^#$=DRw3S&VS3AO4%eh&_4%K50^)w~mhAJ8s^kKVU$VyE$EVGi zvOMdx6uZMwk9&1viR8e`DT6aJEy1ew0K*(LN~zNQ`7wdZF+p>&ECa+1F3v7Dnzb&W z_vZCL$1_3K;F%L7IN;fZ1(L=jay{|27U3--sE*~+AQ9>bO~J}7Se7hYr&*Cw3Mg3} zSQd|kb&R`4Umx0gp2LN2tIXB@-g$MqgL39NR-SUbn zdHTFC1gWF{bWF6zPmG_j6I?VqI=}!OD}9awMHmJNmwqJNs@KG&23AaU`oDse&g~o8 z*IERBGcY%dFwp zzz5%=J1M%7^<^H6D~{_p?wdZcT^vp(`n!?l43Y_{dAXGAb_6jh8xD64bHDcd3*6_?u*~JFa$;5EZ59eS z(89wE+BTZIApK+DkX@es*mFp-C)lJR>S|1nuGLEzkbC*#G5@bl(Gx==PICW+7z6$s zVid~g<|BTlo7-N>rxHS#V-u7$v6@>|S;W{0&(@hNHBhELBEf6XMpm6m>PAdJ@~PGx z%hg&PXZ^7WfYMjaQ{D$z(-TP^O*lb zFoD|}C2-w?WIU0Dxsk}fJTcj6#9GM~(C&xS!j~xV&vD9iJ1KKz({$7XW6)~Zn?~k8SE@MLIv}-C zM7HB24KArePH+h!$Cv}neP+hs%n{^f zo>ouhu@qaiw+&j~PY5zs$0_r9qM9`p!C9X&+-f=wuKhu^$h9HVc}lz>dxPi_)-a11 zeEMBlLS6w&ANEBka=2s(XD{BwLddc zl;@hR6>w4z*sA0c-H8S>+>Ali>qMik>Gd+YHcG}(+6A}gUHBm!ofKe=0biVFI3vA& z$$Ek?hU#G^=7yNA&uA|uakx{hjmHYFigFzWgul8TFc+6I=>r`Cm;;3|>IxxIzJw~Nh}qhP_< zsrR~NNH*J<0LuJ0?Ob2!>=qLK6}+ga0`uW3(7d>qlM_`3g`V=%yz*2|Je>5hk~#%} z1o|g?G`b3Obsy^JB7Q}NCrcwN#QmGOzvieyknmmWuFUk@)4n{RuS%FfEqW3Y(6(INxLk$t8WHwJ&lmdFJxlYo zLBiwhbtl*Fbd^l!IdZ~=B8Qxa7M8F{#}Ox*2uk68&nJ@?`~7R1|KDVQ!EkDj#j zP;DjegQrco&x<4psWHP?X9xGzPrPd00nOA)M;f_cex)}{D` z(Bpc_d|FeJJ_D1Qo2&~*W}9Cr7oC@I7XCfMWvlm!#(=@^US{Pvvsu54r_9uH^E?#Q z*`L4DgRV7ys1Hf>1-Qy)W14us;4w}xzCN(LhHKq(y8 zJ==tN!9uF^H*tC6PbPx~vBy4b_L5cg0X?#{x{ZUjT~*M!Q!z?`Z3ikkzwkM- zXBX_v?7_0d_ilkN^;`??6$k`(T|Rfq(O0IrW$CFTTBZ>W%bO}hl?69E;c1fmxTPFa zu8^P{fTrdF0TM08pR_!}`^+rAopLnIhL!KgHk%Y!nUAA~pE-Gs? zF#Y1nLrKtFlh~1t>J;rb75h!$4Sl2ZhzzqQL-0_Ob!5;+`MzZy@|C>hEmy^Lzn-W{ z1iN#N$udm1pR`Pc=5gB1AAsN6RK7>=K)DgmhmVQLUKs@%ETK^aZca`FP46t;ei$w@5rjw#OCf*{H9*N z9;wqL`|!`2wTym1g<0?E4zqFEICwC>+Q|kXMBZ+mFbIsh0B=1H<<&XXE4Rg5z*<+R zOym-;N*LcV*zKUp4W1^<$S;zB$<+L>P@Jh9?>X(CRMHx*B21s_gSyxsO>i2jOND=$ zA+NQa@gg6lWO#Hi&cpgHlRJ?kKc(oxX{GVz3gj@ly9O z13mFYt;hG)bLGybK`1x}N8 ziF5J?k>HN{vHpB{GrHz)A=m5cmS-a`FE~jA4>rG_#5y`Rviwy zp#|>3jQnDue!^33BADHs!B@(}XVGU9&^v-_rqck#zU4_=T*@_HV149H6eqIfpETd5 z6+^aXDnS%tobU4Ewf||}J;!UJ%aCpT3~2nf0zV=p=2oR!s<~F-G%$~XXbZ9S((I+& z<3NhmIs#uTZ0=PgLDKN*XO?eGJ{yXC(PY$Nxf&I=9(|e^&fJh%{M3~|*IH8q<)rx| z7#s_mo_(@4Aroe(vHr`i0(X+jX0&4HBFG;U-&Oi)!Rl4yB3xtPjMu79LI}B?=0tl9 zD6tcCv+1gGRzX8UmW2G?-gFxJfxJKf5f$2{Hul;=k+*w=f|3IkR>YH6H@FC&8M!0w z<(A9)-iTj!Y+xy-1~+ zG0XQC9)&*m*nGe#6?8k*Or?~jIo~FY+AawX@WS!LVt(`jgrtWIJ?t9b?M+MZbujr% zXg909m}Qqz;6<$@J5R-@r!$Vx;S4`IvwV(8C>9>buDaUVWj`%m|GKVXP@I&K;sMn zL~hqkv&>71DdC&m(i2 zl7=pPbVln!$R_-%w$RgM&V+C?pkH9%QmjT+#*fm#{Dnt>d`6)644VPM=5OAoqU<&C zThX6#drmX!sm;3vk+Q$s;?;a^q#jxw#DmohCWxnM4$^sQEyeZ$txDtX;@aFBS7i+C zXsNa5DNF^zInRKDy!B9?(q{Uj>O|U*B}uzpLSJ|ESim>wAK&Z69q6{4n7_rM^My?u zOJwr@FW#-;5Qzhk-f}>4Z!0Q7_jTO&KH(NMuBPWIFXWfQo_i%Nhlo$mXmVbO*Vbrm z6)Jb$dJMzm7Nu}dnq%<-hj|Y>NpbrOs$9r?Tye!PmL<%k21HXNXPThl{p2V!b;S%W zY_O`PND-4V`Nw!}Tw?QnwYAiShoI(HqHXF6k)VbADPZ7caJ7sY=c%9YRQ&`o%iV0d zJIs+N^1Hw8O9$c`kLpqT3>dz;fvJ58lr`8XxY*d#Hm8;DIafitF>3v6&v2!wq(6-CQsejO%Cx`Zn=-(%M!bev$^b z^zV-p&bbeVIf^Gs4Jf8QzKY*(s+&J=tV$ekD8DMp&GAe_9X%6T z0TIA}+mXp_3ija~Wn&v`AJSRbKhRhi_DDFdqd>79*uWez3-7YlCcllRR({BN(L08F z?bmeS0_~@re}5!siTTKEQYGC5R=x$EapDMP6a!IJn%jLfj2`V6DPu*vQqB5a;4T4BcN?Yeh^WuEZ~Ij@;LGhX<>cdfE-u zOjBgra^IVc*0C#a0jP{DI}Kf@f9>-&PP6z(0v|QPvnuQG0Y)q9&TtMWNe3~Ce zd)^e{M;(7iw-n~zkyR|`5<~LCfYN@dRR$n(yOacv0N6q>m;1h@WOBedTG~H4Y#N~z z=|7Ikd&~CtuV7(QUjy~&fVC{9V20_>c2DxVKkbPpsTzfF)B%}-{K=LDOjP}xnT9N`DYUgV__WX|BJ&Q4451W zE_Yn<#bW@m{q@LwCIUp@RgRIBmS@yL?ne-B1tF%V5eHQl|&MF zmaGo-936_Cw25(o_bLn{8EQy}tOJ1m0zMfxZ%7EP*idLBAvY3I~!pDTSmW`@@2W=}H4g35`Sz*EgTs|yu<%lj@VH4Z^|^#SENagp zBdPKM7ctx%5@UIABLztEhn;(rjB=|W;Y4~sO3&u>Hr5DF;O6xn$cF>I?#EnSwH!|e z=OPPcJBEAuZLh9=UXeP~aT(QkE^f0LMF4PJ>#Wc`%(}=-@HuqSEUU!VU?4!|e%C@> zzvdT(kz6#oG_8T1_{(D!{r}9+ST&FfaYK@cJuZKN22jkYhc{0vP+M(|aZRD9=UHCP zuDy8YcX=MQvbMUMfZQE;TF>t?Ye||}E91$YQGt7~&=w*qYXwQWRPfOtdqyHOapt;k!xXV7e=I|wKhPZy z3SVxmXc`ft`Rnqkj(9FVNo|1_f202RY4T#%(1M$GZ#8P9EeK*R1q2aFW?yD*|Wy_ z#dgLqS{#FMm5=WH51>BT-_#LQcYHn%3>Y)|t@O*&hD!ZT^^i(nb%4!xH|DGOE?=ZAc$?6Q9#~u0=cVbUKD@rW@S>SMzoY>Spjvv(`@tX6SvX~4bFP&B3A z5FM$GVW^2Ybk3yh8t@I9dyDcie7f4MKN}a1PQ(!-#mVDvBk}{5-C`~}K%;>ws~t{} z2GBL8(jI9%08UCzv8KMD1LU7+bh*YEf{P0x$)~clcorTMJ!_jHpPqDr-INq?y*qu~ z{Q~=r)$NalQS))2sa$kS`d_}wwxg%ON* zE68uadq(41gpoWCyTQQid`Ropo^Oz=y{V4l;CG%!Xfo!NV2Y zn;3suRhypX-03v8@QZ=|f+BL$rh)3hy}XYXPzn#&e3YcRWf2y%yo8JcM}x4dp}p+o zZ)uF$!Y4R>@W)`@=>Gs6n*j`eJ!cAx(A)E6-k1{S^NShOI_26dX)ABpiF4bGk#zxQ zvQ`WG_0tBf4!-A;x}q_I_?_HC!LsP*@kXn^nvZ!G{@sXw2@Xcq0K9fioR|tPa(3P^ zdf5f^8{|gRBRbyIQ~jQUjc?x2LQ?S<62Abq%bNzatl8RUt&lTRp-5t8%?X0oF%vTi zVLPEG(VF`YRN{Cv(44;*DD(04>M9zB54?pzVcz#|OS2pJ;pV5zTm*%)ESB~Zu&Iwr zeUG{aS<~L*IxHJqokGg`Y|JfW%Krmcu(Q+ND`;I*mRy*yfidHrMjMY&*GJB?YLE1Q zbE;Lb+TLYN!D~b2+t7VegGL1WQsg2TV#>r%)Wvx-=#lVI33fNT*ZL`O1b$iG*BlJl zGZ3Ywdh7n-6J2cD$n~(TLG1cki%OCyn7F} z1`U|B&vkV|FZ|!^(}DL9YpOD2+s;k1eNxPy*w3sapeuX5PBfuh+2xLgqEX$u?kZlp zolpaps0LJ1EBq`7y)Y_d?(69=$DFzN7_1#FJfAqw`X%CSe!fPt=0>UGjwN1CVRew; zk3s#jr5mu`XQ)vpQc{)%DM6)I9h!p!6;-O2FWc~c6mI2Ew>C>_#am_()x8n_w8UK) zqjdK7SN7txdirP|mH?NA;w2n`{+S!8U)me(y66*i|KWw)UW4<}G%UrF*&JpRxx9z*3`mW!ztUKtmfN#tMswt$CZzVXtdF`%pVqF)wHf{9q#OkM2; zIiRzq@&hq1B|9}4Zeyiq-t|4u13EPaZO5lh6nuC7&MNs{R-Hh!_Jr+$lPvAhw**|0 zh4~%r?^wQa0o4XEl`+QKSGk4X*iIX(T~M+w2loBUDLwZlVt>-}%j!82WmVC|?mt@w zR^`)p{^)i9DZjlQJryMf_3;h*|H$hy`zQ0;JKRU0BMKjL95TLk1!j&JQQ{{x6(jA{B_uxT~X+0o0c2h$U3`BD1vj*<1g&&uqE2eBECd2G+~AuyBcQOU|FhPrBb&y1WoO6 zV3*~o)Y?%>KXb=8nmRrEc1ems_o1?v?IsJpng?3>64bT2`9b;4X0Qh**;WN^Ra)_1 zW7=yi8=o!t)lnBcEcgSOn? zm7R{w)$NoKljrJMhy{_$o}KRo`k<(!RosNlhfFz73fXS9fitCl6?0C#zo$|%ctt$c zBK5Q?)OEy=^6=Vrmi#?d<}5cqu4xe%=Ho}f+S?!(G9WWu#3&1o*jIm0vcoTf48B4U z+SNzhKsV>P1haEjn4EsR)F+Rt2RZW%zMe{Eh~eFdnQ^IAq4~NzE`u}^t4UNwzA698 z&In3Ym%6I1bQ^x!jEBoK7{sx+>Z_MxyQ-Oeh)7Sy`QnP+IQ}SO5f}Hm5Y0FABVOk# zYD|h8eyb{bd|eYX7L>fRMYwKvRsBVilv_w71SE@REgI}M(CZfl?Q`Q#xpwb0`d=Ynpl%<7P!X!X&Z7$IKaT|($cpF={nRMbzOz_QfwlvvH{iBWWvv*SPx6bQd_byAVMUI$7psv}mnp#I) zOB$F7XhCQ>bHw2&y|NPuOFz#TsGcGACskIjA|iXy+nAOQUpve!r%E6!9AjW!_v>>v zezo<%3)HIaUH!+@pd5YqU0k(5ZDY z_%Qs?YB{aFHAT$N=PPWMTTMBtMS22n4gE9^v?nEATF%~vC2=@l26wu4JqgSa+6cxK zTxOpBspflLzva$9rcULPTBd4)hLr9Lo#e}-VqrZ;DdI?AEJk)myoe*zIxiJD%c#$> zqCm~P(1@sHusU!i?`8MzcfwI>1THHjg6&-f)}&o($9x)|n`InxJ1lfl_LgHV#d)RA zlB9O%4k}#|8d?ai%SDWH`^ZdrWYV$rNG$72yCNXBNgvz=9_&;Y&X z7hhtl&1hWf`UQ_Q{(3}y@m-xExIZxLSud~#QuEP~pUaqJ7W4j$rO@dw=(tf`_3yc= zW;P?IVv%;8LSH5||3GfE3CKxW?XZ8hjo|r1Gialh@Q9H);vjKOtlMMbC75EtOtr$M z;uO7ZHz$f}ON5LI$2sZ_(9_kF0HcLepMmTO;|zuhA}f#d$iQ+HLJ&h<+W9=Kpnb(j zG}lz8UFDM7SEPmp@cVzo({ciU%FiZ@fTr5QB z=fQcaYdJlIxfT#riJj_(PY5si!+amOY8Lkg`s6D+$eBx3TE4BHhkCtYNTw!79ZW!< zwdd}UWd%#%gu=IJ7%Y9^!g>=TMjrM){S)iow%I4sYtMK5FlzrRU*idKOv@PI2W+)t zJBcaQX8bz^=}-JF|8GHuiMlgw`8keRu@Q>!5<20PL&LO-7Rg)Y+D zI=Ehw-*^hzv=08*GwMCdnGv&73*MG5C8g3X@63(9Ie!A$5tOXBRr&FZ_|w`J3)WXbssMm6xPIYZgcT;UKQ!M}xesQcY!NyrlHM;8_C%yM{bMBtI|hYxkMVa=&PzQv@q*ovTmpxrfg*&Z#D^_OB;!hD%)VB-Ddz3 zBnvfVix3R=~c z7=A$}Y09VGe{L=b0?Th_uC4MM_9Y4NWbVY_Ld^onz8>Z^jFiDZs1cbk4zNqZDJW~b>`>BkDsZ2F7Oj=sSdp^Dr+jv zA`4>m2wIsE++xz#g=p$3N|J9QQ15i(A=p}Io_Rzi(Ehv#X`Jn3ujHU6oSXI~0-PNL zCZwG)n2EPZyJ($l4|ayFUzpH^ugoRn>z5_=O%Yi%Wj3m6t_&n^t;u4y=`dX@LAk2( zJ3GGlIScpPT;2xsoy(Y@%cC*NbbF>0SP-UE9lx)5nu{X|g~LB&)|T%c`*o!=$W{zr<@d25h6$ER&pQP>z(28z;v+_* z?!6?OW8eVzkCP6l>ai2^!GKN#2K46(DXkBV-B%u8Q!KxY=!nz!WQPMl zW$v&E{yZ&nq@ku15w>VR(QFtXiYD zKj?Id$S`FSuT%`?Qu-U49Seqnkh33TNZ1hrawv11`g`w=z`fe_{7pbC0DAy5a()<1!r4MMt03}$fnnUL31@`Sg$(?d zJ}}w!+r0!YZY=b}anG4>0={-P@ZK@6RSBD@D~N^XM-2b~fs@e7!Us!;9W;t4Lv4Si z*?vj|ZDTHRb#{)o8uGI~9ODw}4rnf-M77j!=c5mk9=?0xhf7E<>6|1~4k}H9?mdux z4>&j92D&__IpUQ_#}P(q<%?gd{D3V7dCfLE)KKDa-W>ATF8P+{1A(pzSaTsZ0fRgO zM#7kx*{d%M4!;d>@Sz$;kMY(;7IMS&&752IN4c|=o{+6JAL|-Ff9uui>nYK@$=+@D zb+{CLnY&-N(S!d9=OuQt@Dk2I(+obwXKhd_Xsh6c;ktcutpGB(-#_T4uZ|txm=-5W zFWd0ab4L8i?JzNFa|vW}Ufz2$s(;j^O}|q%SMMWez~&}p{}ad*v3Y2W>SiRQoj9-v zHo{}w2D3wQoA#!1vvUR9R-2y}g#X}JX6J;L<$TxL`hNjN3c2;XL#dlfHi$7vXB>v= z!`8H2&QV#5JDh@g3gY}ts4k(a&6nRAksc30j-*#*JIUyeDzoZWvPZbMRE&T}Jo;AN zp>USb1zv~c!K#{cD{XEZl3ltH)C#SsP9%35h+*8V_JbG4af=kzCJY%6W;^o^U z@M*G6u`UNpezjM=qe!JA`EfFkH)pL;k-5kjtB)U=OP)PPtyPl*0tN@CrA3SAFdr=B zas4Y}Nx2go9COFLViI=}4;3}O^w{|-Gr<)0)K14!b)@;ww1~Tvdx2c#?w;{vk&*3P zcZPL@kcg!t(1YH!^{cN~m#6lnBDcjb5Jd2V>YVayg)6WOwy62|DCZ&GU7mpveItIXFG4NXhTl)}Y5i4;1kG zdUc}0+haoAhye8+Yi`o+eknedL>*o~Jw!O+Y-b zPeJMHRmSUygZg%>n^Xgqz#N?S_NCC5V;4}jc-%Q&8@Q@1g!Ey4Afb^=Db|I9> z7X)=$FLOl9>M3a{mN_}}tz8>On^wF?V@4#Ak~>s-Hj_51c924)-2OtngTnqBheeD9 z1I)+H-`+Kp>T)BRr9~YGG`$scyJfdTP*?9CQ&*4fq{#VPb``0y6c9p_oE+3rz&=nA z1}Aq~!=C$`_wIN83BMEAG(;B2$s}+qqClKUxa6EE#c{t5+D4jma`Frel^j=SBQXQF zlh`+>ttVrdX>?)VMJ=t%?S2kd3&&cnx`V=Sz_>w=mx|T9kSybmmHh`pQB4$Yv;Y`) zQPkHo;UToqj&)ad9N~!I;;W5HX^4yDaPY`Jl-rBP= z1#qQpCO+TC(NZQnOjQrV**3*2-S4f>nk5h88KjAPomD?2r9 znFDWcTJ$L|cSEggQb-;_o|)r;(yo7JkCjS*0LQ&uNOraX0|y^kvlIbYnVT(~Vzse1 zT}&6WSe1YTamH&S#Hl1&Eb{Fj5CJ}!uH@Pz(i7J!laAHQd_`4{!^8}My+%7z(#1z~ zd0iAyQ*)>&>p|v<06;N88Kt6?fGDDZGe+PBlOK&O6u>B=iY@^~2Z2H9OaRa*rOhD; zj&nt5TU$q{#v@kQaz5Za8``N!3q>O<4Xh3bG_C?hyVwf*SN1~in9{s8d95<9nRepd zFjtbe%N+hy@&|@Axpi$?3wRT7lR6}cgMXRTUAL%0r}-9=`l zV^svSXFqjkwU~j%Fb+A#YQ(*@VxaJ+zI)eYEL(}%r+@H~+PQ1x%et2VvVLrI_N`*y zLpjGoi`1f3mA4XclhU~>;k}2=vjgsd&NE$}u1A}}1axERT&?n%9gp51mGm{}(QZ1k z%BxK~qmtAa<1V<|%6Z39eNAFtPQlLR%w&AK1~7f=e^65!Nrm~Fg3HwVS0jHjt3F8` z%J#*51qXdu^H_-L(N6O0Hg+%_w-d%|Dp`vhmjIKJSD(t6d3aTZ8C0mr&1SvPcR3&v z{oa-A(^?)iLr6-^fH+S70RFmFZO*s>vyezR#ZrQ8xLgCCxbId_20MY;Gtdn4T0M3- zJx5$hsUB6fc_~m$7~iXO?aznM*GB9t~me{1fJCi zbw;sT7#gc);lMfOvQJvSs2#Go&Rf1}hH`UMYZPf|e>`pu03Nj*%y)8g&rwt9sgxu5 zjV0V$55KUfxi4~t-nkj21K0|fCU){M-kvSPjt^=FaxJ{$0;Tdn!RIvf8wOZ_J2f?> z}xkhh)u!jYo@gDznYG(IXp1q1Jbj#9|K$5w2_nm9FDy!Zq_?}B_mIjnMQJlioUb7khID| zx4)sP*I{!XJ9w8ulR?xT7W;>kZ&6wHsAP~XGtGLBk94o?UByvU#5iC(SDo6x(8xy8 z4NFF`mYSLl4Yip#1d)MU{NF5{Fnud->TpPlJTF|;8KCn%VbFJ^RnbXCSRVPS>+;4} zjy43L5|;AJw(||vj*sK+NX|2Wyz(;U@mA3B3;A-N^ zi_~L2ln9z;7~Bc`J!z`p*PQ1)Di1hggWvR~sVyPE+O+o6EKa8c;EzfOBygabbYLP4 z%ilF3k&yd8F2bl*AxNJijD2ZdKb#)ItZo~+FOyRc~&DMJ%twP z!tD-Oe8xer1K%~Huz3XFeQP=+vQpXaz%_kX{hi#f;Yp^Q*qKYhu!*&6hwiz-6|1P( z09cW^^PXylgCYP=H++ye^{uN<<24hLc>Hugan-%)(QCPo3rBpPp+6rYzq zhL%QEPzV`3)J9Bl!zZW%y=Km+`Hx()79TJBB=xCA(wOb-Rs6+alh&jT4gmYZ^rN)RLmA1CV!3U)+oy7c$5zCTOU|=wFp7m}v3gZBR zdjZaBns(0zJuy}mO~R( U)TTw=tivogI5l>*#uwE8*<&4X;s5{u literal 0 HcmV?d00001 diff --git a/data/scripts/download_weights.sh b/data/scripts/download_weights.sh new file mode 100644 index 0000000..e9fa653 --- /dev/null +++ b/data/scripts/download_weights.sh @@ -0,0 +1,20 @@ +#!/bin/bash +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# Download latest models from https://github.com/ultralytics/yolov5/releases +# Example usage: bash path/to/download_weights.sh +# parent +# └── yolov5 +# ├── yolov5s.pt ← downloads here +# ├── yolov5m.pt +# └── ... + +python - <= cls >= 0, f'incorrect class index {cls}' + + # Write YOLO label + if id not in shapes: + shapes[id] = Image.open(file).size + box = xyxy2xywhn(box[None].astype(np.float), w=shapes[id][0], h=shapes[id][1], clip=True) + with open((labels / id).with_suffix('.txt'), 'a') as f: + f.write(f"{cls} {' '.join(f'{x:.6f}' for x in box[0])}\n") # write label.txt + except Exception as e: + print(f'WARNING: skipping one label for {file}: {e}') + + + # Download manually from https://challenge.xviewdataset.org + dir = Path(yaml['path']) # dataset root dir + # urls = ['https://d307kc0mrhucc3.cloudfront.net/train_labels.zip', # train labels + # 'https://d307kc0mrhucc3.cloudfront.net/train_images.zip', # 15G, 847 train images + # 'https://d307kc0mrhucc3.cloudfront.net/val_images.zip'] # 5G, 282 val images (no labels) + # download(urls, dir=dir, delete=False) + + # Convert labels + convert_labels(dir / 'xView_train.geojson') + + # Move images + images = Path(dir / 'images') + images.mkdir(parents=True, exist_ok=True) + Path(dir / 'train_images').rename(dir / 'images' / 'train') + Path(dir / 'val_images').rename(dir / 'images' / 'val') + + # Split + autosplit(dir / 'images' / 'train') diff --git a/detect.py b/detect.py new file mode 100644 index 0000000..15f22f5 --- /dev/null +++ b/detect.py @@ -0,0 +1,260 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Run inference on images, videos, directories, streams, etc. + +Usage - sources: + $ python path/to/detect.py --weights yolov5s.pt --source 0 # webcam + img.jpg # image + vid.mp4 # video + path/ # directory + path/*.jpg # glob + 'https://youtu.be/Zgi9g1ksQHc' # YouTube + 'rtsp://example.com/media.mp4' # RTSP, RTMP, HTTP stream + +Usage - formats: + $ python path/to/detect.py --weights yolov5s.pt # PyTorch + yolov5s.torchscript # TorchScript + yolov5s.onnx # ONNX Runtime or OpenCV DNN with --dnn + yolov5s.xml # OpenVINO + yolov5s.engine # TensorRT + yolov5s.mlmodel # CoreML (MacOS-only) + yolov5s_saved_model # TensorFlow SavedModel + yolov5s.pb # TensorFlow GraphDef + yolov5s.tflite # TensorFlow Lite + yolov5s_edgetpu.tflite # TensorFlow Edge TPU +""" + +import argparse +import os +import sys +import time +from pathlib import Path + +import cv2 +import torch +import torch.backends.cudnn as cudnn + +FILE = Path(__file__).resolve() +ROOT = FILE.parents[0] # YOLOv5 root directory +if str(ROOT) not in sys.path: + sys.path.append(str(ROOT)) # add ROOT to PATH +ROOT = Path(os.path.relpath(ROOT, Path.cwd())) # relative + +from models.common import DetectMultiBackend +from utils.datasets import IMG_FORMATS, VID_FORMATS, LoadImages, LoadStreams +from utils.general import (LOGGER, check_file, check_img_size, check_imshow, check_requirements, colorstr, + increment_path, non_max_suppression, print_args, scale_coords, strip_optimizer, xyxy2xywh) +from utils.plots import Annotator, colors, save_one_box +from utils.torch_utils import select_device, time_sync + + +@torch.no_grad() +def run(weights=ROOT / 'yolov5s.pt', # model.pt path(s) + source=ROOT / 'data/images', # file/dir/URL/glob, 0 for webcam + data=ROOT / 'data/coco128.yaml', # dataset.yaml path + imgsz=(640, 640), # inference size (height, width) + conf_thres=0.25, # confidence threshold + iou_thres=0.45, # NMS IOU threshold + max_det=1000, # maximum detections per image + device='', # cuda device, i.e. 0 or 0,1,2,3 or cpu + view_img=False, # show results + save_txt=False, # save results to *.txt + save_conf=False, # save confidences in --save-txt labels + save_crop=False, # save cropped prediction boxes + nosave=False, # do not save images/videos + classes=None, # filter by class: --class 0, or --class 0 2 3 + agnostic_nms=False, # class-agnostic NMS + augment=False, # augmented inference + visualize=False, # visualize features + update=False, # update all models + project=ROOT / 'runs/detect', # save results to project/name + name='exp', # save results to project/name + exist_ok=False, # existing project/name ok, do not increment + line_thickness=3, # bounding box thickness (pixels) + hide_labels=False, # hide labels + hide_conf=False, # hide confidences + half=False, # use FP16 half-precision inference + dnn=False, # use OpenCV DNN for ONNX inference + ): + source = str(source) + save_img = not nosave and not source.endswith('.txt') # save inference images + is_file = Path(source).suffix[1:] in (IMG_FORMATS + VID_FORMATS) + is_url = source.lower().startswith(('rtsp://', 'rtmp://', 'http://', 'https://')) + webcam = source.isnumeric() or source.endswith('.txt') or (is_url and not is_file) + if is_url and is_file: + source = check_file(source) # download + + # Directories + save_dir = increment_path(Path(project) / name, exist_ok=exist_ok) # increment run + (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True) # make dir + + # Load model + device = select_device(device) + model = DetectMultiBackend(weights, device=device, dnn=dnn, data=data) + stride, names, pt, jit, onnx, engine = model.stride, model.names, model.pt, model.jit, model.onnx, model.engine + imgsz = check_img_size(imgsz, s=stride) # check image size + + # Half + half &= (pt or jit or onnx or engine) and device.type != 'cpu' # FP16 supported on limited backends with CUDA + if pt or jit: + model.model.half() if half else model.model.float() + + # Dataloader + if webcam: + view_img = check_imshow() + cudnn.benchmark = True # set True to speed up constant image size inference + dataset = LoadStreams(source, img_size=imgsz, stride=stride, auto=pt) + bs = len(dataset) # batch_size + else: + dataset = LoadImages(source, img_size=imgsz, stride=stride, auto=pt) + bs = 1 # batch_size + vid_path, vid_writer = [None] * bs, [None] * bs + + # Run inference + model.warmup(imgsz=(1 if pt else bs, 3, *imgsz), half=half) # warmup + dt, seen = [0.0, 0.0, 0.0], 0 + for path, im, im0s, vid_cap, s in dataset: + t1 = time_sync() + im = torch.from_numpy(im).to(device) + im = im.half() if half else im.float() # uint8 to fp16/32 + im /= 255 # 0 - 255 to 0.0 - 1.0 + if len(im.shape) == 3: + im = im[None] # expand for batch dim + t2 = time_sync() + dt[0] += t2 - t1 + + # Inference + visualize = increment_path(save_dir / Path(path).stem, mkdir=True) if visualize else False + pred = model(im, augment=augment, visualize=visualize) + t3 = time_sync() + dt[1] += t3 - t2 + + # NMS + pred = non_max_suppression(pred, conf_thres, iou_thres, classes, agnostic_nms, max_det=max_det) + dt[2] += time_sync() - t3 + + # Second-stage classifier (optional) + # pred = utils.general.apply_classifier(pred, classifier_model, im, im0s) + + # Process predictions + for i, det in enumerate(pred): # per image + seen += 1 + if webcam: # batch_size >= 1 + p, im0, frame = path[i], im0s[i].copy(), dataset.count + s += f'{i}: ' + else: + p, im0, frame = path, im0s.copy(), getattr(dataset, 'frame', 0) + + p = Path(p) # to Path + save_path = str(save_dir / p.name) # im.jpg + txt_path = str(save_dir / 'labels' / p.stem) + ('' if dataset.mode == 'image' else f'_{frame}') # im.txt + s += '%gx%g ' % im.shape[2:] # print string + gn = torch.tensor(im0.shape)[[1, 0, 1, 0]] # normalization gain whwh + imc = im0.copy() if save_crop else im0 # for save_crop + annotator = Annotator(im0, line_width=line_thickness, example=str(names)) + if len(det): + # Rescale boxes from img_size to im0 size + det[:, :4] = scale_coords(im.shape[2:], det[:, :4], im0.shape).round() + + # Print results + for c in det[:, -1].unique(): + n = (det[:, -1] == c).sum() # detections per class + s += f"{n} {names[int(c)]}{'s' * (n > 1)}, " # add to string + + # Write results + for *xyxy, conf, cls in reversed(det): + if save_txt: # Write to file + xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist() # normalized xywh + line = (cls, *xywh, conf) if save_conf else (cls, *xywh) # label format + with open(txt_path + '.txt', 'a') as f: + f.write(('%g ' * len(line)).rstrip() % line + '\n') + + if save_img or save_crop or view_img: # Add bbox to image + c = int(cls) # integer class + label = None if hide_labels else (names[c] if hide_conf else f'{names[c]} {conf:.2f}') + annotator.box_label(xyxy, label, color=colors(c, True)) + if save_crop: + save_one_box(xyxy, imc, file=save_dir / 'crops' / names[c] / f'{p.stem}.jpg', BGR=True) + + # Stream results + im0 = annotator.result() + if view_img: + cv2.imshow(str(p), im0) + cv2.waitKey(1) # 1 millisecond + + # Save results (image with detections) + if save_img: + if dataset.mode == 'image': + cv2.imwrite(save_path, im0) + else: # 'video' or 'stream' + if vid_path[i] != save_path: # new video + vid_path[i] = save_path + if isinstance(vid_writer[i], cv2.VideoWriter): + vid_writer[i].release() # release previous video writer + if vid_cap: # video + fps = vid_cap.get(cv2.CAP_PROP_FPS) + w = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH)) + h = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT)) + else: # stream + fps, w, h = 30, im0.shape[1], im0.shape[0] + save_path = str(Path(save_path).with_suffix('.mp4')) # force *.mp4 suffix on results videos + vid_writer[i] = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*'mp4v'), fps, (w, h)) + vid_writer[i].write(im0) + + # Print time (inference-only) + LOGGER.info(f'{s}Done. ({t3 - t2:.3f}s fps:{1./(t3 - t2)}) ') + #print() + # Print results + t = tuple(x / seen * 1E3 for x in dt) # speeds per image + LOGGER.info(f'Speed: %.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape {(1, 3, *imgsz)}' % t) + if save_txt or save_img: + s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else '' + LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}{s}") + if update: + strip_optimizer(weights) # update model (to fix SourceChangeWarning) + + +def parse_opt(): + parser = argparse.ArgumentParser() + #parser.add_argument('--weights', nargs='+', type=str, default=ROOT / 'yolov5s.pt', help='model path(s)') + parser.add_argument('--weights', nargs='+', type=str, default=ROOT / 'best.pt', help='model path(s)') + #parser.add_argument('--source', type=str, default=ROOT / 'data/images/hand', help='file/dir/URL/glob, 0 for webcam') + parser.add_argument('--source', type=str, default=0, help='file/dir/URL/glob, 0 for webcam') + parser.add_argument('--data', type=str, default=ROOT / 'data/hand.yaml', help='(optional) dataset.yaml path') + parser.add_argument('--imgsz', '--img', '--img-size', nargs='+', type=int, default=[640], help='inference size h,w') + parser.add_argument('--conf-thres', type=float, default=0.25, help='confidence threshold') + parser.add_argument('--iou-thres', type=float, default=0.45, help='NMS IoU threshold') + parser.add_argument('--max-det', type=int, default=1000, help='maximum detections per image') + parser.add_argument('--device', default='0', help='cuda device, i.e. 0 or 0,1,2,3 or cpu') + parser.add_argument('--view-img', action='store_true', help='show results') + parser.add_argument('--save-txt', action='store_true', help='save results to *.txt') + parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels') + parser.add_argument('--save-crop', action='store_true', help='save cropped prediction boxes') + parser.add_argument('--nosave', action='store_true', help='do not save images/videos') + parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --classes 0, or --classes 0 2 3') + parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS') + parser.add_argument('--augment', action='store_true', help='augmented inference') + parser.add_argument('--visualize', action='store_true', help='visualize features') + parser.add_argument('--update', action='store_true', help='update all models') + parser.add_argument('--project', default=ROOT / 'runs/detect', help='save results to project/name') + parser.add_argument('--name', default='exp', help='save results to project/name') + parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment') + parser.add_argument('--line-thickness', default=3, type=int, help='bounding box thickness (pixels)') + parser.add_argument('--hide-labels', default=False, action='store_true', help='hide labels') + parser.add_argument('--hide-conf', default=False, action='store_true', help='hide confidences') + parser.add_argument('--half', action='store_true', help='use FP16 half-precision inference') + parser.add_argument('--dnn', action='store_true', help='use OpenCV DNN for ONNX inference') + opt = parser.parse_args() + opt.imgsz *= 2 if len(opt.imgsz) == 1 else 1 # expand + print_args(FILE.stem, opt) + return opt + + +def main(opt): + check_requirements(exclude=('tensorboard', 'thop')) + run(**vars(opt)) + + +if __name__ == "__main__": + opt = parse_opt() + main(opt) diff --git a/export.py b/export.py new file mode 100644 index 0000000..d75bfb4 --- /dev/null +++ b/export.py @@ -0,0 +1,562 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Export a YOLOv5 PyTorch model to other formats. TensorFlow exports authored by https://github.com/zldrobit + +Format | `export.py --include` | Model +--- | --- | --- +PyTorch | - | yolov5s.pt +TorchScript | `torchscript` | yolov5s.torchscript +ONNX | `onnx` | yolov5s.onnx +OpenVINO | `openvino` | yolov5s_openvino_model/ +TensorRT | `engine` | yolov5s.engine +CoreML | `coreml` | yolov5s.mlmodel +TensorFlow SavedModel | `saved_model` | yolov5s_saved_model/ +TensorFlow GraphDef | `pb` | yolov5s.pb +TensorFlow Lite | `tflite` | yolov5s.tflite +TensorFlow Edge TPU | `edgetpu` | yolov5s_edgetpu.tflite +TensorFlow.js | `tfjs` | yolov5s_web_model/ + +Requirements: + $ pip install -r requirements.txt coremltools onnx onnx-simplifier onnxruntime openvino-dev tensorflow-cpu # CPU + $ pip install -r requirements.txt coremltools onnx onnx-simplifier onnxruntime-gpu openvino-dev tensorflow # GPU + +Usage: + $ python path/to/export.py --weights yolov5s.pt --include torchscript onnx openvino engine coreml tflite ... + +Inference: + $ python path/to/detect.py --weights yolov5s.pt # PyTorch + yolov5s.torchscript # TorchScript + yolov5s.onnx # ONNX Runtime or OpenCV DNN with --dnn + yolov5s.xml # OpenVINO + yolov5s.engine # TensorRT + yolov5s.mlmodel # CoreML (MacOS-only) + yolov5s_saved_model # TensorFlow SavedModel + yolov5s.pb # TensorFlow GraphDef + yolov5s.tflite # TensorFlow Lite + yolov5s_edgetpu.tflite # TensorFlow Edge TPU + +TensorFlow.js: + $ cd .. && git clone https://github.com/zldrobit/tfjs-yolov5-example.git && cd tfjs-yolov5-example + $ npm install + $ ln -s ../../yolov5/yolov5s_web_model public/yolov5s_web_model + $ npm start +""" + +import argparse +import json +import os +import platform +import subprocess +import sys +import time +import warnings +from pathlib import Path + +import pandas as pd +import torch +import torch.nn as nn +from torch.utils.mobile_optimizer import optimize_for_mobile + +FILE = Path(__file__).resolve() +ROOT = FILE.parents[0] # YOLOv5 root directory +if str(ROOT) not in sys.path: + sys.path.append(str(ROOT)) # add ROOT to PATH +ROOT = Path(os.path.relpath(ROOT, Path.cwd())) # relative + +from models.common import Conv +from models.experimental import attempt_load +from models.yolo import Detect +from utils.activations import SiLU +from utils.datasets import LoadImages +from utils.general import (LOGGER, check_dataset, check_img_size, check_requirements, check_version, colorstr, + file_size, print_args, url2file) +from utils.torch_utils import select_device + + +def export_formats(): + # YOLOv5 export formats + x = [['PyTorch', '-', '.pt'], + ['TorchScript', 'torchscript', '.torchscript'], + ['ONNX', 'onnx', '.onnx'], + ['OpenVINO', 'openvino', '_openvino_model'], + ['TensorRT', 'engine', '.engine'], + ['CoreML', 'coreml', '.mlmodel'], + ['TensorFlow SavedModel', 'saved_model', '_saved_model'], + ['TensorFlow GraphDef', 'pb', '.pb'], + ['TensorFlow Lite', 'tflite', '.tflite'], + ['TensorFlow Edge TPU', 'edgetpu', '_edgetpu.tflite'], + ['TensorFlow.js', 'tfjs', '_web_model']] + return pd.DataFrame(x, columns=['Format', 'Argument', 'Suffix']) + + +def export_torchscript(model, im, file, optimize, prefix=colorstr('TorchScript:')): + # YOLOv5 TorchScript model export + try: + LOGGER.info(f'\n{prefix} starting export with torch {torch.__version__}...') + f = file.with_suffix('.torchscript') + + ts = torch.jit.trace(model, im, strict=False) + d = {"shape": im.shape, "stride": int(max(model.stride)), "names": model.names} + extra_files = {'config.txt': json.dumps(d)} # torch._C.ExtraFilesMap() + if optimize: # https://pytorch.org/tutorials/recipes/mobile_interpreter.html + optimize_for_mobile(ts)._save_for_lite_interpreter(str(f), _extra_files=extra_files) + else: + ts.save(str(f), _extra_files=extra_files) + + LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') + return f + except Exception as e: + LOGGER.info(f'{prefix} export failure: {e}') + + +def export_onnx(model, im, file, opset, train, dynamic, simplify, prefix=colorstr('ONNX:')): + # YOLOv5 ONNX export + try: + check_requirements(('onnx',)) + import onnx + + LOGGER.info(f'\n{prefix} starting export with onnx {onnx.__version__}...') + f = file.with_suffix('.onnx') + + torch.onnx.export(model, im, f, verbose=False, opset_version=opset, + training=torch.onnx.TrainingMode.TRAINING if train else torch.onnx.TrainingMode.EVAL, + do_constant_folding=not train, + input_names=['images'], + output_names=['output'], + dynamic_axes={'images': {0: 'batch', 2: 'height', 3: 'width'}, # shape(1,3,640,640) + 'output': {0: 'batch', 1: 'anchors'} # shape(1,25200,85) + } if dynamic else None) + + # Checks + model_onnx = onnx.load(f) # load onnx model + onnx.checker.check_model(model_onnx) # check onnx model + # LOGGER.info(onnx.helper.printable_graph(model_onnx.graph)) # print + + # Simplify + if simplify: + try: + check_requirements(('onnx-simplifier',)) + import onnxsim + + LOGGER.info(f'{prefix} simplifying with onnx-simplifier {onnxsim.__version__}...') + model_onnx, check = onnxsim.simplify( + model_onnx, + dynamic_input_shape=dynamic, + input_shapes={'images': list(im.shape)} if dynamic else None) + assert check, 'assert check failed' + onnx.save(model_onnx, f) + except Exception as e: + LOGGER.info(f'{prefix} simplifier failure: {e}') + LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') + return f + except Exception as e: + LOGGER.info(f'{prefix} export failure: {e}') + + +def export_openvino(model, im, file, prefix=colorstr('OpenVINO:')): + # YOLOv5 OpenVINO export + try: + check_requirements(('openvino-dev',)) # requires openvino-dev: https://pypi.org/project/openvino-dev/ + import openvino.inference_engine as ie + + LOGGER.info(f'\n{prefix} starting export with openvino {ie.__version__}...') + f = str(file).replace('.pt', '_openvino_model' + os.sep) + + cmd = f"mo --input_model {file.with_suffix('.onnx')} --output_dir {f}" + subprocess.check_output(cmd, shell=True) + + LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') + return f + except Exception as e: + LOGGER.info(f'\n{prefix} export failure: {e}') + + +def export_coreml(model, im, file, prefix=colorstr('CoreML:')): + # YOLOv5 CoreML export + try: + check_requirements(('coremltools',)) + import coremltools as ct + + LOGGER.info(f'\n{prefix} starting export with coremltools {ct.__version__}...') + f = file.with_suffix('.mlmodel') + + ts = torch.jit.trace(model, im, strict=False) # TorchScript model + ct_model = ct.convert(ts, inputs=[ct.ImageType('image', shape=im.shape, scale=1 / 255, bias=[0, 0, 0])]) + ct_model.save(f) + + LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') + return ct_model, f + except Exception as e: + LOGGER.info(f'\n{prefix} export failure: {e}') + return None, None + + +def export_engine(model, im, file, train, half, simplify, workspace=4, verbose=False, prefix=colorstr('TensorRT:')): + # YOLOv5 TensorRT export https://developer.nvidia.com/tensorrt + try: + check_requirements(('tensorrt',)) + import tensorrt as trt + + if trt.__version__[0] == '7': # TensorRT 7 handling https://github.com/ultralytics/yolov5/issues/6012 + grid = model.model[-1].anchor_grid + model.model[-1].anchor_grid = [a[..., :1, :1, :] for a in grid] + export_onnx(model, im, file, 12, train, False, simplify) # opset 12 + model.model[-1].anchor_grid = grid + else: # TensorRT >= 8 + check_version(trt.__version__, '8.0.0', hard=True) # require tensorrt>=8.0.0 + export_onnx(model, im, file, 13, train, False, simplify) # opset 13 + onnx = file.with_suffix('.onnx') + + LOGGER.info(f'\n{prefix} starting export with TensorRT {trt.__version__}...') + assert im.device.type != 'cpu', 'export running on CPU but must be on GPU, i.e. `python export.py --device 0`' + assert onnx.exists(), f'failed to export ONNX file: {onnx}' + f = file.with_suffix('.engine') # TensorRT engine file + logger = trt.Logger(trt.Logger.INFO) + if verbose: + logger.min_severity = trt.Logger.Severity.VERBOSE + + builder = trt.Builder(logger) + config = builder.create_builder_config() + config.max_workspace_size = workspace * 1 << 30 + + flag = (1 << int(trt.NetworkDefinitionCreationFlag.EXPLICIT_BATCH)) + network = builder.create_network(flag) + parser = trt.OnnxParser(network, logger) + if not parser.parse_from_file(str(onnx)): + raise RuntimeError(f'failed to load ONNX file: {onnx}') + + inputs = [network.get_input(i) for i in range(network.num_inputs)] + outputs = [network.get_output(i) for i in range(network.num_outputs)] + LOGGER.info(f'{prefix} Network Description:') + for inp in inputs: + LOGGER.info(f'{prefix}\tinput "{inp.name}" with shape {inp.shape} and dtype {inp.dtype}') + for out in outputs: + LOGGER.info(f'{prefix}\toutput "{out.name}" with shape {out.shape} and dtype {out.dtype}') + + half &= builder.platform_has_fast_fp16 + LOGGER.info(f'{prefix} building FP{16 if half else 32} engine in {f}') + if half: + config.set_flag(trt.BuilderFlag.FP16) + with builder.build_engine(network, config) as engine, open(f, 'wb') as t: + t.write(engine.serialize()) + LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') + return f + except Exception as e: + LOGGER.info(f'\n{prefix} export failure: {e}') + + +def export_saved_model(model, im, file, dynamic, + tf_nms=False, agnostic_nms=False, topk_per_class=100, topk_all=100, iou_thres=0.45, + conf_thres=0.25, keras=False, prefix=colorstr('TensorFlow SavedModel:')): + # YOLOv5 TensorFlow SavedModel export + try: + import tensorflow as tf + from tensorflow.python.framework.convert_to_constants import convert_variables_to_constants_v2 + + from models.tf import TFDetect, TFModel + + LOGGER.info(f'\n{prefix} starting export with tensorflow {tf.__version__}...') + f = str(file).replace('.pt', '_saved_model') + batch_size, ch, *imgsz = list(im.shape) # BCHW + + tf_model = TFModel(cfg=model.yaml, model=model, nc=model.nc, imgsz=imgsz) + im = tf.zeros((batch_size, *imgsz, 3)) # BHWC order for TensorFlow + _ = tf_model.predict(im, tf_nms, agnostic_nms, topk_per_class, topk_all, iou_thres, conf_thres) + inputs = tf.keras.Input(shape=(*imgsz, 3), batch_size=None if dynamic else batch_size) + outputs = tf_model.predict(inputs, tf_nms, agnostic_nms, topk_per_class, topk_all, iou_thres, conf_thres) + keras_model = tf.keras.Model(inputs=inputs, outputs=outputs) + keras_model.trainable = False + keras_model.summary() + if keras: + keras_model.save(f, save_format='tf') + else: + m = tf.function(lambda x: keras_model(x)) # full model + spec = tf.TensorSpec(keras_model.inputs[0].shape, keras_model.inputs[0].dtype) + m = m.get_concrete_function(spec) + frozen_func = convert_variables_to_constants_v2(m) + tfm = tf.Module() + tfm.__call__ = tf.function(lambda x: frozen_func(x), [spec]) + tfm.__call__(im) + tf.saved_model.save( + tfm, + f, + options=tf.saved_model.SaveOptions(experimental_custom_gradients=False) if + check_version(tf.__version__, '2.6') else tf.saved_model.SaveOptions()) + LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') + return keras_model, f + except Exception as e: + LOGGER.info(f'\n{prefix} export failure: {e}') + return None, None + + +def export_pb(keras_model, im, file, prefix=colorstr('TensorFlow GraphDef:')): + # YOLOv5 TensorFlow GraphDef *.pb export https://github.com/leimao/Frozen_Graph_TensorFlow + try: + import tensorflow as tf + from tensorflow.python.framework.convert_to_constants import convert_variables_to_constants_v2 + + LOGGER.info(f'\n{prefix} starting export with tensorflow {tf.__version__}...') + f = file.with_suffix('.pb') + + m = tf.function(lambda x: keras_model(x)) # full model + m = m.get_concrete_function(tf.TensorSpec(keras_model.inputs[0].shape, keras_model.inputs[0].dtype)) + frozen_func = convert_variables_to_constants_v2(m) + frozen_func.graph.as_graph_def() + tf.io.write_graph(graph_or_graph_def=frozen_func.graph, logdir=str(f.parent), name=f.name, as_text=False) + + LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') + return f + except Exception as e: + LOGGER.info(f'\n{prefix} export failure: {e}') + + +def export_tflite(keras_model, im, file, int8, data, ncalib, prefix=colorstr('TensorFlow Lite:')): + # YOLOv5 TensorFlow Lite export + try: + import tensorflow as tf + + LOGGER.info(f'\n{prefix} starting export with tensorflow {tf.__version__}...') + batch_size, ch, *imgsz = list(im.shape) # BCHW + f = str(file).replace('.pt', '-fp16.tflite') + + converter = tf.lite.TFLiteConverter.from_keras_model(keras_model) + converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS] + converter.target_spec.supported_types = [tf.float16] + converter.optimizations = [tf.lite.Optimize.DEFAULT] + if int8: + from models.tf import representative_dataset_gen + dataset = LoadImages(check_dataset(data)['train'], img_size=imgsz, auto=False) # representative data + converter.representative_dataset = lambda: representative_dataset_gen(dataset, ncalib) + converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8] + converter.target_spec.supported_types = [] + converter.inference_input_type = tf.uint8 # or tf.int8 + converter.inference_output_type = tf.uint8 # or tf.int8 + converter.experimental_new_quantizer = False + f = str(file).replace('.pt', '-int8.tflite') + + tflite_model = converter.convert() + open(f, "wb").write(tflite_model) + LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') + return f + except Exception as e: + LOGGER.info(f'\n{prefix} export failure: {e}') + + +def export_edgetpu(keras_model, im, file, prefix=colorstr('Edge TPU:')): + # YOLOv5 Edge TPU export https://coral.ai/docs/edgetpu/models-intro/ + try: + cmd = 'edgetpu_compiler --version' + help_url = 'https://coral.ai/docs/edgetpu/compiler/' + assert platform.system() == 'Linux', f'export only supported on Linux. See {help_url}' + if subprocess.run(cmd + ' >/dev/null', shell=True).returncode != 0: + LOGGER.info(f'\n{prefix} export requires Edge TPU compiler. Attempting install from {help_url}') + sudo = subprocess.run('sudo --version >/dev/null', shell=True).returncode == 0 # sudo installed on system + for c in ['curl https://packages.cloud.google.com/apt/doc/apt-key.gpg | sudo apt-key add -', + 'echo "deb https://packages.cloud.google.com/apt coral-edgetpu-stable main" | sudo tee /etc/apt/sources.list.d/coral-edgetpu.list', + 'sudo apt-get update', + 'sudo apt-get install edgetpu-compiler']: + subprocess.run(c if sudo else c.replace('sudo ', ''), shell=True, check=True) + ver = subprocess.run(cmd, shell=True, capture_output=True, check=True).stdout.decode().split()[-1] + + LOGGER.info(f'\n{prefix} starting export with Edge TPU compiler {ver}...') + f = str(file).replace('.pt', '-int8_edgetpu.tflite') # Edge TPU model + f_tfl = str(file).replace('.pt', '-int8.tflite') # TFLite model + + cmd = f"edgetpu_compiler -s {f_tfl}" + subprocess.run(cmd, shell=True, check=True) + + LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') + return f + except Exception as e: + LOGGER.info(f'\n{prefix} export failure: {e}') + + +def export_tfjs(keras_model, im, file, prefix=colorstr('TensorFlow.js:')): + # YOLOv5 TensorFlow.js export + try: + check_requirements(('tensorflowjs',)) + import re + + import tensorflowjs as tfjs + + LOGGER.info(f'\n{prefix} starting export with tensorflowjs {tfjs.__version__}...') + f = str(file).replace('.pt', '_web_model') # js dir + f_pb = file.with_suffix('.pb') # *.pb path + f_json = f + '/model.json' # *.json path + + cmd = f'tensorflowjs_converter --input_format=tf_frozen_model ' \ + f'--output_node_names="Identity,Identity_1,Identity_2,Identity_3" {f_pb} {f}' + subprocess.run(cmd, shell=True) + + json = open(f_json).read() + with open(f_json, 'w') as j: # sort JSON Identity_* in ascending order + subst = re.sub( + r'{"outputs": {"Identity.?.?": {"name": "Identity.?.?"}, ' + r'"Identity.?.?": {"name": "Identity.?.?"}, ' + r'"Identity.?.?": {"name": "Identity.?.?"}, ' + r'"Identity.?.?": {"name": "Identity.?.?"}}}', + r'{"outputs": {"Identity": {"name": "Identity"}, ' + r'"Identity_1": {"name": "Identity_1"}, ' + r'"Identity_2": {"name": "Identity_2"}, ' + r'"Identity_3": {"name": "Identity_3"}}}', + json) + j.write(subst) + + LOGGER.info(f'{prefix} export success, saved as {f} ({file_size(f):.1f} MB)') + return f + except Exception as e: + LOGGER.info(f'\n{prefix} export failure: {e}') + + +@torch.no_grad() +def run(data=ROOT / 'data/coco128.yaml', # 'dataset.yaml path' + weights=ROOT / 'yolov5s.pt', # weights path + imgsz=(640, 640), # image (height, width) + batch_size=1, # batch size + device='cpu', # cuda device, i.e. 0 or 0,1,2,3 or cpu + include=('torchscript', 'onnx'), # include formats + half=False, # FP16 half-precision export + inplace=False, # set YOLOv5 Detect() inplace=True + train=False, # model.train() mode + optimize=False, # TorchScript: optimize for mobile + int8=False, # CoreML/TF INT8 quantization + dynamic=False, # ONNX/TF: dynamic axes + simplify=False, # ONNX: simplify model + opset=12, # ONNX: opset version + verbose=False, # TensorRT: verbose log + workspace=4, # TensorRT: workspace size (GB) + nms=False, # TF: add NMS to model + agnostic_nms=False, # TF: add agnostic NMS to model + topk_per_class=100, # TF.js NMS: topk per class to keep + topk_all=100, # TF.js NMS: topk for all classes to keep + iou_thres=0.45, # TF.js NMS: IoU threshold + conf_thres=0.25 # TF.js NMS: confidence threshold + ): + t = time.time() + include = [x.lower() for x in include] # to lowercase + formats = tuple(export_formats()['Argument'][1:]) # --include arguments + flags = [x in include for x in formats] + assert sum(flags) == len(include), f'ERROR: Invalid --include {include}, valid --include arguments are {formats}' + jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs = flags # export booleans + file = Path(url2file(weights) if str(weights).startswith(('http:/', 'https:/')) else weights) # PyTorch weights + + # Load PyTorch model + device = select_device(device) + assert not (device.type == 'cpu' and half), '--half only compatible with GPU export, i.e. use --device 0' + model = attempt_load(weights, map_location=device, inplace=True, fuse=True) # load FP32 model + nc, names = model.nc, model.names # number of classes, class names + + # Checks + imgsz *= 2 if len(imgsz) == 1 else 1 # expand + opset = 12 if ('openvino' in include) else opset # OpenVINO requires opset <= 12 + assert nc == len(names), f'Model class count {nc} != len(names) {len(names)}' + + # Input + gs = int(max(model.stride)) # grid size (max stride) + imgsz = [check_img_size(x, gs) for x in imgsz] # verify img_size are gs-multiples + im = torch.zeros(batch_size, 3, *imgsz).to(device) # image size(1,3,320,192) BCHW iDetection + + # Update model + if half: + im, model = im.half(), model.half() # to FP16 + model.train() if train else model.eval() # training mode = no Detect() layer grid construction + for k, m in model.named_modules(): + if isinstance(m, Conv): # assign export-friendly activations + if isinstance(m.act, nn.SiLU): + m.act = SiLU() + elif isinstance(m, Detect): + m.inplace = inplace + m.onnx_dynamic = dynamic + if hasattr(m, 'forward_export'): + m.forward = m.forward_export # assign custom forward (optional) + + for _ in range(2): + y = model(im) # dry runs + shape = tuple(y[0].shape) # model output shape + LOGGER.info(f"\n{colorstr('PyTorch:')} starting from {file} with output shape {shape} ({file_size(file):.1f} MB)") + + # Exports + f = [''] * 10 # exported filenames + warnings.filterwarnings(action='ignore', category=torch.jit.TracerWarning) # suppress TracerWarning + if jit: + f[0] = export_torchscript(model, im, file, optimize) + if engine: # TensorRT required before ONNX + f[1] = export_engine(model, im, file, train, half, simplify, workspace, verbose) + if onnx or xml: # OpenVINO requires ONNX + f[2] = export_onnx(model, im, file, opset, train, dynamic, simplify) + if xml: # OpenVINO + f[3] = export_openvino(model, im, file) + if coreml: + _, f[4] = export_coreml(model, im, file) + + # TensorFlow Exports + if any((saved_model, pb, tflite, edgetpu, tfjs)): + if int8 or edgetpu: # TFLite --int8 bug https://github.com/ultralytics/yolov5/issues/5707 + check_requirements(('flatbuffers==1.12',)) # required before `import tensorflow` + assert not (tflite and tfjs), 'TFLite and TF.js models must be exported separately, please pass only one type.' + model, f[5] = export_saved_model(model, im, file, dynamic, tf_nms=nms or agnostic_nms or tfjs, + agnostic_nms=agnostic_nms or tfjs, topk_per_class=topk_per_class, + topk_all=topk_all, conf_thres=conf_thres, iou_thres=iou_thres) # keras model + if pb or tfjs: # pb prerequisite to tfjs + f[6] = export_pb(model, im, file) + if tflite or edgetpu: + f[7] = export_tflite(model, im, file, int8=int8 or edgetpu, data=data, ncalib=100) + if edgetpu: + f[8] = export_edgetpu(model, im, file) + if tfjs: + f[9] = export_tfjs(model, im, file) + + # Finish + f = [str(x) for x in f if x] # filter out '' and None + if any(f): + LOGGER.info(f'\nExport complete ({time.time() - t:.2f}s)' + f"\nResults saved to {colorstr('bold', file.parent.resolve())}" + f"\nDetect: python detect.py --weights {f[-1]}" + f"\nPyTorch Hub: model = torch.hub.load('ultralytics/yolov5', 'custom', '{f[-1]}')" + f"\nValidate: python val.py --weights {f[-1]}" + f"\nVisualize: https://netron.app") + return f # return list of exported files/dirs + + +def parse_opt(): + parser = argparse.ArgumentParser() + #parser.add_argument('--data', type=str, default=ROOT / 'data/hand.yaml', help='dataset.yaml path') + + parser.add_argument('--data', type=str, default=ROOT / 'data/hand.yaml', help='dataset.yaml path') + parser.add_argument('--weights', nargs='+', type=str, default=ROOT / 'best.pt', help='model.pt path(s)') + #parser.add_argument('--weights', nargs='+', type=str, default=ROOT / 'weights/v5lite/best_5s.pt', help='model.pt path(s)') + parser.add_argument('--imgsz', '--img', '--img-size', nargs='+', type=int, default=[640, 640], help='image (h, w)') + parser.add_argument('--batch-size', type=int, default=1, help='batch size') + parser.add_argument('--device', default='cpu', help='cuda device, i.e. 0 or 0,1,2,3 or cpu') + parser.add_argument('--half', action='store_true', help='FP16 half-precision export') + parser.add_argument('--inplace', action='store_true', help='set YOLOv5 Detect() inplace=True') + parser.add_argument('--train', action='store_true', help='model.train() mode') + parser.add_argument('--optimize', action='store_true', help='TorchScript: optimize for mobile') + parser.add_argument('--int8', action='store_true', help='CoreML/TF INT8 quantization') + parser.add_argument('--dynamic', action='store_true', help='ONNX/TF: dynamic axes') + parser.add_argument('--simplify', action='store_true', help='ONNX: simplify model') + parser.add_argument('--opset', type=int, default=12, help='ONNX: opset version') + parser.add_argument('--verbose', action='store_true', help='TensorRT: verbose log') + parser.add_argument('--workspace', type=int, default=4, help='TensorRT: workspace size (GB)') + parser.add_argument('--nms', action='store_true', help='TF: add NMS to model') + parser.add_argument('--agnostic-nms', action='store_true', help='TF: add agnostic NMS to model') + parser.add_argument('--topk-per-class', type=int, default=100, help='TF.js NMS: topk per class to keep') + parser.add_argument('--topk-all', type=int, default=100, help='TF.js NMS: topk for all classes to keep') + parser.add_argument('--iou-thres', type=float, default=0.45, help='TF.js NMS: IoU threshold') + parser.add_argument('--conf-thres', type=float, default=0.25, help='TF.js NMS: confidence threshold') + parser.add_argument('--include', nargs='+', + default=['onnx','tflite'], + help='torchscript, onnx, openvino, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs') + opt = parser.parse_args() + print_args(FILE.stem, opt) + return opt + + +def main(opt): + for opt.weights in (opt.weights if isinstance(opt.weights, list) else [opt.weights]): + run(**vars(opt)) + + +if __name__ == "__main__": + opt = parse_opt() + main(opt) diff --git a/hubconf.py b/hubconf.py new file mode 100644 index 0000000..39fa614 --- /dev/null +++ b/hubconf.py @@ -0,0 +1,143 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +PyTorch Hub models https://pytorch.org/hub/ultralytics_yolov5/ + +Usage: + import torch + model = torch.hub.load('ultralytics/yolov5', 'yolov5s') + model = torch.hub.load('ultralytics/yolov5:master', 'custom', 'path/to/yolov5s.onnx') # file from branch +""" + +import torch + + +def _create(name, pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None): + """Creates or loads a YOLOv5 model + + Arguments: + name (str): model name 'yolov5s' or path 'path/to/best.pt' + pretrained (bool): load pretrained weights into the model + channels (int): number of input channels + classes (int): number of model classes + autoshape (bool): apply YOLOv5 .autoshape() wrapper to model + verbose (bool): print all information to screen + device (str, torch.device, None): device to use for model parameters + + Returns: + YOLOv5 model + """ + from pathlib import Path + + from models.common import AutoShape, DetectMultiBackend + from models.yolo import Model + from utils.downloads import attempt_download + from utils.general import LOGGER, check_requirements, intersect_dicts, logging + from utils.torch_utils import select_device + + if not verbose: + LOGGER.setLevel(logging.WARNING) + check_requirements(exclude=('tensorboard', 'thop', 'opencv-python')) + name = Path(name) + path = name.with_suffix('.pt') if name.suffix == '' else name # checkpoint path + try: + device = select_device(('0' if torch.cuda.is_available() else 'cpu') if device is None else device) + + if pretrained and channels == 3 and classes == 80: + model = DetectMultiBackend(path, device=device) # download/load FP32 model + # model = models.experimental.attempt_load(path, map_location=device) # download/load FP32 model + else: + cfg = list((Path(__file__).parent / 'models').rglob(f'{path.stem}.yaml'))[0] # model.yaml path + model = Model(cfg, channels, classes) # create model + if pretrained: + ckpt = torch.load(attempt_download(path), map_location=device) # load + csd = ckpt['model'].float().state_dict() # checkpoint state_dict as FP32 + csd = intersect_dicts(csd, model.state_dict(), exclude=['anchors']) # intersect + model.load_state_dict(csd, strict=False) # load + if len(ckpt['model'].names) == classes: + model.names = ckpt['model'].names # set class names attribute + if autoshape: + model = AutoShape(model) # for file/URI/PIL/cv2/np inputs and NMS + return model.to(device) + + except Exception as e: + help_url = 'https://github.com/ultralytics/yolov5/issues/36' + s = f'{e}. Cache may be out of date, try `force_reload=True` or see {help_url} for help.' + raise Exception(s) from e + + +def custom(path='path/to/model.pt', autoshape=True, verbose=True, device=None): + # YOLOv5 custom or local model + return _create(path, autoshape=autoshape, verbose=verbose, device=device) + + +def yolov5n(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None): + # YOLOv5-nano model https://github.com/ultralytics/yolov5 + return _create('yolov5n', pretrained, channels, classes, autoshape, verbose, device) + + +def yolov5s(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None): + # YOLOv5-small model https://github.com/ultralytics/yolov5 + return _create('yolov5s', pretrained, channels, classes, autoshape, verbose, device) + + +def yolov5m(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None): + # YOLOv5-medium model https://github.com/ultralytics/yolov5 + return _create('yolov5m', pretrained, channels, classes, autoshape, verbose, device) + + +def yolov5l(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None): + # YOLOv5-large model https://github.com/ultralytics/yolov5 + return _create('yolov5l', pretrained, channels, classes, autoshape, verbose, device) + + +def yolov5x(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None): + # YOLOv5-xlarge model https://github.com/ultralytics/yolov5 + return _create('yolov5x', pretrained, channels, classes, autoshape, verbose, device) + + +def yolov5n6(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None): + # YOLOv5-nano-P6 model https://github.com/ultralytics/yolov5 + return _create('yolov5n6', pretrained, channels, classes, autoshape, verbose, device) + + +def yolov5s6(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None): + # YOLOv5-small-P6 model https://github.com/ultralytics/yolov5 + return _create('yolov5s6', pretrained, channels, classes, autoshape, verbose, device) + + +def yolov5m6(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None): + # YOLOv5-medium-P6 model https://github.com/ultralytics/yolov5 + return _create('yolov5m6', pretrained, channels, classes, autoshape, verbose, device) + + +def yolov5l6(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None): + # YOLOv5-large-P6 model https://github.com/ultralytics/yolov5 + return _create('yolov5l6', pretrained, channels, classes, autoshape, verbose, device) + + +def yolov5x6(pretrained=True, channels=3, classes=80, autoshape=True, verbose=True, device=None): + # YOLOv5-xlarge-P6 model https://github.com/ultralytics/yolov5 + return _create('yolov5x6', pretrained, channels, classes, autoshape, verbose, device) + + +if __name__ == '__main__': + model = _create(name='yolov5s', pretrained=True, channels=3, classes=80, autoshape=True, verbose=True) # pretrained + # model = custom(path='path/to/model.pt') # custom + + # Verify inference + from pathlib import Path + + import cv2 + import numpy as np + from PIL import Image + + imgs = ['data/images/zidane.jpg', # filename + Path('data/images/zidane.jpg'), # Path + 'https://ultralytics.com/images/zidane.jpg', # URI + cv2.imread('data/images/bus.jpg')[:, :, ::-1], # OpenCV + Image.open('data/images/bus.jpg'), # PIL + np.zeros((320, 640, 3))] # numpy + + results = model(imgs, size=320) # batched inference + results.print() + results.save() diff --git a/models/__init__.py b/models/__init__.py new file mode 100644 index 0000000..e69de29 diff --git a/models/common.py b/models/common.py new file mode 100644 index 0000000..0dae024 --- /dev/null +++ b/models/common.py @@ -0,0 +1,677 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Common modules +""" + +import json +import math +import platform +import warnings +from collections import OrderedDict, namedtuple +from copy import copy +from pathlib import Path + +import cv2 +import numpy as np +import pandas as pd +import requests +import torch +import torch.nn as nn +import yaml +from PIL import Image +from torch.cuda import amp + +from utils.datasets import exif_transpose, letterbox +from utils.general import (LOGGER, check_requirements, check_suffix, check_version, colorstr, increment_path, + make_divisible, non_max_suppression, scale_coords, xywh2xyxy, xyxy2xywh) +from utils.plots import Annotator, colors, save_one_box +from utils.torch_utils import copy_attr, time_sync + + +def autopad(k, p=None): # kernel, padding + # Pad to 'same' + if p is None: + p = k // 2 if isinstance(k, int) else [x // 2 for x in k] # auto-pad + return p + + +class Conv(nn.Module): + # Standard convolution + def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True): # ch_in, ch_out, kernel, stride, padding, groups + super().__init__() + self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g, bias=False) + self.bn = nn.BatchNorm2d(c2) + self.act = nn.SiLU() if act is True else (act if isinstance(act, nn.Module) else nn.Identity()) + + def forward(self, x): + return self.act(self.bn(self.conv(x))) + + def forward_fuse(self, x): + return self.act(self.conv(x)) + + +class DWConv(Conv): + # Depth-wise convolution class + def __init__(self, c1, c2, k=1, s=1, act=True): # ch_in, ch_out, kernel, stride, padding, groups + super().__init__(c1, c2, k, s, g=math.gcd(c1, c2), act=act) + + +class TransformerLayer(nn.Module): + # Transformer layer https://arxiv.org/abs/2010.11929 (LayerNorm layers removed for better performance) + def __init__(self, c, num_heads): + super().__init__() + self.q = nn.Linear(c, c, bias=False) + self.k = nn.Linear(c, c, bias=False) + self.v = nn.Linear(c, c, bias=False) + self.ma = nn.MultiheadAttention(embed_dim=c, num_heads=num_heads) + self.fc1 = nn.Linear(c, c, bias=False) + self.fc2 = nn.Linear(c, c, bias=False) + + def forward(self, x): + x = self.ma(self.q(x), self.k(x), self.v(x))[0] + x + x = self.fc2(self.fc1(x)) + x + return x + + +class TransformerBlock(nn.Module): + # Vision Transformer https://arxiv.org/abs/2010.11929 + def __init__(self, c1, c2, num_heads, num_layers): + super().__init__() + self.conv = None + if c1 != c2: + self.conv = Conv(c1, c2) + self.linear = nn.Linear(c2, c2) # learnable position embedding + self.tr = nn.Sequential(*(TransformerLayer(c2, num_heads) for _ in range(num_layers))) + self.c2 = c2 + + def forward(self, x): + if self.conv is not None: + x = self.conv(x) + b, _, w, h = x.shape + p = x.flatten(2).permute(2, 0, 1) + return self.tr(p + self.linear(p)).permute(1, 2, 0).reshape(b, self.c2, w, h) + + +class Bottleneck(nn.Module): + # Standard bottleneck + def __init__(self, c1, c2, shortcut=True, g=1, e=0.5): # ch_in, ch_out, shortcut, groups, expansion + super().__init__() + c_ = int(c2 * e) # hidden channels + self.cv1 = Conv(c1, c_, 1, 1) + self.cv2 = Conv(c_, c2, 3, 1, g=g) + self.add = shortcut and c1 == c2 + + def forward(self, x): + return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x)) + + +class BottleneckCSP(nn.Module): + # CSP Bottleneck https://github.com/WongKinYiu/CrossStagePartialNetworks + def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5): # ch_in, ch_out, number, shortcut, groups, expansion + super().__init__() + c_ = int(c2 * e) # hidden channels + self.cv1 = Conv(c1, c_, 1, 1) + self.cv2 = nn.Conv2d(c1, c_, 1, 1, bias=False) + self.cv3 = nn.Conv2d(c_, c_, 1, 1, bias=False) + self.cv4 = Conv(2 * c_, c2, 1, 1) + self.bn = nn.BatchNorm2d(2 * c_) # applied to cat(cv2, cv3) + self.act = nn.SiLU() + self.m = nn.Sequential(*(Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n))) + + def forward(self, x): + y1 = self.cv3(self.m(self.cv1(x))) + y2 = self.cv2(x) + return self.cv4(self.act(self.bn(torch.cat((y1, y2), dim=1)))) + + +class C3(nn.Module): + # CSP Bottleneck with 3 convolutions + def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5): # ch_in, ch_out, number, shortcut, groups, expansion + super().__init__() + c_ = int(c2 * e) # hidden channels + self.cv1 = Conv(c1, c_, 1, 1) + self.cv2 = Conv(c1, c_, 1, 1) + self.cv3 = Conv(2 * c_, c2, 1) # act=FReLU(c2) + self.m = nn.Sequential(*(Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n))) + # self.m = nn.Sequential(*[CrossConv(c_, c_, 3, 1, g, 1.0, shortcut) for _ in range(n)]) + + def forward(self, x): + return self.cv3(torch.cat((self.m(self.cv1(x)), self.cv2(x)), dim=1)) + + +class C3TR(C3): + # C3 module with TransformerBlock() + def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5): + super().__init__(c1, c2, n, shortcut, g, e) + c_ = int(c2 * e) + self.m = TransformerBlock(c_, c_, 4, n) + + +class C3SPP(C3): + # C3 module with SPP() + def __init__(self, c1, c2, k=(5, 9, 13), n=1, shortcut=True, g=1, e=0.5): + super().__init__(c1, c2, n, shortcut, g, e) + c_ = int(c2 * e) + self.m = SPP(c_, c_, k) + + +class C3Ghost(C3): + # C3 module with GhostBottleneck() + def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5): + super().__init__(c1, c2, n, shortcut, g, e) + c_ = int(c2 * e) # hidden channels + self.m = nn.Sequential(*(GhostBottleneck(c_, c_) for _ in range(n))) + + +class SPP(nn.Module): + # Spatial Pyramid Pooling (SPP) layer https://arxiv.org/abs/1406.4729 + def __init__(self, c1, c2, k=(5, 9, 13)): + super().__init__() + c_ = c1 // 2 # hidden channels + self.cv1 = Conv(c1, c_, 1, 1) + self.cv2 = Conv(c_ * (len(k) + 1), c2, 1, 1) + self.m = nn.ModuleList([nn.MaxPool2d(kernel_size=x, stride=1, padding=x // 2) for x in k]) + + def forward(self, x): + x = self.cv1(x) + with warnings.catch_warnings(): + warnings.simplefilter('ignore') # suppress torch 1.9.0 max_pool2d() warning + return self.cv2(torch.cat([x] + [m(x) for m in self.m], 1)) + + +class SPPF(nn.Module): + # Spatial Pyramid Pooling - Fast (SPPF) layer for YOLOv5 by Glenn Jocher + def __init__(self, c1, c2, k=5): # equivalent to SPP(k=(5, 9, 13)) + super().__init__() + c_ = c1 // 2 # hidden channels + self.cv1 = Conv(c1, c_, 1, 1) + self.cv2 = Conv(c_ * 4, c2, 1, 1) + self.m = nn.MaxPool2d(kernel_size=k, stride=1, padding=k // 2) + + def forward(self, x): + x = self.cv1(x) + with warnings.catch_warnings(): + warnings.simplefilter('ignore') # suppress torch 1.9.0 max_pool2d() warning + y1 = self.m(x) + y2 = self.m(y1) + return self.cv2(torch.cat([x, y1, y2, self.m(y2)], 1)) + + +class Focus(nn.Module): + # Focus wh information into c-space + def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True): # ch_in, ch_out, kernel, stride, padding, groups + super().__init__() + self.conv = Conv(c1 * 4, c2, k, s, p, g, act) + # self.contract = Contract(gain=2) + + def forward(self, x): # x(b,c,w,h) -> y(b,4c,w/2,h/2) + return self.conv(torch.cat([x[..., ::2, ::2], x[..., 1::2, ::2], x[..., ::2, 1::2], x[..., 1::2, 1::2]], 1)) + # return self.conv(self.contract(x)) + + +class GhostConv(nn.Module): + # Ghost Convolution https://github.com/huawei-noah/ghostnet + def __init__(self, c1, c2, k=1, s=1, g=1, act=True): # ch_in, ch_out, kernel, stride, groups + super().__init__() + c_ = c2 // 2 # hidden channels + self.cv1 = Conv(c1, c_, k, s, None, g, act) + self.cv2 = Conv(c_, c_, 5, 1, None, c_, act) + + def forward(self, x): + y = self.cv1(x) + return torch.cat([y, self.cv2(y)], 1) + + +class GhostBottleneck(nn.Module): + # Ghost Bottleneck https://github.com/huawei-noah/ghostnet + def __init__(self, c1, c2, k=3, s=1): # ch_in, ch_out, kernel, stride + super().__init__() + c_ = c2 // 2 + self.conv = nn.Sequential(GhostConv(c1, c_, 1, 1), # pw + DWConv(c_, c_, k, s, act=False) if s == 2 else nn.Identity(), # dw + GhostConv(c_, c2, 1, 1, act=False)) # pw-linear + self.shortcut = nn.Sequential(DWConv(c1, c1, k, s, act=False), + Conv(c1, c2, 1, 1, act=False)) if s == 2 else nn.Identity() + + def forward(self, x): + return self.conv(x) + self.shortcut(x) + + +class Contract(nn.Module): + # Contract width-height into channels, i.e. x(1,64,80,80) to x(1,256,40,40) + def __init__(self, gain=2): + super().__init__() + self.gain = gain + + def forward(self, x): + b, c, h, w = x.size() # assert (h / s == 0) and (W / s == 0), 'Indivisible gain' + s = self.gain + x = x.view(b, c, h // s, s, w // s, s) # x(1,64,40,2,40,2) + x = x.permute(0, 3, 5, 1, 2, 4).contiguous() # x(1,2,2,64,40,40) + return x.view(b, c * s * s, h // s, w // s) # x(1,256,40,40) + + +class Expand(nn.Module): + # Expand channels into width-height, i.e. x(1,64,80,80) to x(1,16,160,160) + def __init__(self, gain=2): + super().__init__() + self.gain = gain + + def forward(self, x): + b, c, h, w = x.size() # assert C / s ** 2 == 0, 'Indivisible gain' + s = self.gain + x = x.view(b, s, s, c // s ** 2, h, w) # x(1,2,2,16,80,80) + x = x.permute(0, 3, 4, 1, 5, 2).contiguous() # x(1,16,80,2,80,2) + return x.view(b, c // s ** 2, h * s, w * s) # x(1,16,160,160) + + +class Concat(nn.Module): + # Concatenate a list of tensors along dimension + def __init__(self, dimension=1): + super().__init__() + self.d = dimension + + def forward(self, x): + return torch.cat(x, self.d) + + +class DetectMultiBackend(nn.Module): + # YOLOv5 MultiBackend class for python inference on various backends + def __init__(self, weights='yolov5s.pt', device=None, dnn=False, data=None): + # Usage: + # PyTorch: weights = *.pt + # TorchScript: *.torchscript + # ONNX Runtime: *.onnx + # ONNX OpenCV DNN: *.onnx with --dnn + # OpenVINO: *.xml + # CoreML: *.mlmodel + # TensorRT: *.engine + # TensorFlow SavedModel: *_saved_model + # TensorFlow GraphDef: *.pb + # TensorFlow Lite: *.tflite + # TensorFlow Edge TPU: *_edgetpu.tflite + from models.experimental import attempt_download, attempt_load # scoped to avoid circular import + + super().__init__() + w = str(weights[0] if isinstance(weights, list) else weights) + pt, jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs = self.model_type(w) # get backend + stride, names = 64, [f'class{i}' for i in range(1000)] # assign defaults + w = attempt_download(w) # download if not local + if data: # data.yaml path (optional) + with open(data, errors='ignore') as f: + names = yaml.safe_load(f)['names'] # class names + + if pt: # PyTorch + model = attempt_load(weights if isinstance(weights, list) else w, map_location=device) + stride = max(int(model.stride.max()), 32) # model stride + names = model.module.names if hasattr(model, 'module') else model.names # get class names + self.model = model # explicitly assign for to(), cpu(), cuda(), half() + elif jit: # TorchScript + LOGGER.info(f'Loading {w} for TorchScript inference...') + extra_files = {'config.txt': ''} # model metadata + model = torch.jit.load(w, _extra_files=extra_files) + if extra_files['config.txt']: + d = json.loads(extra_files['config.txt']) # extra_files dict + stride, names = int(d['stride']), d['names'] + elif dnn: # ONNX OpenCV DNN + LOGGER.info(f'Loading {w} for ONNX OpenCV DNN inference...') + check_requirements(('opencv-python>=4.5.4',)) + net = cv2.dnn.readNetFromONNX(w) + elif onnx: # ONNX Runtime + LOGGER.info(f'Loading {w} for ONNX Runtime inference...') + cuda = torch.cuda.is_available() + check_requirements(('onnx', 'onnxruntime-gpu' if cuda else 'onnxruntime')) + import onnxruntime + providers = ['CUDAExecutionProvider', 'CPUExecutionProvider'] if cuda else ['CPUExecutionProvider'] + session = onnxruntime.InferenceSession(w, providers=providers) + elif xml: # OpenVINO + LOGGER.info(f'Loading {w} for OpenVINO inference...') + check_requirements(('openvino-dev',)) # requires openvino-dev: https://pypi.org/project/openvino-dev/ + import openvino.inference_engine as ie + core = ie.IECore() + if not Path(w).is_file(): # if not *.xml + w = next(Path(w).glob('*.xml')) # get *.xml file from *_openvino_model dir + network = core.read_network(model=w, weights=Path(w).with_suffix('.bin')) # *.xml, *.bin paths + executable_network = core.load_network(network, device_name='CPU', num_requests=1) + elif engine: # TensorRT + LOGGER.info(f'Loading {w} for TensorRT inference...') + import tensorrt as trt # https://developer.nvidia.com/nvidia-tensorrt-download + check_version(trt.__version__, '7.0.0', hard=True) # require tensorrt>=7.0.0 + Binding = namedtuple('Binding', ('name', 'dtype', 'shape', 'data', 'ptr')) + logger = trt.Logger(trt.Logger.INFO) + with open(w, 'rb') as f, trt.Runtime(logger) as runtime: + model = runtime.deserialize_cuda_engine(f.read()) + bindings = OrderedDict() + for index in range(model.num_bindings): + name = model.get_binding_name(index) + dtype = trt.nptype(model.get_binding_dtype(index)) + shape = tuple(model.get_binding_shape(index)) + data = torch.from_numpy(np.empty(shape, dtype=np.dtype(dtype))).to(device) + bindings[name] = Binding(name, dtype, shape, data, int(data.data_ptr())) + binding_addrs = OrderedDict((n, d.ptr) for n, d in bindings.items()) + context = model.create_execution_context() + batch_size = bindings['images'].shape[0] + elif coreml: # CoreML + LOGGER.info(f'Loading {w} for CoreML inference...') + import coremltools as ct + model = ct.models.MLModel(w) + else: # TensorFlow (SavedModel, GraphDef, Lite, Edge TPU) + if saved_model: # SavedModel + LOGGER.info(f'Loading {w} for TensorFlow SavedModel inference...') + import tensorflow as tf + keras = False # assume TF1 saved_model + model = tf.keras.models.load_model(w) if keras else tf.saved_model.load(w) + elif pb: # GraphDef https://www.tensorflow.org/guide/migrate#a_graphpb_or_graphpbtxt + LOGGER.info(f'Loading {w} for TensorFlow GraphDef inference...') + import tensorflow as tf + + def wrap_frozen_graph(gd, inputs, outputs): + x = tf.compat.v1.wrap_function(lambda: tf.compat.v1.import_graph_def(gd, name=""), []) # wrapped + ge = x.graph.as_graph_element + return x.prune(tf.nest.map_structure(ge, inputs), tf.nest.map_structure(ge, outputs)) + + gd = tf.Graph().as_graph_def() # graph_def + gd.ParseFromString(open(w, 'rb').read()) + frozen_func = wrap_frozen_graph(gd, inputs="x:0", outputs="Identity:0") + elif tflite or edgetpu: # https://www.tensorflow.org/lite/guide/python#install_tensorflow_lite_for_python + try: # https://coral.ai/docs/edgetpu/tflite-python/#update-existing-tf-lite-code-for-the-edge-tpu + from tflite_runtime.interpreter import Interpreter, load_delegate + except ImportError: + import tensorflow as tf + Interpreter, load_delegate = tf.lite.Interpreter, tf.lite.experimental.load_delegate, + if edgetpu: # Edge TPU https://coral.ai/software/#edgetpu-runtime + LOGGER.info(f'Loading {w} for TensorFlow Lite Edge TPU inference...') + delegate = {'Linux': 'libedgetpu.so.1', + 'Darwin': 'libedgetpu.1.dylib', + 'Windows': 'edgetpu.dll'}[platform.system()] + interpreter = Interpreter(model_path=w, experimental_delegates=[load_delegate(delegate)]) + else: # Lite + LOGGER.info(f'Loading {w} for TensorFlow Lite inference...') + interpreter = Interpreter(model_path=w) # load TFLite model + interpreter.allocate_tensors() # allocate + input_details = interpreter.get_input_details() # inputs + output_details = interpreter.get_output_details() # outputs + elif tfjs: + raise Exception('ERROR: YOLOv5 TF.js inference is not supported') + self.__dict__.update(locals()) # assign all variables to self + + def forward(self, im, augment=False, visualize=False, val=False): + # YOLOv5 MultiBackend inference + b, ch, h, w = im.shape # batch, channel, height, width + if self.pt or self.jit: # PyTorch + y = self.model(im) if self.jit else self.model(im, augment=augment, visualize=visualize) + return y if val else y[0] + elif self.dnn: # ONNX OpenCV DNN + im = im.cpu().numpy() # torch to numpy + self.net.setInput(im) + y = self.net.forward() + elif self.onnx: # ONNX Runtime + im = im.cpu().numpy() # torch to numpy + y = self.session.run([self.session.get_outputs()[0].name], {self.session.get_inputs()[0].name: im})[0] + elif self.xml: # OpenVINO + im = im.cpu().numpy() # FP32 + desc = self.ie.TensorDesc(precision='FP32', dims=im.shape, layout='NCHW') # Tensor Description + request = self.executable_network.requests[0] # inference request + request.set_blob(blob_name='images', blob=self.ie.Blob(desc, im)) # name=next(iter(request.input_blobs)) + request.infer() + y = request.output_blobs['output'].buffer # name=next(iter(request.output_blobs)) + elif self.engine: # TensorRT + assert im.shape == self.bindings['images'].shape, (im.shape, self.bindings['images'].shape) + self.binding_addrs['images'] = int(im.data_ptr()) + self.context.execute_v2(list(self.binding_addrs.values())) + y = self.bindings['output'].data + elif self.coreml: # CoreML + im = im.permute(0, 2, 3, 1).cpu().numpy() # torch BCHW to numpy BHWC shape(1,320,192,3) + im = Image.fromarray((im[0] * 255).astype('uint8')) + # im = im.resize((192, 320), Image.ANTIALIAS) + y = self.model.predict({'image': im}) # coordinates are xywh normalized + if 'confidence' in y: + box = xywh2xyxy(y['coordinates'] * [[w, h, w, h]]) # xyxy pixels + conf, cls = y['confidence'].max(1), y['confidence'].argmax(1).astype(np.float) + y = np.concatenate((box, conf.reshape(-1, 1), cls.reshape(-1, 1)), 1) + else: + k = 'var_' + str(sorted(int(k.replace('var_', '')) for k in y)[-1]) # output key + y = y[k] # output + else: # TensorFlow (SavedModel, GraphDef, Lite, Edge TPU) + im = im.permute(0, 2, 3, 1).cpu().numpy() # torch BCHW to numpy BHWC shape(1,320,192,3) + if self.saved_model: # SavedModel + y = (self.model(im, training=False) if self.keras else self.model(im)[0]).numpy() + elif self.pb: # GraphDef + y = self.frozen_func(x=self.tf.constant(im)).numpy() + else: # Lite or Edge TPU + input, output = self.input_details[0], self.output_details[0] + int8 = input['dtype'] == np.uint8 # is TFLite quantized uint8 model + if int8: + scale, zero_point = input['quantization'] + im = (im / scale + zero_point).astype(np.uint8) # de-scale + self.interpreter.set_tensor(input['index'], im) + self.interpreter.invoke() + y = self.interpreter.get_tensor(output['index']) + if int8: + scale, zero_point = output['quantization'] + y = (y.astype(np.float32) - zero_point) * scale # re-scale + y[..., :4] *= [w, h, w, h] # xywh normalized to pixels + + y = torch.tensor(y) if isinstance(y, np.ndarray) else y + return (y, []) if val else y + + def warmup(self, imgsz=(1, 3, 640, 640), half=False): + # Warmup model by running inference once + if self.pt or self.jit or self.onnx or self.engine: # warmup types + if isinstance(self.device, torch.device) and self.device.type != 'cpu': # only warmup GPU models + im = torch.zeros(*imgsz).to(self.device).type(torch.half if half else torch.float) # input image + self.forward(im) # warmup + + @staticmethod + def model_type(p='path/to/model.pt'): + # Return model type from model path, i.e. path='path/to/model.onnx' -> type=onnx + from export import export_formats + suffixes = list(export_formats().Suffix) + ['.xml'] # export suffixes + check_suffix(p, suffixes) # checks + p = Path(p).name # eliminate trailing separators + pt, jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs, xml2 = (s in p for s in suffixes) + xml |= xml2 # *_openvino_model or *.xml + tflite &= not edgetpu # *.tflite + return pt, jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs + + +class AutoShape(nn.Module): + # YOLOv5 input-robust model wrapper for passing cv2/np/PIL/torch inputs. Includes preprocessing, inference and NMS + conf = 0.25 # NMS confidence threshold + iou = 0.45 # NMS IoU threshold + agnostic = False # NMS class-agnostic + multi_label = False # NMS multiple labels per box + classes = None # (optional list) filter by class, i.e. = [0, 15, 16] for COCO persons, cats and dogs + max_det = 1000 # maximum number of detections per image + amp = False # Automatic Mixed Precision (AMP) inference + + def __init__(self, model): + super().__init__() + LOGGER.info('Adding AutoShape... ') + copy_attr(self, model, include=('yaml', 'nc', 'hyp', 'names', 'stride', 'abc'), exclude=()) # copy attributes + self.dmb = isinstance(model, DetectMultiBackend) # DetectMultiBackend() instance + self.pt = not self.dmb or model.pt # PyTorch model + self.model = model.eval() + + def _apply(self, fn): + # Apply to(), cpu(), cuda(), half() to model tensors that are not parameters or registered buffers + self = super()._apply(fn) + if self.pt: + m = self.model.model.model[-1] if self.dmb else self.model.model[-1] # Detect() + m.stride = fn(m.stride) + m.grid = list(map(fn, m.grid)) + if isinstance(m.anchor_grid, list): + m.anchor_grid = list(map(fn, m.anchor_grid)) + return self + + @torch.no_grad() + def forward(self, imgs, size=640, augment=False, profile=False): + # Inference from various sources. For height=640, width=1280, RGB images example inputs are: + # file: imgs = 'data/images/zidane.jpg' # str or PosixPath + # URI: = 'https://ultralytics.com/images/zidane.jpg' + # OpenCV: = cv2.imread('image.jpg')[:,:,::-1] # HWC BGR to RGB x(640,1280,3) + # PIL: = Image.open('image.jpg') or ImageGrab.grab() # HWC x(640,1280,3) + # numpy: = np.zeros((640,1280,3)) # HWC + # torch: = torch.zeros(16,3,320,640) # BCHW (scaled to size=640, 0-1 values) + # multiple: = [Image.open('image1.jpg'), Image.open('image2.jpg'), ...] # list of images + + t = [time_sync()] + p = next(self.model.parameters()) if self.pt else torch.zeros(1) # for device and type + autocast = self.amp and (p.device.type != 'cpu') # Automatic Mixed Precision (AMP) inference + if isinstance(imgs, torch.Tensor): # torch + with amp.autocast(enabled=autocast): + return self.model(imgs.to(p.device).type_as(p), augment, profile) # inference + + # Pre-process + n, imgs = (len(imgs), imgs) if isinstance(imgs, list) else (1, [imgs]) # number of images, list of images + shape0, shape1, files = [], [], [] # image and inference shapes, filenames + for i, im in enumerate(imgs): + f = f'image{i}' # filename + if isinstance(im, (str, Path)): # filename or uri + im, f = Image.open(requests.get(im, stream=True).raw if str(im).startswith('http') else im), im + im = np.asarray(exif_transpose(im)) + elif isinstance(im, Image.Image): # PIL Image + im, f = np.asarray(exif_transpose(im)), getattr(im, 'filename', f) or f + files.append(Path(f).with_suffix('.jpg').name) + if im.shape[0] < 5: # image in CHW + im = im.transpose((1, 2, 0)) # reverse dataloader .transpose(2, 0, 1) + im = im[..., :3] if im.ndim == 3 else np.tile(im[..., None], 3) # enforce 3ch input + s = im.shape[:2] # HWC + shape0.append(s) # image shape + g = (size / max(s)) # gain + shape1.append([y * g for y in s]) + imgs[i] = im if im.data.contiguous else np.ascontiguousarray(im) # update + shape1 = [make_divisible(x, self.stride) for x in np.stack(shape1, 0).max(0)] # inference shape + x = [letterbox(im, new_shape=shape1 if self.pt else size, auto=False)[0] for im in imgs] # pad + x = np.stack(x, 0) if n > 1 else x[0][None] # stack + x = np.ascontiguousarray(x.transpose((0, 3, 1, 2))) # BHWC to BCHW + x = torch.from_numpy(x).to(p.device).type_as(p) / 255 # uint8 to fp16/32 + t.append(time_sync()) + + with amp.autocast(enabled=autocast): + # Inference + y = self.model(x, augment, profile) # forward + t.append(time_sync()) + + # Post-process + y = non_max_suppression(y if self.dmb else y[0], self.conf, iou_thres=self.iou, classes=self.classes, + agnostic=self.agnostic, multi_label=self.multi_label, max_det=self.max_det) # NMS + for i in range(n): + scale_coords(shape1, y[i][:, :4], shape0[i]) + + t.append(time_sync()) + return Detections(imgs, y, files, t, self.names, x.shape) + + +class Detections: + # YOLOv5 detections class for inference results + def __init__(self, imgs, pred, files, times=(0, 0, 0, 0), names=None, shape=None): + super().__init__() + d = pred[0].device # device + gn = [torch.tensor([*(im.shape[i] for i in [1, 0, 1, 0]), 1, 1], device=d) for im in imgs] # normalizations + self.imgs = imgs # list of images as numpy arrays + self.pred = pred # list of tensors pred[0] = (xyxy, conf, cls) + self.names = names # class names + self.files = files # image filenames + self.times = times # profiling times + self.xyxy = pred # xyxy pixels + self.xywh = [xyxy2xywh(x) for x in pred] # xywh pixels + self.xyxyn = [x / g for x, g in zip(self.xyxy, gn)] # xyxy normalized + self.xywhn = [x / g for x, g in zip(self.xywh, gn)] # xywh normalized + self.n = len(self.pred) # number of images (batch size) + self.t = tuple((times[i + 1] - times[i]) * 1000 / self.n for i in range(3)) # timestamps (ms) + self.s = shape # inference BCHW shape + + def display(self, pprint=False, show=False, save=False, crop=False, render=False, save_dir=Path('')): + crops = [] + for i, (im, pred) in enumerate(zip(self.imgs, self.pred)): + s = f'image {i + 1}/{len(self.pred)}: {im.shape[0]}x{im.shape[1]} ' # string + if pred.shape[0]: + for c in pred[:, -1].unique(): + n = (pred[:, -1] == c).sum() # detections per class + s += f"{n} {self.names[int(c)]}{'s' * (n > 1)}, " # add to string + if show or save or render or crop: + annotator = Annotator(im, example=str(self.names)) + for *box, conf, cls in reversed(pred): # xyxy, confidence, class + label = f'{self.names[int(cls)]} {conf:.2f}' + if crop: + file = save_dir / 'crops' / self.names[int(cls)] / self.files[i] if save else None + crops.append({'box': box, 'conf': conf, 'cls': cls, 'label': label, + 'im': save_one_box(box, im, file=file, save=save)}) + else: # all others + annotator.box_label(box, label, color=colors(cls)) + im = annotator.im + else: + s += '(no detections)' + + im = Image.fromarray(im.astype(np.uint8)) if isinstance(im, np.ndarray) else im # from np + if pprint: + LOGGER.info(s.rstrip(', ')) + if show: + im.show(self.files[i]) # show + if save: + f = self.files[i] + im.save(save_dir / f) # save + if i == self.n - 1: + LOGGER.info(f"Saved {self.n} image{'s' * (self.n > 1)} to {colorstr('bold', save_dir)}") + if render: + self.imgs[i] = np.asarray(im) + if crop: + if save: + LOGGER.info(f'Saved results to {save_dir}\n') + return crops + + def print(self): + self.display(pprint=True) # print results + LOGGER.info(f'Speed: %.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape {tuple(self.s)}' % + self.t) + + def show(self): + self.display(show=True) # show results + + def save(self, save_dir='runs/detect/exp'): + save_dir = increment_path(save_dir, exist_ok=save_dir != 'runs/detect/exp', mkdir=True) # increment save_dir + self.display(save=True, save_dir=save_dir) # save results + + def crop(self, save=True, save_dir='runs/detect/exp'): + save_dir = increment_path(save_dir, exist_ok=save_dir != 'runs/detect/exp', mkdir=True) if save else None + return self.display(crop=True, save=save, save_dir=save_dir) # crop results + + def render(self): + self.display(render=True) # render results + return self.imgs + + def pandas(self): + # return detections as pandas DataFrames, i.e. print(results.pandas().xyxy[0]) + new = copy(self) # return copy + ca = 'xmin', 'ymin', 'xmax', 'ymax', 'confidence', 'class', 'name' # xyxy columns + cb = 'xcenter', 'ycenter', 'width', 'height', 'confidence', 'class', 'name' # xywh columns + for k, c in zip(['xyxy', 'xyxyn', 'xywh', 'xywhn'], [ca, ca, cb, cb]): + a = [[x[:5] + [int(x[5]), self.names[int(x[5])]] for x in x.tolist()] for x in getattr(self, k)] # update + setattr(new, k, [pd.DataFrame(x, columns=c) for x in a]) + return new + + def tolist(self): + # return a list of Detections objects, i.e. 'for result in results.tolist():' + r = range(self.n) # iterable + x = [Detections([self.imgs[i]], [self.pred[i]], [self.files[i]], self.times, self.names, self.s) for i in r] + # for d in x: + # for k in ['imgs', 'pred', 'xyxy', 'xyxyn', 'xywh', 'xywhn']: + # setattr(d, k, getattr(d, k)[0]) # pop out of list + return x + + def __len__(self): + return self.n + + +class Classify(nn.Module): + # Classification head, i.e. x(b,c1,20,20) to x(b,c2) + def __init__(self, c1, c2, k=1, s=1, p=None, g=1): # ch_in, ch_out, kernel, stride, padding, groups + super().__init__() + self.aap = nn.AdaptiveAvgPool2d(1) # to x(b,c1,1,1) + self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g) # to x(b,c2,1,1) + self.flat = nn.Flatten() + + def forward(self, x): + z = torch.cat([self.aap(y) for y in (x if isinstance(x, list) else [x])], 1) # cat if list + return self.flat(self.conv(z)) # flatten to x(b,c2) diff --git a/models/experimental.py b/models/experimental.py new file mode 100644 index 0000000..463e551 --- /dev/null +++ b/models/experimental.py @@ -0,0 +1,120 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Experimental modules +""" +import math + +import numpy as np +import torch +import torch.nn as nn + +from models.common import Conv +from utils.downloads import attempt_download + + +class CrossConv(nn.Module): + # Cross Convolution Downsample + def __init__(self, c1, c2, k=3, s=1, g=1, e=1.0, shortcut=False): + # ch_in, ch_out, kernel, stride, groups, expansion, shortcut + super().__init__() + c_ = int(c2 * e) # hidden channels + self.cv1 = Conv(c1, c_, (1, k), (1, s)) + self.cv2 = Conv(c_, c2, (k, 1), (s, 1), g=g) + self.add = shortcut and c1 == c2 + + def forward(self, x): + return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x)) + + +class Sum(nn.Module): + # Weighted sum of 2 or more layers https://arxiv.org/abs/1911.09070 + def __init__(self, n, weight=False): # n: number of inputs + super().__init__() + self.weight = weight # apply weights boolean + self.iter = range(n - 1) # iter object + if weight: + self.w = nn.Parameter(-torch.arange(1.0, n) / 2, requires_grad=True) # layer weights + + def forward(self, x): + y = x[0] # no weight + if self.weight: + w = torch.sigmoid(self.w) * 2 + for i in self.iter: + y = y + x[i + 1] * w[i] + else: + for i in self.iter: + y = y + x[i + 1] + return y + + +class MixConv2d(nn.Module): + # Mixed Depth-wise Conv https://arxiv.org/abs/1907.09595 + def __init__(self, c1, c2, k=(1, 3), s=1, equal_ch=True): # ch_in, ch_out, kernel, stride, ch_strategy + super().__init__() + n = len(k) # number of convolutions + if equal_ch: # equal c_ per group + i = torch.linspace(0, n - 1E-6, c2).floor() # c2 indices + c_ = [(i == g).sum() for g in range(n)] # intermediate channels + else: # equal weight.numel() per group + b = [c2] + [0] * n + a = np.eye(n + 1, n, k=-1) + a -= np.roll(a, 1, axis=1) + a *= np.array(k) ** 2 + a[0] = 1 + c_ = np.linalg.lstsq(a, b, rcond=None)[0].round() # solve for equal weight indices, ax = b + + self.m = nn.ModuleList( + [nn.Conv2d(c1, int(c_), k, s, k // 2, groups=math.gcd(c1, int(c_)), bias=False) for k, c_ in zip(k, c_)]) + self.bn = nn.BatchNorm2d(c2) + self.act = nn.SiLU() + + def forward(self, x): + return self.act(self.bn(torch.cat([m(x) for m in self.m], 1))) + + +class Ensemble(nn.ModuleList): + # Ensemble of models + def __init__(self): + super().__init__() + + def forward(self, x, augment=False, profile=False, visualize=False): + y = [] + for module in self: + y.append(module(x, augment, profile, visualize)[0]) + # y = torch.stack(y).max(0)[0] # max ensemble + # y = torch.stack(y).mean(0) # mean ensemble + y = torch.cat(y, 1) # nms ensemble + return y, None # inference, train output + + +def attempt_load(weights, map_location=None, inplace=True, fuse=True): + from models.yolo import Detect, Model + + # Loads an ensemble of models weights=[a,b,c] or a single model weights=[a] or weights=a + model = Ensemble() + for w in weights if isinstance(weights, list) else [weights]: + ckpt = torch.load(attempt_download(w), map_location=map_location) # load + if fuse: + model.append(ckpt['ema' if ckpt.get('ema') else 'model'].float().fuse().eval()) # FP32 model + else: + model.append(ckpt['ema' if ckpt.get('ema') else 'model'].float().eval()) # without layer fuse + + # Compatibility updates + for m in model.modules(): + if type(m) in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6, nn.SiLU, Detect, Model]: + m.inplace = inplace # pytorch 1.7.0 compatibility + if type(m) is Detect: + if not isinstance(m.anchor_grid, list): # new Detect Layer compatibility + delattr(m, 'anchor_grid') + setattr(m, 'anchor_grid', [torch.zeros(1)] * m.nl) + elif type(m) is Conv: + m._non_persistent_buffers_set = set() # pytorch 1.6.0 compatibility + + if len(model) == 1: + return model[-1] # return model + else: + print(f'Ensemble created with {weights}\n') + for k in ['names']: + setattr(model, k, getattr(model[-1], k)) + model.stride = model[torch.argmax(torch.tensor([m.stride.max() for m in model])).int()].stride # max stride + return model # return ensemble diff --git a/models/hub/anchors.yaml b/models/hub/anchors.yaml new file mode 100644 index 0000000..e4d7beb --- /dev/null +++ b/models/hub/anchors.yaml @@ -0,0 +1,59 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +# Default anchors for COCO data + + +# P5 ------------------------------------------------------------------------------------------------------------------- +# P5-640: +anchors_p5_640: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + + +# P6 ------------------------------------------------------------------------------------------------------------------- +# P6-640: thr=0.25: 0.9964 BPR, 5.54 anchors past thr, n=12, img_size=640, metric_all=0.281/0.716-mean/best, past_thr=0.469-mean: 9,11, 21,19, 17,41, 43,32, 39,70, 86,64, 65,131, 134,130, 120,265, 282,180, 247,354, 512,387 +anchors_p6_640: + - [9,11, 21,19, 17,41] # P3/8 + - [43,32, 39,70, 86,64] # P4/16 + - [65,131, 134,130, 120,265] # P5/32 + - [282,180, 247,354, 512,387] # P6/64 + +# P6-1280: thr=0.25: 0.9950 BPR, 5.55 anchors past thr, n=12, img_size=1280, metric_all=0.281/0.714-mean/best, past_thr=0.468-mean: 19,27, 44,40, 38,94, 96,68, 86,152, 180,137, 140,301, 303,264, 238,542, 436,615, 739,380, 925,792 +anchors_p6_1280: + - [19,27, 44,40, 38,94] # P3/8 + - [96,68, 86,152, 180,137] # P4/16 + - [140,301, 303,264, 238,542] # P5/32 + - [436,615, 739,380, 925,792] # P6/64 + +# P6-1920: thr=0.25: 0.9950 BPR, 5.55 anchors past thr, n=12, img_size=1920, metric_all=0.281/0.714-mean/best, past_thr=0.468-mean: 28,41, 67,59, 57,141, 144,103, 129,227, 270,205, 209,452, 455,396, 358,812, 653,922, 1109,570, 1387,1187 +anchors_p6_1920: + - [28,41, 67,59, 57,141] # P3/8 + - [144,103, 129,227, 270,205] # P4/16 + - [209,452, 455,396, 358,812] # P5/32 + - [653,922, 1109,570, 1387,1187] # P6/64 + + +# P7 ------------------------------------------------------------------------------------------------------------------- +# P7-640: thr=0.25: 0.9962 BPR, 6.76 anchors past thr, n=15, img_size=640, metric_all=0.275/0.733-mean/best, past_thr=0.466-mean: 11,11, 13,30, 29,20, 30,46, 61,38, 39,92, 78,80, 146,66, 79,163, 149,150, 321,143, 157,303, 257,402, 359,290, 524,372 +anchors_p7_640: + - [11,11, 13,30, 29,20] # P3/8 + - [30,46, 61,38, 39,92] # P4/16 + - [78,80, 146,66, 79,163] # P5/32 + - [149,150, 321,143, 157,303] # P6/64 + - [257,402, 359,290, 524,372] # P7/128 + +# P7-1280: thr=0.25: 0.9968 BPR, 6.71 anchors past thr, n=15, img_size=1280, metric_all=0.273/0.732-mean/best, past_thr=0.463-mean: 19,22, 54,36, 32,77, 70,83, 138,71, 75,173, 165,159, 148,334, 375,151, 334,317, 251,626, 499,474, 750,326, 534,814, 1079,818 +anchors_p7_1280: + - [19,22, 54,36, 32,77] # P3/8 + - [70,83, 138,71, 75,173] # P4/16 + - [165,159, 148,334, 375,151] # P5/32 + - [334,317, 251,626, 499,474] # P6/64 + - [750,326, 534,814, 1079,818] # P7/128 + +# P7-1920: thr=0.25: 0.9968 BPR, 6.71 anchors past thr, n=15, img_size=1920, metric_all=0.273/0.732-mean/best, past_thr=0.463-mean: 29,34, 81,55, 47,115, 105,124, 207,107, 113,259, 247,238, 222,500, 563,227, 501,476, 376,939, 749,711, 1126,489, 801,1222, 1618,1227 +anchors_p7_1920: + - [29,34, 81,55, 47,115] # P3/8 + - [105,124, 207,107, 113,259] # P4/16 + - [247,238, 222,500, 563,227] # P5/32 + - [501,476, 376,939, 749,711] # P6/64 + - [1126,489, 801,1222, 1618,1227] # P7/128 diff --git a/models/hub/yolov3-spp.yaml b/models/hub/yolov3-spp.yaml new file mode 100644 index 0000000..c669821 --- /dev/null +++ b/models/hub/yolov3-spp.yaml @@ -0,0 +1,51 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.0 # model depth multiple +width_multiple: 1.0 # layer channel multiple +anchors: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + +# darknet53 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [32, 3, 1]], # 0 + [-1, 1, Conv, [64, 3, 2]], # 1-P1/2 + [-1, 1, Bottleneck, [64]], + [-1, 1, Conv, [128, 3, 2]], # 3-P2/4 + [-1, 2, Bottleneck, [128]], + [-1, 1, Conv, [256, 3, 2]], # 5-P3/8 + [-1, 8, Bottleneck, [256]], + [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 + [-1, 8, Bottleneck, [512]], + [-1, 1, Conv, [1024, 3, 2]], # 9-P5/32 + [-1, 4, Bottleneck, [1024]], # 10 + ] + +# YOLOv3-SPP head +head: + [[-1, 1, Bottleneck, [1024, False]], + [-1, 1, SPP, [512, [5, 9, 13]]], + [-1, 1, Conv, [1024, 3, 1]], + [-1, 1, Conv, [512, 1, 1]], + [-1, 1, Conv, [1024, 3, 1]], # 15 (P5/32-large) + + [-2, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 8], 1, Concat, [1]], # cat backbone P4 + [-1, 1, Bottleneck, [512, False]], + [-1, 1, Bottleneck, [512, False]], + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, Conv, [512, 3, 1]], # 22 (P4/16-medium) + + [-2, 1, Conv, [128, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P3 + [-1, 1, Bottleneck, [256, False]], + [-1, 2, Bottleneck, [256, False]], # 27 (P3/8-small) + + [[27, 22, 15], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) + ] diff --git a/models/hub/yolov3-tiny.yaml b/models/hub/yolov3-tiny.yaml new file mode 100644 index 0000000..b28b443 --- /dev/null +++ b/models/hub/yolov3-tiny.yaml @@ -0,0 +1,41 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.0 # model depth multiple +width_multiple: 1.0 # layer channel multiple +anchors: + - [10,14, 23,27, 37,58] # P4/16 + - [81,82, 135,169, 344,319] # P5/32 + +# YOLOv3-tiny backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [16, 3, 1]], # 0 + [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 1-P1/2 + [-1, 1, Conv, [32, 3, 1]], + [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 3-P2/4 + [-1, 1, Conv, [64, 3, 1]], + [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 5-P3/8 + [-1, 1, Conv, [128, 3, 1]], + [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 7-P4/16 + [-1, 1, Conv, [256, 3, 1]], + [-1, 1, nn.MaxPool2d, [2, 2, 0]], # 9-P5/32 + [-1, 1, Conv, [512, 3, 1]], + [-1, 1, nn.ZeroPad2d, [[0, 1, 0, 1]]], # 11 + [-1, 1, nn.MaxPool2d, [2, 1, 0]], # 12 + ] + +# YOLOv3-tiny head +head: + [[-1, 1, Conv, [1024, 3, 1]], + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, Conv, [512, 3, 1]], # 15 (P5/32-large) + + [-2, 1, Conv, [128, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 8], 1, Concat, [1]], # cat backbone P4 + [-1, 1, Conv, [256, 3, 1]], # 19 (P4/16-medium) + + [[19, 15], 1, Detect, [nc, anchors]], # Detect(P4, P5) + ] diff --git a/models/hub/yolov3.yaml b/models/hub/yolov3.yaml new file mode 100644 index 0000000..d1ef912 --- /dev/null +++ b/models/hub/yolov3.yaml @@ -0,0 +1,51 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.0 # model depth multiple +width_multiple: 1.0 # layer channel multiple +anchors: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + +# darknet53 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [32, 3, 1]], # 0 + [-1, 1, Conv, [64, 3, 2]], # 1-P1/2 + [-1, 1, Bottleneck, [64]], + [-1, 1, Conv, [128, 3, 2]], # 3-P2/4 + [-1, 2, Bottleneck, [128]], + [-1, 1, Conv, [256, 3, 2]], # 5-P3/8 + [-1, 8, Bottleneck, [256]], + [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 + [-1, 8, Bottleneck, [512]], + [-1, 1, Conv, [1024, 3, 2]], # 9-P5/32 + [-1, 4, Bottleneck, [1024]], # 10 + ] + +# YOLOv3 head +head: + [[-1, 1, Bottleneck, [1024, False]], + [-1, 1, Conv, [512, 1, 1]], + [-1, 1, Conv, [1024, 3, 1]], + [-1, 1, Conv, [512, 1, 1]], + [-1, 1, Conv, [1024, 3, 1]], # 15 (P5/32-large) + + [-2, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 8], 1, Concat, [1]], # cat backbone P4 + [-1, 1, Bottleneck, [512, False]], + [-1, 1, Bottleneck, [512, False]], + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, Conv, [512, 3, 1]], # 22 (P4/16-medium) + + [-2, 1, Conv, [128, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P3 + [-1, 1, Bottleneck, [256, False]], + [-1, 2, Bottleneck, [256, False]], # 27 (P3/8-small) + + [[27, 22, 15], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) + ] diff --git a/models/hub/yolov5-bifpn.yaml b/models/hub/yolov5-bifpn.yaml new file mode 100644 index 0000000..504815f --- /dev/null +++ b/models/hub/yolov5-bifpn.yaml @@ -0,0 +1,48 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.0 # model depth multiple +width_multiple: 1.0 # layer channel multiple +anchors: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 9 + ] + +# YOLOv5 v6.0 BiFPN head +head: + [[-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 13 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 17 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 14, 6], 1, Concat, [1]], # cat P4 <--- BiFPN change + [-1, 3, C3, [512, False]], # 20 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 10], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [1024, False]], # 23 (P5/32-large) + + [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) + ] diff --git a/models/hub/yolov5-fpn.yaml b/models/hub/yolov5-fpn.yaml new file mode 100644 index 0000000..a23e9c6 --- /dev/null +++ b/models/hub/yolov5-fpn.yaml @@ -0,0 +1,42 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.0 # model depth multiple +width_multiple: 1.0 # layer channel multiple +anchors: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 9 + ] + +# YOLOv5 v6.0 FPN head +head: + [[-1, 3, C3, [1024, False]], # 10 (P5/32-large) + + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 1, Conv, [512, 1, 1]], + [-1, 3, C3, [512, False]], # 14 (P4/16-medium) + + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 1, Conv, [256, 1, 1]], + [-1, 3, C3, [256, False]], # 18 (P3/8-small) + + [[18, 14, 10], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) + ] diff --git a/models/hub/yolov5-p2.yaml b/models/hub/yolov5-p2.yaml new file mode 100644 index 0000000..554117d --- /dev/null +++ b/models/hub/yolov5-p2.yaml @@ -0,0 +1,54 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.0 # model depth multiple +width_multiple: 1.0 # layer channel multiple +anchors: 3 # AutoAnchor evolves 3 anchors per P output layer + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 9 + ] + +# YOLOv5 v6.0 head with (P2, P3, P4, P5) outputs +head: + [[-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 13 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 17 (P3/8-small) + + [-1, 1, Conv, [128, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 2], 1, Concat, [1]], # cat backbone P2 + [-1, 1, C3, [128, False]], # 21 (P2/4-xsmall) + + [-1, 1, Conv, [128, 3, 2]], + [[-1, 18], 1, Concat, [1]], # cat head P3 + [-1, 3, C3, [256, False]], # 24 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 14], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 27 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 10], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [1024, False]], # 30 (P5/32-large) + + [[21, 24, 27, 30], 1, Detect, [nc, anchors]], # Detect(P2, P3, P4, P5) + ] diff --git a/models/hub/yolov5-p34.yaml b/models/hub/yolov5-p34.yaml new file mode 100644 index 0000000..dbf0f85 --- /dev/null +++ b/models/hub/yolov5-p34.yaml @@ -0,0 +1,41 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 0.33 # model depth multiple +width_multiple: 0.50 # layer channel multiple +anchors: 3 # AutoAnchor evolves 3 anchors per P output layer + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [ [ -1, 1, Conv, [ 64, 6, 2, 2 ] ], # 0-P1/2 + [ -1, 1, Conv, [ 128, 3, 2 ] ], # 1-P2/4 + [ -1, 3, C3, [ 128 ] ], + [ -1, 1, Conv, [ 256, 3, 2 ] ], # 3-P3/8 + [ -1, 6, C3, [ 256 ] ], + [ -1, 1, Conv, [ 512, 3, 2 ] ], # 5-P4/16 + [ -1, 9, C3, [ 512 ] ], + [ -1, 1, Conv, [ 1024, 3, 2 ] ], # 7-P5/32 + [ -1, 3, C3, [ 1024 ] ], + [ -1, 1, SPPF, [ 1024, 5 ] ], # 9 + ] + +# YOLOv5 v6.0 head with (P3, P4) outputs +head: + [ [ -1, 1, Conv, [ 512, 1, 1 ] ], + [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], + [ [ -1, 6 ], 1, Concat, [ 1 ] ], # cat backbone P4 + [ -1, 3, C3, [ 512, False ] ], # 13 + + [ -1, 1, Conv, [ 256, 1, 1 ] ], + [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ], + [ [ -1, 4 ], 1, Concat, [ 1 ] ], # cat backbone P3 + [ -1, 3, C3, [ 256, False ] ], # 17 (P3/8-small) + + [ -1, 1, Conv, [ 256, 3, 2 ] ], + [ [ -1, 14 ], 1, Concat, [ 1 ] ], # cat head P4 + [ -1, 3, C3, [ 512, False ] ], # 20 (P4/16-medium) + + [ [ 17, 20 ], 1, Detect, [ nc, anchors ] ], # Detect(P3, P4) + ] diff --git a/models/hub/yolov5-p6.yaml b/models/hub/yolov5-p6.yaml new file mode 100644 index 0000000..a17202f --- /dev/null +++ b/models/hub/yolov5-p6.yaml @@ -0,0 +1,56 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.0 # model depth multiple +width_multiple: 1.0 # layer channel multiple +anchors: 3 # AutoAnchor evolves 3 anchors per P output layer + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [768, 3, 2]], # 7-P5/32 + [-1, 3, C3, [768]], + [-1, 1, Conv, [1024, 3, 2]], # 9-P6/64 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 11 + ] + +# YOLOv5 v6.0 head with (P3, P4, P5, P6) outputs +head: + [[-1, 1, Conv, [768, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 8], 1, Concat, [1]], # cat backbone P5 + [-1, 3, C3, [768, False]], # 15 + + [-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 19 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 23 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 20], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 26 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 16], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [768, False]], # 29 (P5/32-large) + + [-1, 1, Conv, [768, 3, 2]], + [[-1, 12], 1, Concat, [1]], # cat head P6 + [-1, 3, C3, [1024, False]], # 32 (P6/64-xlarge) + + [[23, 26, 29, 32], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5, P6) + ] diff --git a/models/hub/yolov5-p7.yaml b/models/hub/yolov5-p7.yaml new file mode 100644 index 0000000..edd7d13 --- /dev/null +++ b/models/hub/yolov5-p7.yaml @@ -0,0 +1,67 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.0 # model depth multiple +width_multiple: 1.0 # layer channel multiple +anchors: 3 # AutoAnchor evolves 3 anchors per P output layer + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [768, 3, 2]], # 7-P5/32 + [-1, 3, C3, [768]], + [-1, 1, Conv, [1024, 3, 2]], # 9-P6/64 + [-1, 3, C3, [1024]], + [-1, 1, Conv, [1280, 3, 2]], # 11-P7/128 + [-1, 3, C3, [1280]], + [-1, 1, SPPF, [1280, 5]], # 13 + ] + +# YOLOv5 v6.0 head with (P3, P4, P5, P6, P7) outputs +head: + [[-1, 1, Conv, [1024, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 10], 1, Concat, [1]], # cat backbone P6 + [-1, 3, C3, [1024, False]], # 17 + + [-1, 1, Conv, [768, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 8], 1, Concat, [1]], # cat backbone P5 + [-1, 3, C3, [768, False]], # 21 + + [-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 25 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 29 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 26], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 32 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 22], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [768, False]], # 35 (P5/32-large) + + [-1, 1, Conv, [768, 3, 2]], + [[-1, 18], 1, Concat, [1]], # cat head P6 + [-1, 3, C3, [1024, False]], # 38 (P6/64-xlarge) + + [-1, 1, Conv, [1024, 3, 2]], + [[-1, 14], 1, Concat, [1]], # cat head P7 + [-1, 3, C3, [1280, False]], # 41 (P7/128-xxlarge) + + [[29, 32, 35, 38, 41], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5, P6, P7) + ] diff --git a/models/hub/yolov5-panet.yaml b/models/hub/yolov5-panet.yaml new file mode 100644 index 0000000..ccfbf90 --- /dev/null +++ b/models/hub/yolov5-panet.yaml @@ -0,0 +1,48 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.0 # model depth multiple +width_multiple: 1.0 # layer channel multiple +anchors: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 9 + ] + +# YOLOv5 v6.0 PANet head +head: + [[-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 13 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 17 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 14], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 20 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 10], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [1024, False]], # 23 (P5/32-large) + + [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) + ] diff --git a/models/hub/yolov5l6.yaml b/models/hub/yolov5l6.yaml new file mode 100644 index 0000000..632c2cb --- /dev/null +++ b/models/hub/yolov5l6.yaml @@ -0,0 +1,60 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.0 # model depth multiple +width_multiple: 1.0 # layer channel multiple +anchors: + - [19,27, 44,40, 38,94] # P3/8 + - [96,68, 86,152, 180,137] # P4/16 + - [140,301, 303,264, 238,542] # P5/32 + - [436,615, 739,380, 925,792] # P6/64 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [768, 3, 2]], # 7-P5/32 + [-1, 3, C3, [768]], + [-1, 1, Conv, [1024, 3, 2]], # 9-P6/64 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 11 + ] + +# YOLOv5 v6.0 head +head: + [[-1, 1, Conv, [768, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 8], 1, Concat, [1]], # cat backbone P5 + [-1, 3, C3, [768, False]], # 15 + + [-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 19 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 23 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 20], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 26 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 16], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [768, False]], # 29 (P5/32-large) + + [-1, 1, Conv, [768, 3, 2]], + [[-1, 12], 1, Concat, [1]], # cat head P6 + [-1, 3, C3, [1024, False]], # 32 (P6/64-xlarge) + + [[23, 26, 29, 32], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5, P6) + ] diff --git a/models/hub/yolov5m6.yaml b/models/hub/yolov5m6.yaml new file mode 100644 index 0000000..ecc53fd --- /dev/null +++ b/models/hub/yolov5m6.yaml @@ -0,0 +1,60 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 0.67 # model depth multiple +width_multiple: 0.75 # layer channel multiple +anchors: + - [19,27, 44,40, 38,94] # P3/8 + - [96,68, 86,152, 180,137] # P4/16 + - [140,301, 303,264, 238,542] # P5/32 + - [436,615, 739,380, 925,792] # P6/64 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [768, 3, 2]], # 7-P5/32 + [-1, 3, C3, [768]], + [-1, 1, Conv, [1024, 3, 2]], # 9-P6/64 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 11 + ] + +# YOLOv5 v6.0 head +head: + [[-1, 1, Conv, [768, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 8], 1, Concat, [1]], # cat backbone P5 + [-1, 3, C3, [768, False]], # 15 + + [-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 19 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 23 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 20], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 26 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 16], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [768, False]], # 29 (P5/32-large) + + [-1, 1, Conv, [768, 3, 2]], + [[-1, 12], 1, Concat, [1]], # cat head P6 + [-1, 3, C3, [1024, False]], # 32 (P6/64-xlarge) + + [[23, 26, 29, 32], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5, P6) + ] diff --git a/models/hub/yolov5n6.yaml b/models/hub/yolov5n6.yaml new file mode 100644 index 0000000..0c0c71d --- /dev/null +++ b/models/hub/yolov5n6.yaml @@ -0,0 +1,60 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 0.33 # model depth multiple +width_multiple: 0.25 # layer channel multiple +anchors: + - [19,27, 44,40, 38,94] # P3/8 + - [96,68, 86,152, 180,137] # P4/16 + - [140,301, 303,264, 238,542] # P5/32 + - [436,615, 739,380, 925,792] # P6/64 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [768, 3, 2]], # 7-P5/32 + [-1, 3, C3, [768]], + [-1, 1, Conv, [1024, 3, 2]], # 9-P6/64 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 11 + ] + +# YOLOv5 v6.0 head +head: + [[-1, 1, Conv, [768, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 8], 1, Concat, [1]], # cat backbone P5 + [-1, 3, C3, [768, False]], # 15 + + [-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 19 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 23 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 20], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 26 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 16], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [768, False]], # 29 (P5/32-large) + + [-1, 1, Conv, [768, 3, 2]], + [[-1, 12], 1, Concat, [1]], # cat head P6 + [-1, 3, C3, [1024, False]], # 32 (P6/64-xlarge) + + [[23, 26, 29, 32], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5, P6) + ] diff --git a/models/hub/yolov5s-ghost.yaml b/models/hub/yolov5s-ghost.yaml new file mode 100644 index 0000000..ff9519c --- /dev/null +++ b/models/hub/yolov5s-ghost.yaml @@ -0,0 +1,48 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 0.33 # model depth multiple +width_multiple: 0.50 # layer channel multiple +anchors: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, GhostConv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3Ghost, [128]], + [-1, 1, GhostConv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3Ghost, [256]], + [-1, 1, GhostConv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3Ghost, [512]], + [-1, 1, GhostConv, [1024, 3, 2]], # 7-P5/32 + [-1, 3, C3Ghost, [1024]], + [-1, 1, SPPF, [1024, 5]], # 9 + ] + +# YOLOv5 v6.0 head +head: + [[-1, 1, GhostConv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3Ghost, [512, False]], # 13 + + [-1, 1, GhostConv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3Ghost, [256, False]], # 17 (P3/8-small) + + [-1, 1, GhostConv, [256, 3, 2]], + [[-1, 14], 1, Concat, [1]], # cat head P4 + [-1, 3, C3Ghost, [512, False]], # 20 (P4/16-medium) + + [-1, 1, GhostConv, [512, 3, 2]], + [[-1, 10], 1, Concat, [1]], # cat head P5 + [-1, 3, C3Ghost, [1024, False]], # 23 (P5/32-large) + + [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) + ] diff --git a/models/hub/yolov5s-transformer.yaml b/models/hub/yolov5s-transformer.yaml new file mode 100644 index 0000000..100d7c4 --- /dev/null +++ b/models/hub/yolov5s-transformer.yaml @@ -0,0 +1,48 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 0.33 # model depth multiple +width_multiple: 0.50 # layer channel multiple +anchors: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 + [-1, 3, C3TR, [1024]], # 9 <--- C3TR() Transformer module + [-1, 1, SPPF, [1024, 5]], # 9 + ] + +# YOLOv5 v6.0 head +head: + [[-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 13 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 17 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 14], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 20 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 10], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [1024, False]], # 23 (P5/32-large) + + [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) + ] diff --git a/models/hub/yolov5s6.yaml b/models/hub/yolov5s6.yaml new file mode 100644 index 0000000..a28fb55 --- /dev/null +++ b/models/hub/yolov5s6.yaml @@ -0,0 +1,60 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 0.33 # model depth multiple +width_multiple: 0.50 # layer channel multiple +anchors: + - [19,27, 44,40, 38,94] # P3/8 + - [96,68, 86,152, 180,137] # P4/16 + - [140,301, 303,264, 238,542] # P5/32 + - [436,615, 739,380, 925,792] # P6/64 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [768, 3, 2]], # 7-P5/32 + [-1, 3, C3, [768]], + [-1, 1, Conv, [1024, 3, 2]], # 9-P6/64 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 11 + ] + +# YOLOv5 v6.0 head +head: + [[-1, 1, Conv, [768, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 8], 1, Concat, [1]], # cat backbone P5 + [-1, 3, C3, [768, False]], # 15 + + [-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 19 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 23 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 20], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 26 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 16], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [768, False]], # 29 (P5/32-large) + + [-1, 1, Conv, [768, 3, 2]], + [[-1, 12], 1, Concat, [1]], # cat head P6 + [-1, 3, C3, [1024, False]], # 32 (P6/64-xlarge) + + [[23, 26, 29, 32], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5, P6) + ] diff --git a/models/hub/yolov5x6.yaml b/models/hub/yolov5x6.yaml new file mode 100644 index 0000000..ba795c4 --- /dev/null +++ b/models/hub/yolov5x6.yaml @@ -0,0 +1,60 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.33 # model depth multiple +width_multiple: 1.25 # layer channel multiple +anchors: + - [19,27, 44,40, 38,94] # P3/8 + - [96,68, 86,152, 180,137] # P4/16 + - [140,301, 303,264, 238,542] # P5/32 + - [436,615, 739,380, 925,792] # P6/64 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [768, 3, 2]], # 7-P5/32 + [-1, 3, C3, [768]], + [-1, 1, Conv, [1024, 3, 2]], # 9-P6/64 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 11 + ] + +# YOLOv5 v6.0 head +head: + [[-1, 1, Conv, [768, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 8], 1, Concat, [1]], # cat backbone P5 + [-1, 3, C3, [768, False]], # 15 + + [-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 19 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 23 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 20], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 26 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 16], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [768, False]], # 29 (P5/32-large) + + [-1, 1, Conv, [768, 3, 2]], + [[-1, 12], 1, Concat, [1]], # cat head P6 + [-1, 3, C3, [1024, False]], # 32 (P6/64-xlarge) + + [[23, 26, 29, 32], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5, P6) + ] diff --git a/models/tf.py b/models/tf.py new file mode 100644 index 0000000..45c3547 --- /dev/null +++ b/models/tf.py @@ -0,0 +1,464 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +TensorFlow, Keras and TFLite versions of YOLOv5 +Authored by https://github.com/zldrobit in PR https://github.com/ultralytics/yolov5/pull/1127 + +Usage: + $ python models/tf.py --weights yolov5s.pt + +Export: + $ python path/to/export.py --weights yolov5s.pt --include saved_model pb tflite tfjs +""" + +import argparse +import sys +from copy import deepcopy +from pathlib import Path + +FILE = Path(__file__).resolve() +ROOT = FILE.parents[1] # YOLOv5 root directory +if str(ROOT) not in sys.path: + sys.path.append(str(ROOT)) # add ROOT to PATH +# ROOT = ROOT.relative_to(Path.cwd()) # relative + +import numpy as np +import tensorflow as tf +import torch +import torch.nn as nn +from tensorflow import keras + +from models.common import C3, SPP, SPPF, Bottleneck, BottleneckCSP, Concat, Conv, DWConv, Focus, autopad +from models.experimental import CrossConv, MixConv2d, attempt_load +from models.yolo import Detect +from utils.activations import SiLU +from utils.general import LOGGER, make_divisible, print_args + + +class TFBN(keras.layers.Layer): + # TensorFlow BatchNormalization wrapper + def __init__(self, w=None): + super().__init__() + self.bn = keras.layers.BatchNormalization( + beta_initializer=keras.initializers.Constant(w.bias.numpy()), + gamma_initializer=keras.initializers.Constant(w.weight.numpy()), + moving_mean_initializer=keras.initializers.Constant(w.running_mean.numpy()), + moving_variance_initializer=keras.initializers.Constant(w.running_var.numpy()), + epsilon=w.eps) + + def call(self, inputs): + return self.bn(inputs) + + +class TFPad(keras.layers.Layer): + def __init__(self, pad): + super().__init__() + self.pad = tf.constant([[0, 0], [pad, pad], [pad, pad], [0, 0]]) + + def call(self, inputs): + return tf.pad(inputs, self.pad, mode='constant', constant_values=0) + + +class TFConv(keras.layers.Layer): + # Standard convolution + def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True, w=None): + # ch_in, ch_out, weights, kernel, stride, padding, groups + super().__init__() + assert g == 1, "TF v2.2 Conv2D does not support 'groups' argument" + assert isinstance(k, int), "Convolution with multiple kernels are not allowed." + # TensorFlow convolution padding is inconsistent with PyTorch (e.g. k=3 s=2 'SAME' padding) + # see https://stackoverflow.com/questions/52975843/comparing-conv2d-with-padding-between-tensorflow-and-pytorch + + conv = keras.layers.Conv2D( + c2, k, s, 'SAME' if s == 1 else 'VALID', use_bias=False if hasattr(w, 'bn') else True, + kernel_initializer=keras.initializers.Constant(w.conv.weight.permute(2, 3, 1, 0).numpy()), + bias_initializer='zeros' if hasattr(w, 'bn') else keras.initializers.Constant(w.conv.bias.numpy())) + self.conv = conv if s == 1 else keras.Sequential([TFPad(autopad(k, p)), conv]) + self.bn = TFBN(w.bn) if hasattr(w, 'bn') else tf.identity + + # YOLOv5 activations + if isinstance(w.act, nn.LeakyReLU): + self.act = (lambda x: keras.activations.relu(x, alpha=0.1)) if act else tf.identity + elif isinstance(w.act, nn.Hardswish): + self.act = (lambda x: x * tf.nn.relu6(x + 3) * 0.166666667) if act else tf.identity + elif isinstance(w.act, (nn.SiLU, SiLU)): + self.act = (lambda x: keras.activations.swish(x)) if act else tf.identity + else: + raise Exception(f'no matching TensorFlow activation found for {w.act}') + + def call(self, inputs): + return self.act(self.bn(self.conv(inputs))) + + +class TFFocus(keras.layers.Layer): + # Focus wh information into c-space + def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True, w=None): + # ch_in, ch_out, kernel, stride, padding, groups + super().__init__() + self.conv = TFConv(c1 * 4, c2, k, s, p, g, act, w.conv) + + def call(self, inputs): # x(b,w,h,c) -> y(b,w/2,h/2,4c) + # inputs = inputs / 255 # normalize 0-255 to 0-1 + return self.conv(tf.concat([inputs[:, ::2, ::2, :], + inputs[:, 1::2, ::2, :], + inputs[:, ::2, 1::2, :], + inputs[:, 1::2, 1::2, :]], 3)) + + +class TFBottleneck(keras.layers.Layer): + # Standard bottleneck + def __init__(self, c1, c2, shortcut=True, g=1, e=0.5, w=None): # ch_in, ch_out, shortcut, groups, expansion + super().__init__() + c_ = int(c2 * e) # hidden channels + self.cv1 = TFConv(c1, c_, 1, 1, w=w.cv1) + self.cv2 = TFConv(c_, c2, 3, 1, g=g, w=w.cv2) + self.add = shortcut and c1 == c2 + + def call(self, inputs): + return inputs + self.cv2(self.cv1(inputs)) if self.add else self.cv2(self.cv1(inputs)) + + +class TFConv2d(keras.layers.Layer): + # Substitution for PyTorch nn.Conv2D + def __init__(self, c1, c2, k, s=1, g=1, bias=True, w=None): + super().__init__() + assert g == 1, "TF v2.2 Conv2D does not support 'groups' argument" + self.conv = keras.layers.Conv2D( + c2, k, s, 'VALID', use_bias=bias, + kernel_initializer=keras.initializers.Constant(w.weight.permute(2, 3, 1, 0).numpy()), + bias_initializer=keras.initializers.Constant(w.bias.numpy()) if bias else None, ) + + def call(self, inputs): + return self.conv(inputs) + + +class TFBottleneckCSP(keras.layers.Layer): + # CSP Bottleneck https://github.com/WongKinYiu/CrossStagePartialNetworks + def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5, w=None): + # ch_in, ch_out, number, shortcut, groups, expansion + super().__init__() + c_ = int(c2 * e) # hidden channels + self.cv1 = TFConv(c1, c_, 1, 1, w=w.cv1) + self.cv2 = TFConv2d(c1, c_, 1, 1, bias=False, w=w.cv2) + self.cv3 = TFConv2d(c_, c_, 1, 1, bias=False, w=w.cv3) + self.cv4 = TFConv(2 * c_, c2, 1, 1, w=w.cv4) + self.bn = TFBN(w.bn) + self.act = lambda x: keras.activations.relu(x, alpha=0.1) + self.m = keras.Sequential([TFBottleneck(c_, c_, shortcut, g, e=1.0, w=w.m[j]) for j in range(n)]) + + def call(self, inputs): + y1 = self.cv3(self.m(self.cv1(inputs))) + y2 = self.cv2(inputs) + return self.cv4(self.act(self.bn(tf.concat((y1, y2), axis=3)))) + + +class TFC3(keras.layers.Layer): + # CSP Bottleneck with 3 convolutions + def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5, w=None): + # ch_in, ch_out, number, shortcut, groups, expansion + super().__init__() + c_ = int(c2 * e) # hidden channels + self.cv1 = TFConv(c1, c_, 1, 1, w=w.cv1) + self.cv2 = TFConv(c1, c_, 1, 1, w=w.cv2) + self.cv3 = TFConv(2 * c_, c2, 1, 1, w=w.cv3) + self.m = keras.Sequential([TFBottleneck(c_, c_, shortcut, g, e=1.0, w=w.m[j]) for j in range(n)]) + + def call(self, inputs): + return self.cv3(tf.concat((self.m(self.cv1(inputs)), self.cv2(inputs)), axis=3)) + + +class TFSPP(keras.layers.Layer): + # Spatial pyramid pooling layer used in YOLOv3-SPP + def __init__(self, c1, c2, k=(5, 9, 13), w=None): + super().__init__() + c_ = c1 // 2 # hidden channels + self.cv1 = TFConv(c1, c_, 1, 1, w=w.cv1) + self.cv2 = TFConv(c_ * (len(k) + 1), c2, 1, 1, w=w.cv2) + self.m = [keras.layers.MaxPool2D(pool_size=x, strides=1, padding='SAME') for x in k] + + def call(self, inputs): + x = self.cv1(inputs) + return self.cv2(tf.concat([x] + [m(x) for m in self.m], 3)) + + +class TFSPPF(keras.layers.Layer): + # Spatial pyramid pooling-Fast layer + def __init__(self, c1, c2, k=5, w=None): + super().__init__() + c_ = c1 // 2 # hidden channels + self.cv1 = TFConv(c1, c_, 1, 1, w=w.cv1) + self.cv2 = TFConv(c_ * 4, c2, 1, 1, w=w.cv2) + self.m = keras.layers.MaxPool2D(pool_size=k, strides=1, padding='SAME') + + def call(self, inputs): + x = self.cv1(inputs) + y1 = self.m(x) + y2 = self.m(y1) + return self.cv2(tf.concat([x, y1, y2, self.m(y2)], 3)) + + +class TFDetect(keras.layers.Layer): + def __init__(self, nc=80, anchors=(), ch=(), imgsz=(640, 640), w=None): # detection layer + super().__init__() + self.stride = tf.convert_to_tensor(w.stride.numpy(), dtype=tf.float32) + self.nc = nc # number of classes + self.no = nc + 5 # number of outputs per anchor + self.nl = len(anchors) # number of detection layers + self.na = len(anchors[0]) // 2 # number of anchors + self.grid = [tf.zeros(1)] * self.nl # init grid + self.anchors = tf.convert_to_tensor(w.anchors.numpy(), dtype=tf.float32) + self.anchor_grid = tf.reshape(self.anchors * tf.reshape(self.stride, [self.nl, 1, 1]), + [self.nl, 1, -1, 1, 2]) + self.m = [TFConv2d(x, self.no * self.na, 1, w=w.m[i]) for i, x in enumerate(ch)] + self.training = False # set to False after building model + self.imgsz = imgsz + for i in range(self.nl): + ny, nx = self.imgsz[0] // self.stride[i], self.imgsz[1] // self.stride[i] + self.grid[i] = self._make_grid(nx, ny) + + def call(self, inputs): + z = [] # inference output + x = [] + for i in range(self.nl): + x.append(self.m[i](inputs[i])) + # x(bs,20,20,255) to x(bs,3,20,20,85) + ny, nx = self.imgsz[0] // self.stride[i], self.imgsz[1] // self.stride[i] + x[i] = tf.transpose(tf.reshape(x[i], [-1, ny * nx, self.na, self.no]), [0, 2, 1, 3]) + + if not self.training: # inference + y = tf.sigmoid(x[i]) + xy = (y[..., 0:2] * 2 - 0.5 + self.grid[i]) * self.stride[i] # xy + wh = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i] + # Normalize xywh to 0-1 to reduce calibration error + xy /= tf.constant([[self.imgsz[1], self.imgsz[0]]], dtype=tf.float32) + wh /= tf.constant([[self.imgsz[1], self.imgsz[0]]], dtype=tf.float32) + y = tf.concat([xy, wh, y[..., 4:]], -1) + z.append(tf.reshape(y, [-1, self.na * ny * nx, self.no])) + + return x if self.training else (tf.concat(z, 1), x) + + @staticmethod + def _make_grid(nx=20, ny=20): + # yv, xv = torch.meshgrid([torch.arange(ny), torch.arange(nx)]) + # return torch.stack((xv, yv), 2).view((1, 1, ny, nx, 2)).float() + xv, yv = tf.meshgrid(tf.range(nx), tf.range(ny)) + return tf.cast(tf.reshape(tf.stack([xv, yv], 2), [1, 1, ny * nx, 2]), dtype=tf.float32) + + +class TFUpsample(keras.layers.Layer): + def __init__(self, size, scale_factor, mode, w=None): # warning: all arguments needed including 'w' + super().__init__() + assert scale_factor == 2, "scale_factor must be 2" + self.upsample = lambda x: tf.image.resize(x, (x.shape[1] * 2, x.shape[2] * 2), method=mode) + # self.upsample = keras.layers.UpSampling2D(size=scale_factor, interpolation=mode) + # with default arguments: align_corners=False, half_pixel_centers=False + # self.upsample = lambda x: tf.raw_ops.ResizeNearestNeighbor(images=x, + # size=(x.shape[1] * 2, x.shape[2] * 2)) + + def call(self, inputs): + return self.upsample(inputs) + + +class TFConcat(keras.layers.Layer): + def __init__(self, dimension=1, w=None): + super().__init__() + assert dimension == 1, "convert only NCHW to NHWC concat" + self.d = 3 + + def call(self, inputs): + return tf.concat(inputs, self.d) + + +def parse_model(d, ch, model, imgsz): # model_dict, input_channels(3) + LOGGER.info(f"\n{'':>3}{'from':>18}{'n':>3}{'params':>10} {'module':<40}{'arguments':<30}") + anchors, nc, gd, gw = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple'] + na = (len(anchors[0]) // 2) if isinstance(anchors, list) else anchors # number of anchors + no = na * (nc + 5) # number of outputs = anchors * (classes + 5) + + layers, save, c2 = [], [], ch[-1] # layers, savelist, ch out + for i, (f, n, m, args) in enumerate(d['backbone'] + d['head']): # from, number, module, args + m_str = m + m = eval(m) if isinstance(m, str) else m # eval strings + for j, a in enumerate(args): + try: + args[j] = eval(a) if isinstance(a, str) else a # eval strings + except NameError: + pass + + n = max(round(n * gd), 1) if n > 1 else n # depth gain + if m in [nn.Conv2d, Conv, Bottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3]: + c1, c2 = ch[f], args[0] + c2 = make_divisible(c2 * gw, 8) if c2 != no else c2 + + args = [c1, c2, *args[1:]] + if m in [BottleneckCSP, C3]: + args.insert(2, n) + n = 1 + elif m is nn.BatchNorm2d: + args = [ch[f]] + elif m is Concat: + c2 = sum(ch[-1 if x == -1 else x + 1] for x in f) + elif m is Detect: + args.append([ch[x + 1] for x in f]) + if isinstance(args[1], int): # number of anchors + args[1] = [list(range(args[1] * 2))] * len(f) + args.append(imgsz) + else: + c2 = ch[f] + + tf_m = eval('TF' + m_str.replace('nn.', '')) + m_ = keras.Sequential([tf_m(*args, w=model.model[i][j]) for j in range(n)]) if n > 1 \ + else tf_m(*args, w=model.model[i]) # module + + torch_m_ = nn.Sequential(*(m(*args) for _ in range(n))) if n > 1 else m(*args) # module + t = str(m)[8:-2].replace('__main__.', '') # module type + np = sum(x.numel() for x in torch_m_.parameters()) # number params + m_.i, m_.f, m_.type, m_.np = i, f, t, np # attach index, 'from' index, type, number params + LOGGER.info(f'{i:>3}{str(f):>18}{str(n):>3}{np:>10} {t:<40}{str(args):<30}') # print + save.extend(x % i for x in ([f] if isinstance(f, int) else f) if x != -1) # append to savelist + layers.append(m_) + ch.append(c2) + return keras.Sequential(layers), sorted(save) + + +class TFModel: + def __init__(self, cfg='yolov5s.yaml', ch=3, nc=None, model=None, imgsz=(640, 640)): # model, channels, classes + super().__init__() + if isinstance(cfg, dict): + self.yaml = cfg # model dict + else: # is *.yaml + import yaml # for torch hub + self.yaml_file = Path(cfg).name + with open(cfg) as f: + self.yaml = yaml.load(f, Loader=yaml.FullLoader) # model dict + + # Define model + if nc and nc != self.yaml['nc']: + LOGGER.info(f"Overriding {cfg} nc={self.yaml['nc']} with nc={nc}") + self.yaml['nc'] = nc # override yaml value + self.model, self.savelist = parse_model(deepcopy(self.yaml), ch=[ch], model=model, imgsz=imgsz) + + def predict(self, inputs, tf_nms=False, agnostic_nms=False, topk_per_class=100, topk_all=100, iou_thres=0.45, + conf_thres=0.25): + y = [] # outputs + x = inputs + for i, m in enumerate(self.model.layers): + if m.f != -1: # if not from previous layer + x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f] # from earlier layers + + x = m(x) # run + y.append(x if m.i in self.savelist else None) # save output + + # Add TensorFlow NMS + if tf_nms: + boxes = self._xywh2xyxy(x[0][..., :4]) + probs = x[0][:, :, 4:5] + classes = x[0][:, :, 5:] + scores = probs * classes + if agnostic_nms: + nms = AgnosticNMS()((boxes, classes, scores), topk_all, iou_thres, conf_thres) + return nms, x[1] + else: + boxes = tf.expand_dims(boxes, 2) + nms = tf.image.combined_non_max_suppression( + boxes, scores, topk_per_class, topk_all, iou_thres, conf_thres, clip_boxes=False) + return nms, x[1] + + return x[0] # output only first tensor [1,6300,85] = [xywh, conf, class0, class1, ...] + # x = x[0][0] # [x(1,6300,85), ...] to x(6300,85) + # xywh = x[..., :4] # x(6300,4) boxes + # conf = x[..., 4:5] # x(6300,1) confidences + # cls = tf.reshape(tf.cast(tf.argmax(x[..., 5:], axis=1), tf.float32), (-1, 1)) # x(6300,1) classes + # return tf.concat([conf, cls, xywh], 1) + + @staticmethod + def _xywh2xyxy(xywh): + # Convert nx4 boxes from [x, y, w, h] to [x1, y1, x2, y2] where xy1=top-left, xy2=bottom-right + x, y, w, h = tf.split(xywh, num_or_size_splits=4, axis=-1) + return tf.concat([x - w / 2, y - h / 2, x + w / 2, y + h / 2], axis=-1) + + +class AgnosticNMS(keras.layers.Layer): + # TF Agnostic NMS + def call(self, input, topk_all, iou_thres, conf_thres): + # wrap map_fn to avoid TypeSpec related error https://stackoverflow.com/a/65809989/3036450 + return tf.map_fn(lambda x: self._nms(x, topk_all, iou_thres, conf_thres), input, + fn_output_signature=(tf.float32, tf.float32, tf.float32, tf.int32), + name='agnostic_nms') + + @staticmethod + def _nms(x, topk_all=100, iou_thres=0.45, conf_thres=0.25): # agnostic NMS + boxes, classes, scores = x + class_inds = tf.cast(tf.argmax(classes, axis=-1), tf.float32) + scores_inp = tf.reduce_max(scores, -1) + selected_inds = tf.image.non_max_suppression( + boxes, scores_inp, max_output_size=topk_all, iou_threshold=iou_thres, score_threshold=conf_thres) + selected_boxes = tf.gather(boxes, selected_inds) + padded_boxes = tf.pad(selected_boxes, + paddings=[[0, topk_all - tf.shape(selected_boxes)[0]], [0, 0]], + mode="CONSTANT", constant_values=0.0) + selected_scores = tf.gather(scores_inp, selected_inds) + padded_scores = tf.pad(selected_scores, + paddings=[[0, topk_all - tf.shape(selected_boxes)[0]]], + mode="CONSTANT", constant_values=-1.0) + selected_classes = tf.gather(class_inds, selected_inds) + padded_classes = tf.pad(selected_classes, + paddings=[[0, topk_all - tf.shape(selected_boxes)[0]]], + mode="CONSTANT", constant_values=-1.0) + valid_detections = tf.shape(selected_inds)[0] + return padded_boxes, padded_scores, padded_classes, valid_detections + + +def representative_dataset_gen(dataset, ncalib=100): + # Representative dataset generator for use with converter.representative_dataset, returns a generator of np arrays + for n, (path, img, im0s, vid_cap, string) in enumerate(dataset): + input = np.transpose(img, [1, 2, 0]) + input = np.expand_dims(input, axis=0).astype(np.float32) + input /= 255 + yield [input] + if n >= ncalib: + break + + +def run(weights=ROOT / 'yolov5s.pt', # weights path + imgsz=(640, 640), # inference size h,w + batch_size=1, # batch size + dynamic=False, # dynamic batch size + ): + # PyTorch model + im = torch.zeros((batch_size, 3, *imgsz)) # BCHW image + model = attempt_load(weights, map_location=torch.device('cpu'), inplace=True, fuse=False) + _ = model(im) # inference + model.info() + + # TensorFlow model + im = tf.zeros((batch_size, *imgsz, 3)) # BHWC image + tf_model = TFModel(cfg=model.yaml, model=model, nc=model.nc, imgsz=imgsz) + _ = tf_model.predict(im) # inference + + # Keras model + im = keras.Input(shape=(*imgsz, 3), batch_size=None if dynamic else batch_size) + keras_model = keras.Model(inputs=im, outputs=tf_model.predict(im)) + keras_model.summary() + + LOGGER.info('PyTorch, TensorFlow and Keras models successfully verified.\nUse export.py for TF model export.') + + +def parse_opt(): + parser = argparse.ArgumentParser() + parser.add_argument('--weights', type=str, default=ROOT / 'weightyolov5s.pt', help='weights path') + parser.add_argument('--imgsz', '--img', '--img-size', nargs='+', type=int, default=[640], help='inference size h,w') + parser.add_argument('--batch-size', type=int, default=1, help='batch size') + parser.add_argument('--dynamic', action='store_true', help='dynamic batch size') + opt = parser.parse_args() + opt.imgsz *= 2 if len(opt.imgsz) == 1 else 1 # expand + print_args(FILE.stem, opt) + return opt + + +def main(opt): + run(**vars(opt)) + + +if __name__ == "__main__": + opt = parse_opt() + main(opt) diff --git a/models/yolo.py b/models/yolo.py new file mode 100644 index 0000000..b6dfab8 --- /dev/null +++ b/models/yolo.py @@ -0,0 +1,329 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +YOLO-specific modules + +Usage: + $ python path/to/models/yolo.py --cfg yolov5s.yaml +""" + +import argparse +import sys +from copy import deepcopy +from pathlib import Path + +FILE = Path(__file__).resolve() +ROOT = FILE.parents[1] # YOLOv5 root directory +if str(ROOT) not in sys.path: + sys.path.append(str(ROOT)) # add ROOT to PATH +# ROOT = ROOT.relative_to(Path.cwd()) # relative + +from models.common import * +from models.experimental import * +from utils.autoanchor import check_anchor_order +from utils.general import LOGGER, check_version, check_yaml, make_divisible, print_args +from utils.plots import feature_visualization +from utils.torch_utils import fuse_conv_and_bn, initialize_weights, model_info, scale_img, select_device, time_sync + +try: + import thop # for FLOPs computation +except ImportError: + thop = None + + +class Detect(nn.Module): + stride = None # strides computed during build + onnx_dynamic = False # ONNX export parameter + + def __init__(self, nc=80, anchors=(), ch=(), inplace=True): # detection layer + super().__init__() + self.nc = nc # number of classes + self.no = nc + 5 # number of outputs per anchor + self.nl = len(anchors) # number of detection layers + self.na = len(anchors[0]) // 2 # number of anchors + self.grid = [torch.zeros(1)] * self.nl # init grid + self.anchor_grid = [torch.zeros(1)] * self.nl # init anchor grid + self.register_buffer('anchors', torch.tensor(anchors).float().view(self.nl, -1, 2)) # shape(nl,na,2) + self.m = nn.ModuleList(nn.Conv2d(x, self.no * self.na, 1) for x in ch) # output conv + self.inplace = inplace # use in-place ops (e.g. slice assignment) + + def forward(self, x): + z = [] # inference output + for i in range(self.nl): + x[i] = self.m[i](x[i]) # conv + bs, _, ny, nx = x[i].shape # x(bs,255,20,20) to x(bs,3,20,20,85) + x[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous() + + if not self.training: # inference + if self.onnx_dynamic or self.grid[i].shape[2:4] != x[i].shape[2:4]: + self.grid[i], self.anchor_grid[i] = self._make_grid(nx, ny, i) + + y = x[i].sigmoid() + if self.inplace: + y[..., 0:2] = (y[..., 0:2] * 2 - 0.5 + self.grid[i]) * self.stride[i] # xy + y[..., 2:4] = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i] # wh + else: # for YOLOv5 on AWS Inferentia https://github.com/ultralytics/yolov5/pull/2953 + xy = (y[..., 0:2] * 2 - 0.5 + self.grid[i]) * self.stride[i] # xy + wh = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i] # wh + y = torch.cat((xy, wh, y[..., 4:]), -1) + z.append(y.view(bs, -1, self.no)) + + return x if self.training else (torch.cat(z, 1), x) + + def _make_grid(self, nx=20, ny=20, i=0): + d = self.anchors[i].device + if check_version(torch.__version__, '1.10.0'): # torch>=1.10.0 meshgrid workaround for torch>=0.7 compatibility + yv, xv = torch.meshgrid([torch.arange(ny, device=d), torch.arange(nx, device=d)], indexing='ij') + else: + yv, xv = torch.meshgrid([torch.arange(ny, device=d), torch.arange(nx, device=d)]) + grid = torch.stack((xv, yv), 2).expand((1, self.na, ny, nx, 2)).float() + anchor_grid = (self.anchors[i].clone() * self.stride[i]) \ + .view((1, self.na, 1, 1, 2)).expand((1, self.na, ny, nx, 2)).float() + return grid, anchor_grid + + +class Model(nn.Module): + def __init__(self, cfg='yolov5s.yaml', ch=3, nc=None, anchors=None): # model, input channels, number of classes + super().__init__()#super把cfg='yolov5s.yaml'įš„į›´æŽĨ初始化äē† + if isinstance(cfg, dict): + self.yaml = cfg # model dict + else: # is *.yaml + import yaml # for torch hub + self.yaml_file = Path(cfg).name + with open(cfg, encoding='ascii', errors='ignore') as f: + self.yaml = yaml.safe_load(f) # model dict + + # Define model + ch = self.yaml['ch'] = self.yaml.get('ch', ch) # input channels + if nc and nc != self.yaml['nc']: + LOGGER.info(f"Overriding model.yaml nc={self.yaml['nc']} with nc={nc}") + self.yaml['nc'] = nc # override yaml value + if anchors: + LOGGER.info(f'Overriding model.yaml anchors with anchors={anchors}') + self.yaml['anchors'] = round(anchors) # override yaml value + self.model, self.save = parse_model(deepcopy(self.yaml), ch=[ch]) # model, savelist + self.names = [str(i) for i in range(self.yaml['nc'])] # default names + self.inplace = self.yaml.get('inplace', True)#čŋ™æŽĩäģŖį įœ‹čĩˇæĨ是äģŽä¸€ä¸Ē YAML 文äģļä¸­čŽˇå–é”Žä¸ē 'inplace' įš„å€ŧīŧŒåšļäŊŋį”¨éģ˜čŽ¤å€ŧ True åĻ‚æžœč¯Ĩ锎不存在。 + + # Build strides, anchors + m = self.model[-1] # Detect() + if isinstance(m, Detect): + s = 256 # 2x min stride + m.inplace = self.inplace + m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))]) # forward + m.anchors /= m.stride.view(-1, 1, 1) + check_anchor_order(m) + self.stride = m.stride + self._initialize_biases() # only run once + + # Init weights, biases + initialize_weights(self) + self.info() + LOGGER.info('') + + def forward(self, x, augment=False, profile=False, visualize=False): + if augment: + return self._forward_augment(x) # augmented inference, None + return self._forward_once(x, profile, visualize) # single-scale inference, train + + def _forward_augment(self, x): + img_size = x.shape[-2:] # height, width + s = [1, 0.83, 0.67] # scales + f = [None, 3, None] # flips (2-ud, 3-lr) + y = [] # outputs + for si, fi in zip(s, f): + xi = scale_img(x.flip(fi) if fi else x, si, gs=int(self.stride.max())) + yi = self._forward_once(xi)[0] # forward + # cv2.imwrite(f'img_{si}.jpg', 255 * xi[0].cpu().numpy().transpose((1, 2, 0))[:, :, ::-1]) # save + yi = self._descale_pred(yi, fi, si, img_size) + y.append(yi) + y = self._clip_augmented(y) # clip augmented tails + return torch.cat(y, 1), None # augmented inference, train + + def _forward_once(self, x, profile=False, visualize=False): + y, dt = [], [] # outputs + for m in self.model: + if m.f != -1: # if not from previous layer + x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f] # from earlier layers + if profile: + self._profile_one_layer(m, x, dt) + x = m(x) # run + y.append(x if m.i in self.save else None) # save output + if visualize: + feature_visualization(x, m.type, m.i, save_dir=visualize) + return x + + def _descale_pred(self, p, flips, scale, img_size): + # de-scale predictions following augmented inference (inverse operation) + if self.inplace: + p[..., :4] /= scale # de-scale + if flips == 2: + p[..., 1] = img_size[0] - p[..., 1] # de-flip ud + elif flips == 3: + p[..., 0] = img_size[1] - p[..., 0] # de-flip lr + else: + x, y, wh = p[..., 0:1] / scale, p[..., 1:2] / scale, p[..., 2:4] / scale # de-scale + if flips == 2: + y = img_size[0] - y # de-flip ud + elif flips == 3: + x = img_size[1] - x # de-flip lr + p = torch.cat((x, y, wh, p[..., 4:]), -1) + return p + + def _clip_augmented(self, y): + # Clip YOLOv5 augmented inference tails + nl = self.model[-1].nl # number of detection layers (P3-P5) + g = sum(4 ** x for x in range(nl)) # grid points + e = 1 # exclude layer count + i = (y[0].shape[1] // g) * sum(4 ** x for x in range(e)) # indices + y[0] = y[0][:, :-i] # large + i = (y[-1].shape[1] // g) * sum(4 ** (nl - 1 - x) for x in range(e)) # indices + y[-1] = y[-1][:, i:] # small + return y + + def _profile_one_layer(self, m, x, dt): + c = isinstance(m, Detect) # is final layer, copy input as inplace fix + o = thop.profile(m, inputs=(x.copy() if c else x,), verbose=False)[0] / 1E9 * 2 if thop else 0 # FLOPs + t = time_sync() + for _ in range(10): + m(x.copy() if c else x) + dt.append((time_sync() - t) * 100) + if m == self.model[0]: + LOGGER.info(f"{'time (ms)':>10s} {'GFLOPs':>10s} {'params':>10s} {'module'}") + LOGGER.info(f'{dt[-1]:10.2f} {o:10.2f} {m.np:10.0f} {m.type}') + if c: + LOGGER.info(f"{sum(dt):10.2f} {'-':>10s} {'-':>10s} Total") + + def _initialize_biases(self, cf=None): # initialize biases into Detect(), cf is class frequency + # https://arxiv.org/abs/1708.02002 section 3.3 + # cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1. + m = self.model[-1] # Detect() module + for mi, s in zip(m.m, m.stride): # from + b = mi.bias.view(m.na, -1) # conv.bias(255) to (3,85) + b.data[:, 4] += math.log(8 / (640 / s) ** 2) # obj (8 objects per 640 image) + b.data[:, 5:] += math.log(0.6 / (m.nc - 0.999999)) if cf is None else torch.log(cf / cf.sum()) # cls + mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True) + + def _print_biases(self): + m = self.model[-1] # Detect() module + for mi in m.m: # from + b = mi.bias.detach().view(m.na, -1).T # conv.bias(255) to (3,85) + LOGGER.info( + ('%6g Conv2d.bias:' + '%10.3g' * 6) % (mi.weight.shape[1], *b[:5].mean(1).tolist(), b[5:].mean())) + + # def _print_weights(self): + # for m in self.model.modules(): + # if type(m) is Bottleneck: + # LOGGER.info('%10.3g' % (m.w.detach().sigmoid() * 2)) # shortcut weights + + def fuse(self): # fuse model Conv2d() + BatchNorm2d() layers + LOGGER.info('Fusing layers... ') + for m in self.model.modules(): + if isinstance(m, (Conv, DWConv)) and hasattr(m, 'bn'): + m.conv = fuse_conv_and_bn(m.conv, m.bn) # update conv + delattr(m, 'bn') # remove batchnorm + m.forward = m.forward_fuse # update forward + self.info() + return self + + def info(self, verbose=False, img_size=640): # print model information + model_info(self, verbose, img_size) + + def _apply(self, fn): + # Apply to(), cpu(), cuda(), half() to model tensors that are not parameters or registered buffers + self = super()._apply(fn) + m = self.model[-1] # Detect() + if isinstance(m, Detect): + m.stride = fn(m.stride) + m.grid = list(map(fn, m.grid)) + if isinstance(m.anchor_grid, list): + m.anchor_grid = list(map(fn, m.anchor_grid)) + return self + + +def parse_model(d, ch): # model_dict, input_channels(3) + LOGGER.info(f"\n{'':>3}{'from':>18}{'n':>3}{'params':>10} {'module':<40}{'arguments':<30}") + anchors, nc, gd, gw = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple'] + na = (len(anchors[0]) // 2) if isinstance(anchors, list) else anchors # number of anchors + no = na * (nc + 5) # number of outputs = anchors * (classes + 5) + + layers, save, c2 = [], [], ch[-1] # layers, savelist, ch out + for i, (f, n, m, args) in enumerate(d['backbone'] + d['head']): # from, number, module, args + m = eval(m) if isinstance(m, str) else m # eval strings + for j, a in enumerate(args): + try: + args[j] = eval(a) if isinstance(a, str) else a # eval strings + except NameError: + pass + + n = n_ = max(round(n * gd), 1) if n > 1 else n # depth gain + if m in [Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, + BottleneckCSP, C3, C3TR, C3SPP, C3Ghost]: + c1, c2 = ch[f], args[0] + if c2 != no: # if not output + c2 = make_divisible(c2 * gw, 8)#??????? + + args = [c1, c2, *args[1:]] + if m in [BottleneckCSP, C3, C3TR, C3Ghost]: + args.insert(2, n) # number of repeats + n = 1 + elif m is nn.BatchNorm2d: + args = [ch[f]] + elif m is Concat: + c2 = sum(ch[x] for x in f) + elif m is Detect: + args.append([ch[x] for x in f]) + if isinstance(args[1], int): # number of anchors + args[1] = [list(range(args[1] * 2))] * len(f) + elif m is Contract: + c2 = ch[f] * args[0] ** 2 + elif m is Expand: + c2 = ch[f] // args[0] ** 2 + else: + c2 = ch[f] + + m_ = nn.Sequential(*(m(*args) for _ in range(n))) if n > 1 else m(*args) # module + t = str(m)[8:-2].replace('__main__.', '') # module type + np = sum(x.numel() for x in m_.parameters()) # number params + m_.i, m_.f, m_.type, m_.np = i, f, t, np # attach index, 'from' index, type, number params + LOGGER.info(f'{i:>3}{str(f):>18}{n_:>3}{np:10.0f} {t:<40}{str(args):<30}') # print + save.extend(x % i for x in ([f] if isinstance(f, int) else f) if x != -1) # append to savelist + layers.append(m_) + if i == 0: + ch = [] + ch.append(c2) + return nn.Sequential(*layers), sorted(save) + + +if __name__ == '__main__': + parser = argparse.ArgumentParser() + parser.add_argument('--cfg', type=str, default='yolov5s.yaml', help='model.yaml') + parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu') + parser.add_argument('--profile', action='store_true', help='profile model speed') + parser.add_argument('--test', action='store_true', help='test all yolo*.yaml') + opt = parser.parse_args() + opt.cfg = check_yaml(opt.cfg) # check YAML + print_args(FILE.stem, opt) + device = select_device(opt.device) + + # Create model + model = Model(opt.cfg).to(device) + model.train() + + # Profile + if opt.profile: + img = torch.rand(8 if torch.cuda.is_available() else 1, 3, 640, 640).to(device) + y = model(img, profile=True) + + # Test all models + if opt.test: + for cfg in Path(ROOT / 'models').rglob('yolo*.yaml'): + try: + _ = Model(cfg) + except Exception as e: + print(f'Error in {cfg}: {e}') + + # Tensorboard (not working https://github.com/ultralytics/yolov5/issues/2898) + # from torch.utils.tensorboard import SummaryWriter + # tb_writer = SummaryWriter('.') + # LOGGER.info("Run 'tensorboard --logdir=models' to view tensorboard at http://localhost:6006/") + # tb_writer.add_graph(torch.jit.trace(model, img, strict=False), []) # add model graph diff --git a/models/yolov5l.yaml b/models/yolov5l.yaml new file mode 100644 index 0000000..ce8a5de --- /dev/null +++ b/models/yolov5l.yaml @@ -0,0 +1,48 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.0 # model depth multiple +width_multiple: 1.0 # layer channel multiple +anchors: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 9 + ] + +# YOLOv5 v6.0 head +head: + [[-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 13 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 17 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 14], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 20 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 10], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [1024, False]], # 23 (P5/32-large) + + [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) + ] diff --git a/models/yolov5m.yaml b/models/yolov5m.yaml new file mode 100644 index 0000000..ad13ab3 --- /dev/null +++ b/models/yolov5m.yaml @@ -0,0 +1,48 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 0.67 # model depth multiple +width_multiple: 0.75 # layer channel multiple +anchors: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 9 + ] + +# YOLOv5 v6.0 head +head: + [[-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 13 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 17 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 14], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 20 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 10], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [1024, False]], # 23 (P5/32-large) + + [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) + ] diff --git a/models/yolov5n.yaml b/models/yolov5n.yaml new file mode 100644 index 0000000..8a28a40 --- /dev/null +++ b/models/yolov5n.yaml @@ -0,0 +1,48 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 0.33 # model depth multiple +width_multiple: 0.25 # layer channel multiple +anchors: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 9 + ] + +# YOLOv5 v6.0 head +head: + [[-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 13 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 17 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 14], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 20 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 10], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [1024, False]], # 23 (P5/32-large) + + [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) + ] diff --git a/models/yolov5s.yaml b/models/yolov5s.yaml new file mode 100644 index 0000000..f35beab --- /dev/null +++ b/models/yolov5s.yaml @@ -0,0 +1,48 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 0.33 # model depth multiple +width_multiple: 0.50 # layer channel multiple +anchors: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 9 + ] + +# YOLOv5 v6.0 head +head: + [[-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 13 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 17 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 14], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 20 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 10], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [1024, False]], # 23 (P5/32-large) + + [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) + ] diff --git a/models/yolov5x.yaml b/models/yolov5x.yaml new file mode 100644 index 0000000..f617a02 --- /dev/null +++ b/models/yolov5x.yaml @@ -0,0 +1,48 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license + +# Parameters +nc: 80 # number of classes +depth_multiple: 1.33 # model depth multiple +width_multiple: 1.25 # layer channel multiple +anchors: + - [10,13, 16,30, 33,23] # P3/8 + - [30,61, 62,45, 59,119] # P4/16 + - [116,90, 156,198, 373,326] # P5/32 + +# YOLOv5 v6.0 backbone +backbone: + # [from, number, module, args] + [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 + [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 + [-1, 3, C3, [128]], + [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 + [-1, 6, C3, [256]], + [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 + [-1, 9, C3, [512]], + [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 + [-1, 3, C3, [1024]], + [-1, 1, SPPF, [1024, 5]], # 9 + ] + +# YOLOv5 v6.0 head +head: + [[-1, 1, Conv, [512, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 6], 1, Concat, [1]], # cat backbone P4 + [-1, 3, C3, [512, False]], # 13 + + [-1, 1, Conv, [256, 1, 1]], + [-1, 1, nn.Upsample, [None, 2, 'nearest']], + [[-1, 4], 1, Concat, [1]], # cat backbone P3 + [-1, 3, C3, [256, False]], # 17 (P3/8-small) + + [-1, 1, Conv, [256, 3, 2]], + [[-1, 14], 1, Concat, [1]], # cat head P4 + [-1, 3, C3, [512, False]], # 20 (P4/16-medium) + + [-1, 1, Conv, [512, 3, 2]], + [[-1, 10], 1, Concat, [1]], # cat head P5 + [-1, 3, C3, [1024, False]], # 23 (P5/32-large) + + [[17, 20, 23], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) + ] diff --git a/requirements.txt b/requirements.txt new file mode 100644 index 0000000..96fc9d1 --- /dev/null +++ b/requirements.txt @@ -0,0 +1,37 @@ +# pip install -r requirements.txt + +# Base ---------------------------------------- +matplotlib>=3.2.2 +numpy>=1.18.5 +opencv-python>=4.1.2 +Pillow>=7.1.2 +PyYAML>=5.3.1 +requests>=2.23.0 +scipy>=1.4.1 +torch>=1.7.0 +torchvision>=0.8.1 +tqdm>=4.41.0 + +# Logging ------------------------------------- +tensorboard>=2.4.1 +# wandb + +# Plotting ------------------------------------ +pandas>=1.1.4 +seaborn>=0.11.0 + +# Export -------------------------------------- +# coremltools>=4.1 # CoreML export +# onnx>=1.9.0 # ONNX export +# onnx-simplifier>=0.3.6 # ONNX simplifier +# scikit-learn==0.19.2 # CoreML quantization +# tensorflow>=2.4.1 # TFLite export +# tensorflowjs>=3.9.0 # TF.js export +# openvino-dev # OpenVINO export + +# Extras -------------------------------------- +# albumentations>=1.0.3 +# Cython # for pycocotools https://github.com/cocodataset/cocoapi/issues/172 +# pycocotools>=2.0 # COCO mAP +# roboflow +thop # FLOPs computation diff --git a/setup.cfg b/setup.cfg new file mode 100644 index 0000000..20ea49a --- /dev/null +++ b/setup.cfg @@ -0,0 +1,45 @@ +# Project-wide configuration file, can be used for package metadata and other toll configurations +# Example usage: global configuration for PEP8 (via flake8) setting or default pytest arguments + +[metadata] +license_file = LICENSE +description-file = README.md + + +[tool:pytest] +norecursedirs = + .git + dist + build +addopts = + --doctest-modules + --durations=25 + --color=yes + + +[flake8] +max-line-length = 120 +exclude = .tox,*.egg,build,temp +select = E,W,F +doctests = True +verbose = 2 +# https://pep8.readthedocs.io/en/latest/intro.html#error-codes +format = pylint +# see: https://www.flake8rules.com/ +ignore = + E731 # Do not assign a lambda expression, use a def + F405 # name may be undefined, or defined from star imports: module + E402 # module level import not at top of file + F401 # module imported but unused + W504 # line break after binary operator + E127 # continuation line over-indented for visual indent + W504 # line break after binary operator + E231 # missing whitespace after ‘,’, ‘;’, or ‘:’ + E501 # line too long + F403 # ‘from module import *’ used; unable to detect undefined names + + +[isort] +# https://pycqa.github.io/isort/docs/configuration/options.html +line_length = 120 +multi_line_output = 0 diff --git a/train.py b/train.py new file mode 100644 index 0000000..f6480d8 --- /dev/null +++ b/train.py @@ -0,0 +1,644 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Train a YOLOv5 model on a custom dataset. + +Models and datasets download automatically from the latest YOLOv5 release. +Models: https://github.com/ultralytics/yolov5/tree/master/models +Datasets: https://github.com/ultralytics/yolov5/tree/master/data +Tutorial: https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data + +Usage: + $ python path/to/train.py --data coco128.yaml --weights yolov5s.pt --img 640 # from pretrained (RECOMMENDED) + $ python path/to/train.py --data coco128.yaml --weights '' --cfg yolov5s.yaml --img 640 # from scratch +""" + +import argparse +import math +import os +import random +import sys +import time +from copy import deepcopy +from datetime import datetime +from pathlib import Path + +import numpy as np +import torch +import torch.distributed as dist +import torch.nn as nn +import yaml +from torch.cuda import amp +from torch.nn.parallel import DistributedDataParallel as DDP +from torch.optim import SGD, Adam, AdamW, lr_scheduler +from tqdm import tqdm + +FILE = Path(__file__).resolve() +ROOT = FILE.parents[0] # YOLOv5 root directory +if str(ROOT) not in sys.path: + sys.path.append(str(ROOT)) # add ROOT to PATH +ROOT = Path(os.path.relpath(ROOT, Path.cwd())) # relative + +import val # for end-of-epoch mAP +from models.experimental import attempt_load +from models.yolo import Model +from utils.autoanchor import check_anchors +from utils.autobatch import check_train_batch_size +from utils.callbacks import Callbacks +from utils.datasets import create_dataloader +from utils.downloads import attempt_download +from utils.general import (LOGGER, check_dataset, check_file, check_git_status, check_img_size, check_requirements, + check_suffix, check_yaml, colorstr, get_latest_run, increment_path, init_seeds, + intersect_dicts, labels_to_class_weights, labels_to_image_weights, methods, one_cycle, + print_args, print_mutation, strip_optimizer) +from utils.loggers import Loggers +from utils.loggers.wandb.wandb_utils import check_wandb_resume +from utils.loss import ComputeLoss +from utils.metrics import fitness +from utils.plots import plot_evolve, plot_labels +from utils.torch_utils import EarlyStopping, ModelEMA, de_parallel, select_device, torch_distributed_zero_first + +LOCAL_RANK = int(os.getenv('LOCAL_RANK', -1)) # https://pytorch.org/docs/stable/elastic/run.html +RANK = int(os.getenv('RANK', -1)) +WORLD_SIZE = int(os.getenv('WORLD_SIZE', 1)) + + +def train(hyp, # path/to/hyp.yaml or hyp dictionary + opt, + device, + callbacks + ): + save_dir, epochs, batch_size, weights, single_cls, evolve, data, cfg, resume, noval, nosave, workers, freeze = \ + Path(opt.save_dir), opt.epochs, opt.batch_size, opt.weights, opt.single_cls, opt.evolve, opt.data, opt.cfg, \ + opt.resume, opt.noval, opt.nosave, opt.workers, opt.freeze + + # Directories + w = save_dir / 'weights' # weights dir + (w.parent if evolve else w).mkdir(parents=True, exist_ok=True) # make dir + last, best = w / 'last.pt', w / 'best.pt' + + # Hyperparameters + if isinstance(hyp, str): + with open(hyp, errors='ignore') as f: + hyp = yaml.safe_load(f) # load hyps dict + LOGGER.info(colorstr('hyperparameters: ') + ', '.join(f'{k}={v}' for k, v in hyp.items())) + + # Save run settings + if not evolve: + with open(save_dir / 'hyp.yaml', 'w') as f: + yaml.safe_dump(hyp, f, sort_keys=False) + with open(save_dir / 'opt.yaml', 'w') as f: + yaml.safe_dump(vars(opt), f, sort_keys=False) + + # Loggers + data_dict = None + if RANK in [-1, 0]: + loggers = Loggers(save_dir, weights, opt, hyp, LOGGER) # loggers instance + if loggers.wandb: + data_dict = loggers.wandb.data_dict + if resume: + weights, epochs, hyp, batch_size = opt.weights, opt.epochs, opt.hyp, opt.batch_size + + # Register actions + for k in methods(loggers): + callbacks.register_action(k, callback=getattr(loggers, k)) + + # Config + plots = not evolve # create plots + cuda = device.type != 'cpu' + init_seeds(1 + RANK)#初始化随æœē数į§å­ ä¸ēäē†čŽŠæ¯æŦĄéƒŊčƒŊ复įŽ°å‡ēä¸€æ ˇįš„įģ“æžœ + with torch_distributed_zero_first(LOCAL_RANK): + data_dict = data_dict or check_dataset(data) # check if None + train_path, val_path = data_dict['train'], data_dict['val'] + nc = 1 if single_cls else int(data_dict['nc']) # number of classes + names = ['item'] if single_cls and len(data_dict['names']) != 1 else data_dict['names'] # class names + assert len(names) == nc, f'{len(names)} names found for nc={nc} dataset in {data}' # check + is_coco = isinstance(val_path, str) and val_path.endswith('coco/val2017.txt') # COCO dataset + + # Model + check_suffix(weights, '.pt') # check weights + pretrained = weights.endswith('.pt') + if pretrained: + with torch_distributed_zero_first(LOCAL_RANK): + weights = attempt_download(weights) # download if not found locally + ckpt = torch.load(weights, map_location='cpu') # load checkpoint to CPU to avoid CUDA memory leak + model = Model(cfg or ckpt['model'].yaml, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device) # create + exclude = ['anchor'] if (cfg or hyp.get('anchors')) and not resume else [] # exclude keys + csd = ckpt['model'].float().state_dict() # checkpoint state_dict as FP32 + csd = intersect_dicts(csd, model.state_dict(), exclude=exclude) # intersect + model.load_state_dict(csd, strict=False) # load + LOGGER.info(f'Transferred {len(csd)}/{len(model.state_dict())} items from {weights}') # report + else: + model = Model(cfg, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device) # create + + # Freeze + freeze = [f'model.{x}.' for x in (freeze if len(freeze) > 1 else range(freeze[0]))] # layers to freeze + for k, v in model.named_parameters(): + v.requires_grad = True # train all layers + if any(x in k for x in freeze): + LOGGER.info(f'freezing {k}') + v.requires_grad = False + + # Image size + gs = max(int(model.stride.max()), 32) # grid size (max stride) + imgsz = check_img_size(opt.imgsz, gs, floor=gs * 2) # verify imgsz is gs-multiple + + # Batch size + if RANK == -1 and batch_size == -1: # single-GPU only, estimate best batch size + batch_size = check_train_batch_size(model, imgsz) + loggers.on_params_update({"batch_size": batch_size}) + + # Optimizer + nbs = 64 # nominal batch size 名䚉上įš„ + accumulate = max(round(nbs / batch_size), 1) # accumulate loss before optimizing + hyp['weight_decay'] *= batch_size * accumulate / nbs # scale weight_decay + LOGGER.info(f"Scaled weight_decay = {hyp['weight_decay']}") + + g0, g1, g2 = [], [], [] # optimizer parameter groups + for v in model.modules(): + if hasattr(v, 'bias') and isinstance(v.bias, nn.Parameter): # bias + g2.append(v.bias) + if isinstance(v, nn.BatchNorm2d): # weight (no decay) + g0.append(v.weight) + elif hasattr(v, 'weight') and isinstance(v.weight, nn.Parameter): # weight (with decay) + g1.append(v.weight) + + if opt.optimizer == 'Adam': + optimizer = Adam(g0, lr=hyp['lr0'], betas=(hyp['momentum'], 0.999)) # adjust beta1 to momentum + elif opt.optimizer == 'AdamW': + optimizer = AdamW(g0, lr=hyp['lr0'], betas=(hyp['momentum'], 0.999)) # adjust beta1 to momentum + else: + optimizer = SGD(g0, lr=hyp['lr0'], momentum=hyp['momentum'], nesterov=True) + + optimizer.add_param_group({'params': g1, 'weight_decay': hyp['weight_decay']}) # add g1 with weight_decay + optimizer.add_param_group({'params': g2}) # add g2 (biases) + LOGGER.info(f"{colorstr('optimizer:')} {type(optimizer).__name__} with parameter groups " + f"{len(g0)} weight (no decay), {len(g1)} weight, {len(g2)} bias") + del g0, g1, g2 + + # Scheduler + if opt.cos_lr: + lf = one_cycle(1, hyp['lrf'], epochs) # cosine 1->hyp['lrf'] + else: + lf = lambda x: (1 - x / epochs) * (1.0 - hyp['lrf']) + hyp['lrf'] # linear + scheduler = lr_scheduler.LambdaLR(optimizer, lr_lambda=lf) # plot_lr_scheduler(optimizer, scheduler, epochs) + + # EMA + ema = ModelEMA(model) if RANK in [-1, 0] else None + + # Resume + start_epoch, best_fitness = 0, 0.0 + if pretrained: + # Optimizer + if ckpt['optimizer'] is not None: + optimizer.load_state_dict(ckpt['optimizer']) + best_fitness = ckpt['best_fitness'] + + # EMA + if ema and ckpt.get('ema'): + ema.ema.load_state_dict(ckpt['ema'].float().state_dict()) + ema.updates = ckpt['updates'] + + # Epochs + start_epoch = ckpt['epoch'] + 1 + if resume: + assert start_epoch > 0, f'{weights} training to {epochs} epochs is finished, nothing to resume.' + if epochs < start_epoch: + LOGGER.info(f"{weights} has been trained for {ckpt['epoch']} epochs. Fine-tuning for {epochs} more epochs.") + epochs += ckpt['epoch'] # finetune additional epochs + + del ckpt, csd + + # DP mode + if cuda and RANK == -1 and torch.cuda.device_count() > 1: + LOGGER.warning('WARNING: DP not recommended, use torch.distributed.run for best DDP Multi-GPU results.\n' + 'See Multi-GPU Tutorial at https://github.com/ultralytics/yolov5/issues/475 to get started.') + model = torch.nn.DataParallel(model) + + # SyncBatchNorm + if opt.sync_bn and cuda and RANK != -1: + model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model).to(device) + LOGGER.info('Using SyncBatchNorm()') + + # Trainloader + train_loader, dataset = create_dataloader(train_path, imgsz, batch_size // WORLD_SIZE, gs, single_cls, + hyp=hyp, augment=True, cache=None if opt.cache == 'val' else opt.cache, + rect=opt.rect, rank=LOCAL_RANK, workers=workers, + image_weights=opt.image_weights, quad=opt.quad, + prefix=colorstr('train: '), shuffle=True) + mlc = int(np.concatenate(dataset.labels, 0)[:, 0].max()) # max label class ?有äģ€äšˆæ„äš‰ + nb = len(train_loader) # number of batches ä¸€å…ą128åŧ  batchsize 16 一扚16åŧ  所äģĨ有8扚 + assert mlc < nc, f'Label class {mlc} exceeds nc={nc} in {data}. Possible class labels are 0-{nc - 1}'#į”¨æĨæ–­č¨€é˜˛æ­ĸå‡ē错įš„ + + # Process 0 + if RANK in [-1, 0]: + val_loader = create_dataloader(val_path, imgsz, batch_size // WORLD_SIZE * 2, gs, single_cls, + hyp=hyp, cache=None if noval else opt.cache, + rect=True, rank=-1, workers=workers * 2, pad=0.5, + prefix=colorstr('val: '))[0] + + if not resume: + labels = np.concatenate(dataset.labels, 0) + # c = torch.tensor(labels[:, 0]) # classes + # cf = torch.bincount(c.long(), minlength=nc) + 1. # frequency + # model._initialize_biases(cf.to(device)) + if plots: + plot_labels(labels, names, save_dir) + + # Anchors + if not opt.noautoanchor: + check_anchors(dataset, model=model, thr=hyp['anchor_t'], imgsz=imgsz) + model.half().float() # pre-reduce anchor precision åŗ将pytorchéģ˜čŽ¤įš„32äŊæĩŽį‚šåž‹éƒŊ攚成16äŊæĩŽį‚šåž‹ã€‚ + + callbacks.run('on_pretrain_routine_end') + + # DDP mode + if cuda and RANK != -1: + model = DDP(model, device_ids=[LOCAL_RANK], output_device=LOCAL_RANK) + + # Model attributes + nl = de_parallel(model).model[-1].nl # number of detection layers (to scale hyps) + hyp['box'] *= 3 / nl # scale to layers + hyp['cls'] *= nc / 80 * 3 / nl # scale to classes and layers + hyp['obj'] *= (imgsz / 640) ** 2 * 3 / nl # scale to image size and layers + hyp['label_smoothing'] = opt.label_smoothing + model.nc = nc # attach number of classes to model + model.hyp = hyp # attach hyperparameters to model + model.class_weights = labels_to_class_weights(dataset.labels, nc).to(device) * nc # attach class weights + model.names = names + + # Start training + t0 = time.time() + nw = max(round(hyp['warmup_epochs'] * nb), 1000) # number of warmup iterations, max(3 epochs, 1k iterations) + # nw = min(nw, (epochs - start_epoch) / 2 * nb) # limit warmup to < 1/2 of training + last_opt_step = -1 + maps = np.zeros(nc) # mAP per class + results = (0, 0, 0, 0, 0, 0, 0) # P, R, mAP@.5, mAP@.5-.95, val_loss(box, obj, cls) + scheduler.last_epoch = start_epoch - 1 # do not move + scaler = amp.GradScaler(enabled=cuda) + stopper = EarlyStopping(patience=opt.patience) + compute_loss = ComputeLoss(model) # init loss class + LOGGER.info(f'Image sizes {imgsz} train, {imgsz} val\n' + f'Using {train_loader.num_workers * WORLD_SIZE} dataloader workers\n' + f"Logging results to {colorstr('bold', save_dir)}\n" + f'Starting training for {epochs} epochs...') + for epoch in range(start_epoch, epochs): # epoch ------------------------------------------------------------------ + model.train() + + # Update image weights (optional, single-GPU only) + if opt.image_weights: + cw = model.class_weights.cpu().numpy() * (1 - maps) ** 2 / nc # class weights + iw = labels_to_image_weights(dataset.labels, nc=nc, class_weights=cw) # image weights + dataset.indices = random.choices(range(dataset.n), weights=iw, k=dataset.n) # rand weighted idx + + # Update mosaic border (optional) + # b = int(random.uniform(0.25 * imgsz, 0.75 * imgsz + gs) // gs * gs) + # dataset.mosaic_border = [b - imgsz, -b] # height, width borders + + mloss = torch.zeros(3, device=device) # mean losses + if RANK != -1: + train_loader.sampler.set_epoch(epoch) + pbar = enumerate(train_loader) + LOGGER.info(('\n' + '%10s' * 7) % ('Epoch', 'gpu_mem', 'box', 'obj', 'cls', 'labels', 'img_size')) + if RANK in [-1, 0]: + pbar = tqdm(pbar, total=nb, bar_format='{l_bar}{bar:10}{r_bar}{bar:-10b}') # progress bar + optimizer.zero_grad() + for i, (imgs, targets, paths, _) in pbar: # batch ------------------------------------------------------------- + ni = i + nb * epoch # number integrated batches (since train start) + imgs = imgs.to(device, non_blocking=True).float() / 255 # uint8 to float32, 0-255 to 0.0-1.0 + + # Warmup + if ni <= nw: + xi = [0, nw] # x interp + # compute_loss.gr = np.interp(ni, xi, [0.0, 1.0]) # iou loss ratio (obj_loss = 1.0 or iou) + accumulate = max(1, np.interp(ni, xi, [1, nbs / batch_size]).round()) + for j, x in enumerate(optimizer.param_groups): + # bias lr falls from 0.1 to lr0, all other lrs rise from 0.0 to lr0 + x['lr'] = np.interp(ni, xi, [hyp['warmup_bias_lr'] if j == 2 else 0.0, x['initial_lr'] * lf(epoch)]) + if 'momentum' in x: + x['momentum'] = np.interp(ni, xi, [hyp['warmup_momentum'], hyp['momentum']]) + + # Multi-scale + if opt.multi_scale: + sz = random.randrange(imgsz * 0.5, imgsz * 1.5 + gs) // gs * gs # size + sf = sz / max(imgs.shape[2:]) # scale factor + if sf != 1: + ns = [math.ceil(x * sf / gs) * gs for x in imgs.shape[2:]] # new shape (stretched to gs-multiple) + imgs = nn.functional.interpolate(imgs, size=ns, mode='bilinear', align_corners=False) + + # Forward + with amp.autocast(enabled=cuda): + pred = model(imgs) # forward + loss, loss_items = compute_loss(pred, targets.to(device)) # loss scaled by batch_size + if RANK != -1: + loss *= WORLD_SIZE # gradient averaged between devices in DDP mode + if opt.quad: + loss *= 4. + + # Backward + scaler.scale(loss).backward() + + # Optimize + if ni - last_opt_step >= accumulate: + scaler.step(optimizer) # optimizer.step + scaler.update() + optimizer.zero_grad() + if ema: + ema.update(model) + last_opt_step = ni + + # Log + if RANK in [-1, 0]: + mloss = (mloss * i + loss_items) / (i + 1) # update mean losses + mem = f'{torch.cuda.memory_reserved() / 1E9 if torch.cuda.is_available() else 0:.3g}G' # (GB) + pbar.set_description(('%10s' * 2 + '%10.4g' * 5) % ( + f'{epoch}/{epochs - 1}', mem, *mloss, targets.shape[0], imgs.shape[-1])) + callbacks.run('on_train_batch_end', ni, model, imgs, targets, paths, plots, opt.sync_bn) + if callbacks.stop_training: + return + # end batch ------------------------------------------------------------------------------------------------ + + # Scheduler + lr = [x['lr'] for x in optimizer.param_groups] # for loggers + scheduler.step() + + if RANK in [-1, 0]: + # mAP + callbacks.run('on_train_epoch_end', epoch=epoch) + ema.update_attr(model, include=['yaml', 'nc', 'hyp', 'names', 'stride', 'class_weights']) + final_epoch = (epoch + 1 == epochs) or stopper.possible_stop + if not noval or final_epoch: # Calculate mAP + results, maps, _ = val.run(data_dict, + batch_size=batch_size // WORLD_SIZE * 2, + imgsz=imgsz, + model=ema.ema, + single_cls=single_cls, + dataloader=val_loader, + save_dir=save_dir, + plots=False, + callbacks=callbacks, + compute_loss=compute_loss) + + # Update best mAP + fi = fitness(np.array(results).reshape(1, -1)) # weighted combination of [P, R, mAP@.5, mAP@.5-.95] + if fi > best_fitness: + best_fitness = fi + log_vals = list(mloss) + list(results) + lr + callbacks.run('on_fit_epoch_end', log_vals, epoch, best_fitness, fi) + + # Save model + if (not nosave) or (final_epoch and not evolve): # if save + ckpt = {'epoch': epoch, + 'best_fitness': best_fitness, + 'model': deepcopy(de_parallel(model)).half(), + 'ema': deepcopy(ema.ema).half(), + 'updates': ema.updates, + 'optimizer': optimizer.state_dict(), + 'wandb_id': loggers.wandb.wandb_run.id if loggers.wandb else None, + 'date': datetime.now().isoformat()} + + # Save last, best and delete + torch.save(ckpt, last) + if best_fitness == fi: + torch.save(ckpt, best) + if (epoch > 0) and (opt.save_period > 0) and (epoch % opt.save_period == 0): + torch.save(ckpt, w / f'epoch{epoch}.pt') + del ckpt + callbacks.run('on_model_save', last, epoch, final_epoch, best_fitness, fi) + + # Stop Single-GPU + if RANK == -1 and stopper(epoch=epoch, fitness=fi): + break + + # Stop DDP TODO: known issues shttps://github.com/ultralytics/yolov5/pull/4576 + # stop = stopper(epoch=epoch, fitness=fi) + # if RANK == 0: + # dist.broadcast_object_list([stop], 0) # broadcast 'stop' to all ranks + + # Stop DPP + # with torch_distributed_zero_first(RANK): + # if stop: + # break # must break all DDP ranks + + # end epoch ---------------------------------------------------------------------------------------------------- + # end training ----------------------------------------------------------------------------------------------------- + if RANK in [-1, 0]: + LOGGER.info(f'\n{epoch - start_epoch + 1} epochs completed in {(time.time() - t0) / 3600:.3f} hours.') + for f in last, best: + if f.exists(): + strip_optimizer(f) # strip optimizers + if f is best: + LOGGER.info(f'\nValidating {f}...') + results, _, _ = val.run(data_dict, + batch_size=batch_size // WORLD_SIZE * 2, + imgsz=imgsz, + model=attempt_load(f, device).half(), + iou_thres=0.65 if is_coco else 0.60, # best pycocotools results at 0.65 + single_cls=single_cls, + dataloader=val_loader, + save_dir=save_dir, + save_json=is_coco, + verbose=True, + plots=True, + callbacks=callbacks, + compute_loss=compute_loss) # val best model with plots + if is_coco: + callbacks.run('on_fit_epoch_end', list(mloss) + list(results) + lr, epoch, best_fitness, fi) + + callbacks.run('on_train_end', last, best, plots, epoch, results) + LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}") + + torch.cuda.empty_cache() + return results + + +def parse_opt(known=False): + parser = argparse.ArgumentParser() + parser.add_argument('--weights', type=str, default=ROOT / 'yolov5s.pt', help='initial weights path') + parser.add_argument('--cfg', type=str, default='', help='model.yaml path') + #parser.add_argument('--data', type=str, default=ROOT / 'data/coco128.yaml', help='dataset.yaml path') + parser.add_argument('--data', type=str, default=ROOT / 'data/hand.yaml', help='dataset.yaml path') + parser.add_argument('--hyp', type=str, default=ROOT / 'data/hyps/hyp.scratch-low.yaml', help='hyperparameters path') + parser.add_argument('--epochs', type=int, default=300) + parser.add_argument('--batch-size', type=int, default=8, help='total batch size for all GPUs, -1 for autobatch') + parser.add_argument('--imgsz', '--img', '--img-size', type=int, default=640, help='train, val image size (pixels)') + parser.add_argument('--rect', action='store_true', help='rectangular training') + parser.add_argument('--resume', nargs='?', const=True, default=False, help='resume most recent training') + parser.add_argument('--nosave', action='store_true', help='only save final checkpoint') + parser.add_argument('--noval', default=True,action='store_true', help='only validate final epoch') + parser.add_argument('--noautoanchor', action='store_true', help='disable AutoAnchor') + parser.add_argument('--evolve', type=int, nargs='?', const=300, help='evolve hyperparameters for x generations') + parser.add_argument('--bucket', type=str, default='', help='gsutil bucket') + parser.add_argument('--cache', type=str, nargs='?', const='ram', help='--cache images in "ram" (default) or "disk"') + parser.add_argument('--image-weights', action='store_true', help='use weighted image selection for training') + parser.add_argument('--device', default='0', help='cuda device, i.e. 0 or 0,1,2,3 or cpu') + parser.add_argument('--multi-scale',default=True, action='store_true', help='vary img-size +/- 50%%') + parser.add_argument('--single-cls', default=True,action='store_true', help='train multi-class data as single-class') + parser.add_argument('--optimizer', type=str, choices=['SGD', 'Adam', 'AdamW'], default='SGD', help='optimizer') + parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode') + parser.add_argument('--workers', type=int, default=8, help='max dataloader workers (per RANK in DDP mode)') + parser.add_argument('--project', default=ROOT / 'runs/train', help='save to project/name') + parser.add_argument('--name', default='exp', help='save to project/name') + parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment') + parser.add_argument('--quad', action='store_true', help='quad dataloader') + parser.add_argument('--cos-lr', action='store_true', help='cosine LR scheduler') + parser.add_argument('--label-smoothing', type=float, default=0.0, help='Label smoothing epsilon') + parser.add_argument('--patience', type=int, default=100, help='EarlyStopping patience (epochs without improvement)') + parser.add_argument('--freeze', nargs='+', type=int, default=[0], help='Freeze layers: backbone=10, first3=0 1 2') + parser.add_argument('--save-period', type=int, default=-1, help='Save checkpoint every x epochs (disabled if < 1)') + parser.add_argument('--local_rank', type=int, default=-1, help='DDP parameter, do not modify') + + # Weights & Biases arguments + parser.add_argument('--entity', default=None, help='W&B: Entity') + parser.add_argument('--upload_dataset', nargs='?', const=True, default=False, help='W&B: Upload data, "val" option') + parser.add_argument('--bbox_interval', type=int, default=-1, help='W&B: Set bounding-box image logging interval') + parser.add_argument('--artifact_alias', type=str, default='latest', help='W&B: Version of dataset artifact to use') + + opt = parser.parse_known_args()[0] if known else parser.parse_args() + return opt + + +def main(opt, callbacks=Callbacks()): + # Checks + if RANK in [-1, 0]: + print_args(FILE.stem, opt) + check_git_status() + check_requirements(exclude=['thop']) + + # Resume + if opt.resume and not check_wandb_resume(opt) and not opt.evolve: # resume an interrupted run + ckpt = opt.resume if isinstance(opt.resume, str) else get_latest_run() # specified or most recent path + assert os.path.isfile(ckpt), 'ERROR: --resume checkpoint does not exist' + with open(Path(ckpt).parent.parent / 'opt.yaml', errors='ignore') as f: + opt = argparse.Namespace(**yaml.safe_load(f)) # replace + opt.cfg, opt.weights, opt.resume = '', ckpt, True # reinstate + LOGGER.info(f'Resuming training from {ckpt}') + else: + opt.data, opt.cfg, opt.hyp, opt.weights, opt.project = \ + check_file(opt.data), check_yaml(opt.cfg), check_yaml(opt.hyp), str(opt.weights), str(opt.project) # checks + assert len(opt.cfg) or len(opt.weights), 'either --cfg or --weights must be specified' + if opt.evolve: + if opt.project == str(ROOT / 'runs/train'): # if default project name, rename to runs/evolve + opt.project = str(ROOT / 'runs/evolve') + opt.exist_ok, opt.resume = opt.resume, False # pass resume to exist_ok and disable resume + opt.save_dir = str(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok)) + + # DDP mode + device = select_device(opt.device, batch_size=opt.batch_size) + if LOCAL_RANK != -1: + msg = 'is not compatible with YOLOv5 Multi-GPU DDP training' + assert not opt.image_weights, f'--image-weights {msg}' + assert not opt.evolve, f'--evolve {msg}' + assert opt.batch_size != -1, f'AutoBatch with --batch-size -1 {msg}, please pass a valid --batch-size' + assert opt.batch_size % WORLD_SIZE == 0, f'--batch-size {opt.batch_size} must be multiple of WORLD_SIZE' + assert torch.cuda.device_count() > LOCAL_RANK, 'insufficient CUDA devices for DDP command' + torch.cuda.set_device(LOCAL_RANK) + device = torch.device('cuda', LOCAL_RANK) + dist.init_process_group(backend="nccl" if dist.is_nccl_available() else "gloo") + + # Train + if not opt.evolve: + train(opt.hyp, opt, device, callbacks) + if WORLD_SIZE > 1 and RANK == 0: + LOGGER.info('Destroying process group... ') + dist.destroy_process_group() + + # Evolve hyperparameters (optional) + else: + # Hyperparameter evolution metadata (mutation scale 0-1, lower_limit, upper_limit) + meta = {'lr0': (1, 1e-5, 1e-1), # initial learning rate (SGD=1E-2, Adam=1E-3) + 'lrf': (1, 0.01, 1.0), # final OneCycleLR learning rate (lr0 * lrf) + 'momentum': (0.3, 0.6, 0.98), # SGD momentum/Adam beta1 + 'weight_decay': (1, 0.0, 0.001), # optimizer weight decay + 'warmup_epochs': (1, 0.0, 5.0), # warmup epochs (fractions ok) + 'warmup_momentum': (1, 0.0, 0.95), # warmup initial momentum + 'warmup_bias_lr': (1, 0.0, 0.2), # warmup initial bias lr + 'box': (1, 0.02, 0.2), # box loss gain + 'cls': (1, 0.2, 4.0), # cls loss gain + 'cls_pw': (1, 0.5, 2.0), # cls BCELoss positive_weight + 'obj': (1, 0.2, 4.0), # obj loss gain (scale with pixels) + 'obj_pw': (1, 0.5, 2.0), # obj BCELoss positive_weight + 'iou_t': (0, 0.1, 0.7), # IoU training threshold + 'anchor_t': (1, 2.0, 8.0), # anchor-multiple threshold + 'anchors': (2, 2.0, 10.0), # anchors per output grid (0 to ignore) + 'fl_gamma': (0, 0.0, 2.0), # focal loss gamma (efficientDet default gamma=1.5) + 'hsv_h': (1, 0.0, 0.1), # image HSV-Hue augmentation (fraction) + 'hsv_s': (1, 0.0, 0.9), # image HSV-Saturation augmentation (fraction) + 'hsv_v': (1, 0.0, 0.9), # image HSV-Value augmentation (fraction) + 'degrees': (1, 0.0, 45.0), # image rotation (+/- deg) + 'translate': (1, 0.0, 0.9), # image translation (+/- fraction) + 'scale': (1, 0.0, 0.9), # image scale (+/- gain) + 'shear': (1, 0.0, 10.0), # image shear (+/- deg) + 'perspective': (0, 0.0, 0.001), # image perspective (+/- fraction), range 0-0.001 + 'flipud': (1, 0.0, 1.0), # image flip up-down (probability) + 'fliplr': (0, 0.0, 1.0), # image flip left-right (probability) + 'mosaic': (1, 0.0, 1.0), # image mixup (probability) + 'mixup': (1, 0.0, 1.0), # image mixup (probability) + 'copy_paste': (1, 0.0, 1.0)} # segment copy-paste (probability) + + with open(opt.hyp, errors='ignore') as f: + hyp = yaml.safe_load(f) # load hyps dict + if 'anchors' not in hyp: # anchors commented in hyp.yaml + hyp['anchors'] = 3 + opt.noval, opt.nosave, save_dir = True, True, Path(opt.save_dir) # only val/save final epoch + # ei = [isinstance(x, (int, float)) for x in hyp.values()] # evolvable indices + evolve_yaml, evolve_csv = save_dir / 'hyp_evolve.yaml', save_dir / 'evolve.csv' + if opt.bucket: + os.system(f'gsutil cp gs://{opt.bucket}/evolve.csv {evolve_csv}') # download evolve.csv if exists + + for _ in range(opt.evolve): # generations to evolve + if evolve_csv.exists(): # if evolve.csv exists: select best hyps and mutate + # Select parent(s) + parent = 'single' # parent selection method: 'single' or 'weighted' + x = np.loadtxt(evolve_csv, ndmin=2, delimiter=',', skiprows=1) + n = min(5, len(x)) # number of previous results to consider + x = x[np.argsort(-fitness(x))][:n] # top n mutations + w = fitness(x) - fitness(x).min() + 1E-6 # weights (sum > 0) + if parent == 'single' or len(x) == 1: + # x = x[random.randint(0, n - 1)] # random selection + x = x[random.choices(range(n), weights=w)[0]] # weighted selection + elif parent == 'weighted': + x = (x * w.reshape(n, 1)).sum(0) / w.sum() # weighted combination + + # Mutate + mp, s = 0.8, 0.2 # mutation probability, sigma + npr = np.random + npr.seed(int(time.time())) + g = np.array([meta[k][0] for k in hyp.keys()]) # gains 0-1 + ng = len(meta) + v = np.ones(ng) + while all(v == 1): # mutate until a change occurs (prevent duplicates) + v = (g * (npr.random(ng) < mp) * npr.randn(ng) * npr.random() * s + 1).clip(0.3, 3.0) + for i, k in enumerate(hyp.keys()): # plt.hist(v.ravel(), 300) + hyp[k] = float(x[i + 7] * v[i]) # mutate + + # Constrain to limits + for k, v in meta.items(): + hyp[k] = max(hyp[k], v[1]) # lower limit + hyp[k] = min(hyp[k], v[2]) # upper limit + hyp[k] = round(hyp[k], 5) # significant digits + + # Train mutation + results = train(hyp.copy(), opt, device, callbacks) + callbacks = Callbacks() + # Write mutation results + print_mutation(results, hyp.copy(), save_dir, opt.bucket) + + # Plot results + plot_evolve(evolve_csv) + LOGGER.info(f'Hyperparameter evolution finished {opt.evolve} generations\n' + f"Results saved to {colorstr('bold', save_dir)}\n" + f'Usage example: $ python train.py --hyp {evolve_yaml}') + + +def run(**kwargs): + # Usage: import train; train.run(data='coco128.yaml', imgsz=320, weights='yolov5m.pt') + opt = parse_opt(True) + for k, v in kwargs.items(): + setattr(opt, k, v) + main(opt) + return opt + + +if __name__ == "__main__": + opt = parse_opt() + main(opt) diff --git a/tutorial.ipynb b/tutorial.ipynb new file mode 100644 index 0000000..09b2b33 --- /dev/null +++ b/tutorial.ipynb @@ -0,0 +1,1102 @@ +{ + "nbformat": 4, + "nbformat_minor": 0, + "metadata": { + "colab": { + "name": "YOLOv5 Tutorial", + "provenance": [], + "collapsed_sections": [], + "include_colab_link": true + }, + "kernelspec": { + "name": "python3", + "display_name": "Python 3" + }, + "accelerator": "GPU", + "widgets": { + "application/vnd.jupyter.widget-state+json": { + "eb95db7cae194218b3fcefb439b6352f": { + "model_module": "@jupyter-widgets/controls", + "model_name": "HBoxModel", + "model_module_version": "1.5.0", + "state": { + "_view_name": "HBoxView", + "_dom_classes": [], + "_model_name": "HBoxModel", + "_view_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_view_count": null, + "_view_module_version": "1.5.0", + "box_style": "", + "layout": "IPY_MODEL_769ecde6f2e64bacb596ce972f8d3d2d", + "_model_module": "@jupyter-widgets/controls", + "children": [ + "IPY_MODEL_384a001876054c93b0af45cd1e960bfe", + "IPY_MODEL_dded0aeae74440f7ba2ffa0beb8dd612", + "IPY_MODEL_5296d28be75740b2892ae421bbec3657" + ] + } + }, + "769ecde6f2e64bacb596ce972f8d3d2d": { + "model_module": "@jupyter-widgets/base", + "model_name": "LayoutModel", + "model_module_version": "1.2.0", + "state": { + "_view_name": "LayoutView", + "grid_template_rows": null, + "right": null, + "justify_content": null, + "_view_module": "@jupyter-widgets/base", + "overflow": null, + "_model_module_version": "1.2.0", + "_view_count": null, + "flex_flow": null, + "width": null, + "min_width": null, + "border": null, + "align_items": null, + "bottom": null, + "_model_module": "@jupyter-widgets/base", + "top": null, + "grid_column": null, + "overflow_y": null, + "overflow_x": null, + "grid_auto_flow": null, + "grid_area": null, + "grid_template_columns": null, + "flex": null, + "_model_name": "LayoutModel", + "justify_items": null, + "grid_row": null, + "max_height": null, + "align_content": null, + "visibility": null, + "align_self": null, + "height": null, + "min_height": null, + "padding": null, + "grid_auto_rows": null, + "grid_gap": null, + "max_width": null, + "order": null, + "_view_module_version": "1.2.0", + "grid_template_areas": null, + "object_position": null, + "object_fit": null, + "grid_auto_columns": null, + "margin": null, + "display": null, + "left": null + } + }, + "384a001876054c93b0af45cd1e960bfe": { + "model_module": "@jupyter-widgets/controls", + "model_name": "HTMLModel", + "model_module_version": "1.5.0", + "state": { + "_view_name": "HTMLView", + "style": "IPY_MODEL_9f09facb2a6c4a7096810d327c8b551c", + "_dom_classes": [], + "description": "", + "_model_name": "HTMLModel", + "placeholder": "​", + "_view_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "value": "100%", + "_view_count": null, + "_view_module_version": "1.5.0", + "description_tooltip": null, + "_model_module": "@jupyter-widgets/controls", + "layout": "IPY_MODEL_25621cff5d16448cb7260e839fd0f543" + } + }, + "dded0aeae74440f7ba2ffa0beb8dd612": { + "model_module": "@jupyter-widgets/controls", + "model_name": "FloatProgressModel", + "model_module_version": "1.5.0", + "state": { + "_view_name": "ProgressView", + "style": "IPY_MODEL_0ce7164fc0c74bb9a2b5c7037375a727", + "_dom_classes": [], + "description": "", + "_model_name": "FloatProgressModel", + "bar_style": "success", + "max": 818322941, + "_view_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "value": 818322941, + "_view_count": null, + "_view_module_version": "1.5.0", + "orientation": "horizontal", + "min": 0, + "description_tooltip": null, + "_model_module": "@jupyter-widgets/controls", + "layout": "IPY_MODEL_c4c4593c10904cb5b8a5724d60c7e181" + } + }, + "5296d28be75740b2892ae421bbec3657": { + "model_module": "@jupyter-widgets/controls", + "model_name": "HTMLModel", + "model_module_version": "1.5.0", + "state": { + "_view_name": "HTMLView", + "style": "IPY_MODEL_473371611126476c88d5d42ec7031ed6", + "_dom_classes": [], + "description": "", + "_model_name": "HTMLModel", + "placeholder": "​", + "_view_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "value": " 780M/780M [00:11<00:00, 91.9MB/s]", + "_view_count": null, + "_view_module_version": "1.5.0", + "description_tooltip": null, + "_model_module": "@jupyter-widgets/controls", + "layout": "IPY_MODEL_65efdfd0d26c46e79c8c5ff3b77126cc" + } + }, + "9f09facb2a6c4a7096810d327c8b551c": { + "model_module": "@jupyter-widgets/controls", + "model_name": "DescriptionStyleModel", + "model_module_version": "1.5.0", + "state": { + "_view_name": "StyleView", + "_model_name": "DescriptionStyleModel", + "description_width": "", + "_view_module": "@jupyter-widgets/base", + "_model_module_version": "1.5.0", + "_view_count": null, + "_view_module_version": "1.2.0", + "_model_module": "@jupyter-widgets/controls" + } + }, + "25621cff5d16448cb7260e839fd0f543": { + "model_module": "@jupyter-widgets/base", + "model_name": "LayoutModel", + "model_module_version": "1.2.0", + "state": { + "_view_name": "LayoutView", + "grid_template_rows": null, + "right": null, + "justify_content": null, + "_view_module": "@jupyter-widgets/base", + "overflow": null, + "_model_module_version": "1.2.0", + "_view_count": null, + "flex_flow": null, + "width": null, + "min_width": null, + "border": null, + "align_items": null, + "bottom": null, + "_model_module": "@jupyter-widgets/base", + "top": null, + "grid_column": null, + "overflow_y": null, + "overflow_x": null, + "grid_auto_flow": null, + "grid_area": null, + "grid_template_columns": null, + "flex": null, + "_model_name": "LayoutModel", + "justify_items": null, + "grid_row": null, + "max_height": null, + "align_content": null, + "visibility": null, + "align_self": null, + "height": null, + "min_height": null, + "padding": null, + "grid_auto_rows": null, + "grid_gap": null, + "max_width": null, + "order": null, + "_view_module_version": "1.2.0", + "grid_template_areas": null, + "object_position": null, + "object_fit": null, + "grid_auto_columns": null, + "margin": null, + "display": null, + "left": null + } + }, + "0ce7164fc0c74bb9a2b5c7037375a727": { + "model_module": "@jupyter-widgets/controls", + "model_name": "ProgressStyleModel", + "model_module_version": "1.5.0", + "state": { + "_view_name": "StyleView", + "_model_name": "ProgressStyleModel", + "description_width": "", + "_view_module": "@jupyter-widgets/base", + "_model_module_version": "1.5.0", + "_view_count": null, + "_view_module_version": "1.2.0", + "bar_color": null, + "_model_module": "@jupyter-widgets/controls" + } + }, + "c4c4593c10904cb5b8a5724d60c7e181": { + "model_module": "@jupyter-widgets/base", + "model_name": "LayoutModel", + "model_module_version": "1.2.0", + "state": { + "_view_name": "LayoutView", + "grid_template_rows": null, + "right": null, + "justify_content": null, + "_view_module": "@jupyter-widgets/base", + "overflow": null, + "_model_module_version": "1.2.0", + "_view_count": null, + "flex_flow": null, + "width": null, + "min_width": null, + "border": null, + "align_items": null, + "bottom": null, + "_model_module": "@jupyter-widgets/base", + "top": null, + "grid_column": null, + "overflow_y": null, + "overflow_x": null, + "grid_auto_flow": null, + "grid_area": null, + "grid_template_columns": null, + "flex": null, + "_model_name": "LayoutModel", + "justify_items": null, + "grid_row": null, + "max_height": null, + "align_content": null, + "visibility": null, + "align_self": null, + "height": null, + "min_height": null, + "padding": null, + "grid_auto_rows": null, + "grid_gap": null, + "max_width": null, + "order": null, + "_view_module_version": "1.2.0", + "grid_template_areas": null, + "object_position": null, + "object_fit": null, + "grid_auto_columns": null, + "margin": null, + "display": null, + "left": null + } + }, + "473371611126476c88d5d42ec7031ed6": { + "model_module": "@jupyter-widgets/controls", + "model_name": "DescriptionStyleModel", + "model_module_version": "1.5.0", + "state": { + "_view_name": "StyleView", + "_model_name": "DescriptionStyleModel", + "description_width": "", + "_view_module": "@jupyter-widgets/base", + "_model_module_version": "1.5.0", + "_view_count": null, + "_view_module_version": "1.2.0", + "_model_module": "@jupyter-widgets/controls" + } + }, + "65efdfd0d26c46e79c8c5ff3b77126cc": { + "model_module": "@jupyter-widgets/base", + "model_name": "LayoutModel", + "model_module_version": "1.2.0", + "state": { + "_view_name": "LayoutView", + "grid_template_rows": null, + "right": null, + "justify_content": null, + "_view_module": "@jupyter-widgets/base", + "overflow": null, + "_model_module_version": "1.2.0", + "_view_count": null, + "flex_flow": null, + "width": null, + "min_width": null, + "border": null, + "align_items": null, + "bottom": null, + "_model_module": "@jupyter-widgets/base", + "top": null, + "grid_column": null, + "overflow_y": null, + "overflow_x": null, + "grid_auto_flow": null, + "grid_area": null, + "grid_template_columns": null, + "flex": null, + "_model_name": "LayoutModel", + "justify_items": null, + "grid_row": null, + "max_height": null, + "align_content": null, + "visibility": null, + "align_self": null, + "height": null, + "min_height": null, + "padding": null, + "grid_auto_rows": null, + "grid_gap": null, + "max_width": null, + "order": null, + "_view_module_version": "1.2.0", + "grid_template_areas": null, + "object_position": null, + "object_fit": null, + "grid_auto_columns": null, + "margin": null, + "display": null, + "left": null + } + } + } + } + }, + "cells": [ + { + "cell_type": "markdown", + "metadata": { + "id": "view-in-github", + "colab_type": "text" + }, + "source": [ + "\"Open" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "t6MPjfT5NrKQ" + }, + "source": [ + "\n", + "\n", + "\n", + "This is the **official YOLOv5 🚀 notebook** by **Ultralytics**, and is freely available for redistribution under the [GPL-3.0 license](https://choosealicense.com/licenses/gpl-3.0/). \n", + "For more information please visit https://github.com/ultralytics/yolov5 and https://ultralytics.com. Thank you!" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "7mGmQbAO5pQb" + }, + "source": [ + "# Setup\n", + "\n", + "Clone repo, install dependencies and check PyTorch and GPU." + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "wbvMlHd_QwMG", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "3809e5a9-dd41-4577-fe62-5531abf7cca2" + }, + "source": [ + "!git clone https://github.com/ultralytics/yolov5 # clone\n", + "%cd yolov5\n", + "%pip install -qr requirements.txt # install\n", + "\n", + "import torch\n", + "from yolov5 import utils\n", + "display = utils.notebook_init() # checks" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "name": "stdout", + "text": [ + "YOLOv5 🚀 v6.0-48-g84a8099 torch 1.10.0+cu102 CUDA:0 (Tesla V100-SXM2-16GB, 16160MiB)\n", + "Setup complete ✅\n" + ] + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "4JnkELT0cIJg" + }, + "source": [ + "# 1. Inference\n", + "\n", + "`detect.py` runs YOLOv5 inference on a variety of sources, downloading models automatically from the [latest YOLOv5 release](https://github.com/ultralytics/yolov5/releases), and saving results to `runs/detect`. Example inference sources are:\n", + "\n", + "```shell\n", + "python detect.py --source 0 # webcam\n", + " img.jpg # image \n", + " vid.mp4 # video\n", + " path/ # directory\n", + " path/*.jpg # glob\n", + " 'https://youtu.be/Zgi9g1ksQHc' # YouTube\n", + " 'rtsp://example.com/media.mp4' # RTSP, RTMP, HTTP stream\n", + "```" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "zR9ZbuQCH7FX", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "8f7e6588-215d-4ebd-93af-88b871e770a7" + }, + "source": [ + "!python detect.py --weights yolov5s.pt --img 640 --conf 0.25 --source data/images\n", + "display.Image(filename='runs/detect/exp/zidane.jpg', width=600)" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "name": "stdout", + "text": [ + "\u001b[34m\u001b[1mdetect: \u001b[0mweights=['yolov5s.pt'], source=data/images, imgsz=[640, 640], conf_thres=0.25, iou_thres=0.45, max_det=1000, device=, view_img=False, save_txt=False, save_conf=False, save_crop=False, nosave=False, classes=None, agnostic_nms=False, augment=False, visualize=False, update=False, project=runs/detect, name=exp, exist_ok=False, line_thickness=3, hide_labels=False, hide_conf=False, half=False, dnn=False\n", + "YOLOv5 🚀 v6.0-48-g84a8099 torch 1.10.0+cu102 CUDA:0 (Tesla V100-SXM2-16GB, 16160MiB)\n", + "\n", + "Fusing layers... \n", + "Model Summary: 213 layers, 7225885 parameters, 0 gradients\n", + "image 1/2 /content/yolov5/data/images/bus.jpg: 640x480 4 persons, 1 bus, Done. (0.007s)\n", + "image 2/2 /content/yolov5/data/images/zidane.jpg: 384x640 2 persons, 1 tie, Done. (0.007s)\n", + "Speed: 0.5ms pre-process, 6.9ms inference, 1.3ms NMS per image at shape (1, 3, 640, 640)\n", + "Results saved to \u001b[1mruns/detect/exp\u001b[0m\n" + ] + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "hkAzDWJ7cWTr" + }, + "source": [ + "        \n", + "" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "0eq1SMWl6Sfn" + }, + "source": [ + "# 2. Validate\n", + "Validate a model's accuracy on [COCO](https://cocodataset.org/#home) val or test-dev datasets. Models are downloaded automatically from the [latest YOLOv5 release](https://github.com/ultralytics/yolov5/releases). To show results by class use the `--verbose` flag. Note that `pycocotools` metrics may be ~1% better than the equivalent repo metrics, as is visible below, due to slight differences in mAP computation." + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "eyTZYGgRjnMc" + }, + "source": [ + "## COCO val\n", + "Download [COCO val 2017](https://github.com/ultralytics/yolov5/blob/74b34872fdf41941cddcf243951cdb090fbac17b/data/coco.yaml#L14) dataset (1GB - 5000 images), and test model accuracy." + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "WQPtK1QYVaD_", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 48, + "referenced_widgets": [ + "eb95db7cae194218b3fcefb439b6352f", + "769ecde6f2e64bacb596ce972f8d3d2d", + "384a001876054c93b0af45cd1e960bfe", + "dded0aeae74440f7ba2ffa0beb8dd612", + "5296d28be75740b2892ae421bbec3657", + "9f09facb2a6c4a7096810d327c8b551c", + "25621cff5d16448cb7260e839fd0f543", + "0ce7164fc0c74bb9a2b5c7037375a727", + "c4c4593c10904cb5b8a5724d60c7e181", + "473371611126476c88d5d42ec7031ed6", + "65efdfd0d26c46e79c8c5ff3b77126cc" + ] + }, + "outputId": "bcf9a448-1f9b-4a41-ad49-12f181faf05a" + }, + "source": [ + "# Download COCO val\n", + "torch.hub.download_url_to_file('https://ultralytics.com/assets/coco2017val.zip', 'tmp.zip')\n", + "!unzip -q tmp.zip -d ../datasets && rm tmp.zip" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "display_data", + "data": { + "application/vnd.jupyter.widget-view+json": { + "model_id": "eb95db7cae194218b3fcefb439b6352f", + "version_minor": 0, + "version_major": 2 + }, + "text/plain": [ + " 0%| | 0.00/780M [00:00

\n", + "Close the active learning loop by sampling images from your inference conditions with the `roboflow` pip package\n", + "

\n", + "\n", + "Train a YOLOv5s model on the [COCO128](https://www.kaggle.com/ultralytics/coco128) dataset with `--data coco128.yaml`, starting from pretrained `--weights yolov5s.pt`, or from randomly initialized `--weights '' --cfg yolov5s.yaml`.\n", + "\n", + "- **Pretrained [Models](https://github.com/ultralytics/yolov5/tree/master/models)** are downloaded\n", + "automatically from the [latest YOLOv5 release](https://github.com/ultralytics/yolov5/releases)\n", + "- **[Datasets](https://github.com/ultralytics/yolov5/tree/master/data)** available for autodownload include: [COCO](https://github.com/ultralytics/yolov5/blob/master/data/coco.yaml), [COCO128](https://github.com/ultralytics/yolov5/blob/master/data/coco128.yaml), [VOC](https://github.com/ultralytics/yolov5/blob/master/data/VOC.yaml), [Argoverse](https://github.com/ultralytics/yolov5/blob/master/data/Argoverse.yaml), [VisDrone](https://github.com/ultralytics/yolov5/blob/master/data/VisDrone.yaml), [GlobalWheat](https://github.com/ultralytics/yolov5/blob/master/data/GlobalWheat2020.yaml), [xView](https://github.com/ultralytics/yolov5/blob/master/data/xView.yaml), [Objects365](https://github.com/ultralytics/yolov5/blob/master/data/Objects365.yaml), [SKU-110K](https://github.com/ultralytics/yolov5/blob/master/data/SKU-110K.yaml).\n", + "- **Training Results** are saved to `runs/train/` with incrementing run directories, i.e. `runs/train/exp2`, `runs/train/exp3` etc.\n", + "

\n", + "\n", + "## Train on Custom Data with Roboflow 🌟 NEW\n", + "\n", + "[Roboflow](https://roboflow.com/?ref=ultralytics) enables you to easily **organize, label, and prepare** a high quality dataset with your own custom data. Roboflow also makes it easy to establish an active learning pipeline, collaborate with your team on dataset improvement, and integrate directly into your model building workflow with the `roboflow` pip package.\n", + "\n", + "- Custom Training Example: [https://blog.roboflow.com/how-to-train-yolov5-on-a-custom-dataset/](https://blog.roboflow.com/how-to-train-yolov5-on-a-custom-dataset/?ref=ultralytics)\n", + "- Custom Training Notebook: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/roboflow-ai/yolov5-custom-training-tutorial/blob/main/yolov5-custom-training.ipynb)\n", + "
\n", + "\n", + "

Label images lightning fast (including with model-assisted labeling)" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "bOy5KI2ncnWd" + }, + "source": [ + "# Tensorboard (optional)\n", + "%load_ext tensorboard\n", + "%tensorboard --logdir runs/train" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "2fLAV42oNb7M" + }, + "source": [ + "# Weights & Biases (optional)\n", + "%pip install -q wandb\n", + "import wandb\n", + "wandb.login()" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "1NcFxRcFdJ_O", + "colab": { + "base_uri": "https://localhost:8080/" + }, + "outputId": "8724d13d-6711-4a12-d96a-1c655e5c3549" + }, + "source": [ + "# Train YOLOv5s on COCO128 for 3 epochs\n", + "!python train.py --img 640 --batch 16 --epochs 3 --data coco128.yaml --weights yolov5s.pt --cache" + ], + "execution_count": null, + "outputs": [ + { + "output_type": "stream", + "name": "stdout", + "text": [ + "\u001b[34m\u001b[1mtrain: \u001b[0mweights=yolov5s.pt, cfg=, data=coco128.yaml, hyp=data/hyps/hyp.scratch.yaml, epochs=3, batch_size=16, imgsz=640, rect=False, resume=False, nosave=False, noval=False, noautoanchor=False, evolve=None, bucket=, cache=ram, image_weights=False, device=, multi_scale=False, single_cls=False, adam=False, sync_bn=False, workers=8, project=runs/train, name=exp, exist_ok=False, quad=False, linear_lr=False, label_smoothing=0.0, patience=100, freeze=0, save_period=-1, local_rank=-1, entity=None, upload_dataset=False, bbox_interval=-1, artifact_alias=latest\n", + "\u001b[34m\u001b[1mgithub: \u001b[0mup to date with https://github.com/ultralytics/yolov5 ✅\n", + "YOLOv5 🚀 v6.0-48-g84a8099 torch 1.10.0+cu102 CUDA:0 (Tesla V100-SXM2-16GB, 16160MiB)\n", + "\n", + "\u001b[34m\u001b[1mhyperparameters: \u001b[0mlr0=0.01, lrf=0.1, momentum=0.937, weight_decay=0.0005, warmup_epochs=3.0, warmup_momentum=0.8, warmup_bias_lr=0.1, box=0.05, cls=0.5, cls_pw=1.0, obj=1.0, obj_pw=1.0, iou_t=0.2, anchor_t=4.0, fl_gamma=0.0, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, degrees=0.0, translate=0.1, scale=0.5, shear=0.0, perspective=0.0, flipud=0.0, fliplr=0.5, mosaic=1.0, mixup=0.0, copy_paste=0.0\n", + "\u001b[34m\u001b[1mWeights & Biases: \u001b[0mrun 'pip install wandb' to automatically track and visualize YOLOv5 🚀 runs (RECOMMENDED)\n", + "\u001b[34m\u001b[1mTensorBoard: \u001b[0mStart with 'tensorboard --logdir runs/train', view at http://localhost:6006/\n", + "\n", + " from n params module arguments \n", + " 0 -1 1 3520 models.common.Conv [3, 32, 6, 2, 2] \n", + " 1 -1 1 18560 models.common.Conv [32, 64, 3, 2] \n", + " 2 -1 1 18816 models.common.C3 [64, 64, 1] \n", + " 3 -1 1 73984 models.common.Conv [64, 128, 3, 2] \n", + " 4 -1 2 115712 models.common.C3 [128, 128, 2] \n", + " 5 -1 1 295424 models.common.Conv [128, 256, 3, 2] \n", + " 6 -1 3 625152 models.common.C3 [256, 256, 3] \n", + " 7 -1 1 1180672 models.common.Conv [256, 512, 3, 2] \n", + " 8 -1 1 1182720 models.common.C3 [512, 512, 1] \n", + " 9 -1 1 656896 models.common.SPPF [512, 512, 5] \n", + " 10 -1 1 131584 models.common.Conv [512, 256, 1, 1] \n", + " 11 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest'] \n", + " 12 [-1, 6] 1 0 models.common.Concat [1] \n", + " 13 -1 1 361984 models.common.C3 [512, 256, 1, False] \n", + " 14 -1 1 33024 models.common.Conv [256, 128, 1, 1] \n", + " 15 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest'] \n", + " 16 [-1, 4] 1 0 models.common.Concat [1] \n", + " 17 -1 1 90880 models.common.C3 [256, 128, 1, False] \n", + " 18 -1 1 147712 models.common.Conv [128, 128, 3, 2] \n", + " 19 [-1, 14] 1 0 models.common.Concat [1] \n", + " 20 -1 1 296448 models.common.C3 [256, 256, 1, False] \n", + " 21 -1 1 590336 models.common.Conv [256, 256, 3, 2] \n", + " 22 [-1, 10] 1 0 models.common.Concat [1] \n", + " 23 -1 1 1182720 models.common.C3 [512, 512, 1, False] \n", + " 24 [17, 20, 23] 1 229245 models.yolo.Detect [80, [[10, 13, 16, 30, 33, 23], [30, 61, 62, 45, 59, 119], [116, 90, 156, 198, 373, 326]], [128, 256, 512]]\n", + "Model Summary: 270 layers, 7235389 parameters, 7235389 gradients, 16.5 GFLOPs\n", + "\n", + "Transferred 349/349 items from yolov5s.pt\n", + "Scaled weight_decay = 0.0005\n", + "\u001b[34m\u001b[1moptimizer:\u001b[0m SGD with parameter groups 57 weight, 60 weight (no decay), 60 bias\n", + "\u001b[34m\u001b[1malbumentations: \u001b[0mversion 1.0.3 required by YOLOv5, but version 0.1.12 is currently installed\n", + "\u001b[34m\u001b[1mtrain: \u001b[0mScanning '../datasets/coco128/labels/train2017.cache' images and labels... 128 found, 0 missing, 2 empty, 0 corrupted: 100% 128/128 [00:00\"Weights

" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "-WPvRbS5Swl6" + }, + "source": [ + "## Local Logging\n", + "\n", + "All results are logged by default to `runs/train`, with a new experiment directory created for each new training as `runs/train/exp2`, `runs/train/exp3`, etc. View train and val jpgs to see mosaics, labels, predictions and augmentation effects. Note an Ultralytics **Mosaic Dataloader** is used for training (shown below), which combines 4 images into 1 mosaic during training.\n", + "\n", + "> \n", + "`train_batch0.jpg` shows train batch 0 mosaics and labels\n", + "\n", + "> \n", + "`test_batch0_labels.jpg` shows val batch 0 labels\n", + "\n", + "> \n", + "`test_batch0_pred.jpg` shows val batch 0 _predictions_\n", + "\n", + "Training results are automatically logged to [Tensorboard](https://www.tensorflow.org/tensorboard) and [CSV](https://github.com/ultralytics/yolov5/pull/4148) as `results.csv`, which is plotted as `results.png` (below) after training completes. You can also plot any `results.csv` file manually:\n", + "\n", + "```python\n", + "from utils.plots import plot_results \n", + "plot_results('path/to/results.csv') # plot 'results.csv' as 'results.png'\n", + "```\n", + "\n", + "\"COCO128" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "Zelyeqbyt3GD" + }, + "source": [ + "# Environments\n", + "\n", + "YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including [CUDA](https://developer.nvidia.com/cuda)/[CUDNN](https://developer.nvidia.com/cudnn), [Python](https://www.python.org/) and [PyTorch](https://pytorch.org/) preinstalled):\n", + "\n", + "- **Google Colab and Kaggle** notebooks with free GPU: \"Open \"Open\n", + "- **Google Cloud** Deep Learning VM. See [GCP Quickstart Guide](https://github.com/ultralytics/yolov5/wiki/GCP-Quickstart)\n", + "- **Amazon** Deep Learning AMI. See [AWS Quickstart Guide](https://github.com/ultralytics/yolov5/wiki/AWS-Quickstart)\n", + "- **Docker Image**. See [Docker Quickstart Guide](https://github.com/ultralytics/yolov5/wiki/Docker-Quickstart) \"Docker\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "6Qu7Iesl0p54" + }, + "source": [ + "# Status\n", + "\n", + "![CI CPU testing](https://github.com/ultralytics/yolov5/workflows/CI%20CPU%20testing/badge.svg)\n", + "\n", + "If this badge is green, all [YOLOv5 GitHub Actions](https://github.com/ultralytics/yolov5/actions) Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training ([train.py](https://github.com/ultralytics/yolov5/blob/master/train.py)), testing ([val.py](https://github.com/ultralytics/yolov5/blob/master/val.py)), inference ([detect.py](https://github.com/ultralytics/yolov5/blob/master/detect.py)) and export ([export.py](https://github.com/ultralytics/yolov5/blob/master/export.py)) on MacOS, Windows, and Ubuntu every 24 hours and on every commit.\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "IEijrePND_2I" + }, + "source": [ + "# Appendix\n", + "\n", + "Optional extras below. Unit tests validate repo functionality and should be run on any PRs submitted.\n" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "mcKoSIK2WSzj" + }, + "source": [ + "# Reproduce\n", + "for x in 'yolov5n', 'yolov5s', 'yolov5m', 'yolov5l', 'yolov5x':\n", + " !python val.py --weights {x}.pt --data coco.yaml --img 640 --task speed # speed\n", + " !python val.py --weights {x}.pt --data coco.yaml --img 640 --conf 0.001 --iou 0.65 # mAP" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "GMusP4OAxFu6" + }, + "source": [ + "# PyTorch Hub\n", + "import torch\n", + "\n", + "# Model\n", + "model = torch.hub.load('ultralytics/yolov5', 'yolov5s')\n", + "\n", + "# Images\n", + "dir = 'https://ultralytics.com/images/'\n", + "imgs = [dir + f for f in ('zidane.jpg', 'bus.jpg')] # batch of images\n", + "\n", + "# Inference\n", + "results = model(imgs)\n", + "results.print() # or .show(), .save()" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "FGH0ZjkGjejy" + }, + "source": [ + "# CI Checks\n", + "%%shell\n", + "export PYTHONPATH=\"$PWD\" # to run *.py. files in subdirectories\n", + "rm -rf runs # remove runs/\n", + "for m in yolov5n; do # models\n", + " python train.py --img 64 --batch 32 --weights $m.pt --epochs 1 --device 0 # train pretrained\n", + " python train.py --img 64 --batch 32 --weights '' --cfg $m.yaml --epochs 1 --device 0 # train scratch\n", + " for d in 0 cpu; do # devices\n", + " python val.py --weights $m.pt --device $d # val official\n", + " python val.py --weights runs/train/exp/weights/best.pt --device $d # val custom\n", + " python detect.py --weights $m.pt --device $d # detect official\n", + " python detect.py --weights runs/train/exp/weights/best.pt --device $d # detect custom\n", + " done\n", + " python hubconf.py # hub\n", + " python models/yolo.py --cfg $m.yaml # build PyTorch model\n", + " python models/tf.py --weights $m.pt # build TensorFlow model\n", + " python export.py --img 64 --batch 1 --weights $m.pt --include torchscript onnx # export\n", + "done" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "gogI-kwi3Tye" + }, + "source": [ + "# Profile\n", + "from utils.torch_utils import profile\n", + "\n", + "m1 = lambda x: x * torch.sigmoid(x)\n", + "m2 = torch.nn.SiLU()\n", + "results = profile(input=torch.randn(16, 3, 640, 640), ops=[m1, m2], n=100)" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "RVRSOhEvUdb5" + }, + "source": [ + "# Evolve\n", + "!python train.py --img 640 --batch 64 --epochs 100 --data coco128.yaml --weights yolov5s.pt --cache --noautoanchor --evolve\n", + "!d=runs/train/evolve && cp evolve.* $d && zip -r evolve.zip $d && gsutil mv evolve.zip gs://bucket # upload results (optional)" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "BSgFCAcMbk1R" + }, + "source": [ + "# VOC\n", + "for b, m in zip([64, 64, 32, 16], ['yolov5s', 'yolov5m', 'yolov5l', 'yolov5x']): # zip(batch_size, model)\n", + " !python train.py --batch {b} --weights {m}.pt --data VOC.yaml --epochs 50 --cache --img 512 --nosave --hyp hyp.finetune.yaml --project VOC --name {m}" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "VTRwsvA9u7ln" + }, + "source": [ + "# TensorRT \n", + "# https://docs.nvidia.com/deeplearning/tensorrt/install-guide/index.html#installing-pip\n", + "!pip install -U nvidia-tensorrt --index-url https://pypi.ngc.nvidia.com # install\n", + "!python export.py --weights yolov5s.pt --include engine --imgsz 640 640 --device 0 # export\n", + "!python detect.py --weights yolov5s.engine --imgsz 640 640 --device 0 # inference" + ], + "execution_count": null, + "outputs": [] + } + ] +} diff --git a/utils/__init__.py b/utils/__init__.py new file mode 100644 index 0000000..295aebf --- /dev/null +++ b/utils/__init__.py @@ -0,0 +1,37 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +utils/initialization +""" + + +def notebook_init(verbose=True): + # Check system software and hardware + print('Checking setup...') + + import os + import shutil + + from utils.general import check_requirements, emojis, is_colab + from utils.torch_utils import select_device # imports + + check_requirements(('psutil', 'IPython')) + import psutil + from IPython import display # to display images and clear console output + + if is_colab(): + shutil.rmtree('/content/sample_data', ignore_errors=True) # remove colab /sample_data directory + + if verbose: + # System info + # gb = 1 / 1000 ** 3 # bytes to GB + gib = 1 / 1024 ** 3 # bytes to GiB + ram = psutil.virtual_memory().total + total, used, free = shutil.disk_usage("/") + display.clear_output() + s = f'({os.cpu_count()} CPUs, {ram * gib:.1f} GB RAM, {(total - free) * gib:.1f}/{total * gib:.1f} GB disk)' + else: + s = '' + + select_device(newline=False) + print(emojis(f'Setup complete ✅ {s}')) + return display diff --git a/utils/activations.py b/utils/activations.py new file mode 100644 index 0000000..a4ff789 --- /dev/null +++ b/utils/activations.py @@ -0,0 +1,101 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Activation functions +""" + +import torch +import torch.nn as nn +import torch.nn.functional as F + + +# SiLU https://arxiv.org/pdf/1606.08415.pdf ---------------------------------------------------------------------------- +class SiLU(nn.Module): # export-friendly version of nn.SiLU() + @staticmethod + def forward(x): + return x * torch.sigmoid(x) + + +class Hardswish(nn.Module): # export-friendly version of nn.Hardswish() + @staticmethod + def forward(x): + # return x * F.hardsigmoid(x) # for TorchScript and CoreML + return x * F.hardtanh(x + 3, 0.0, 6.0) / 6.0 # for TorchScript, CoreML and ONNX + + +# Mish https://github.com/digantamisra98/Mish -------------------------------------------------------------------------- +class Mish(nn.Module): + @staticmethod + def forward(x): + return x * F.softplus(x).tanh() + + +class MemoryEfficientMish(nn.Module): + class F(torch.autograd.Function): + @staticmethod + def forward(ctx, x): + ctx.save_for_backward(x) + return x.mul(torch.tanh(F.softplus(x))) # x * tanh(ln(1 + exp(x))) + + @staticmethod + def backward(ctx, grad_output): + x = ctx.saved_tensors[0] + sx = torch.sigmoid(x) + fx = F.softplus(x).tanh() + return grad_output * (fx + x * sx * (1 - fx * fx)) + + def forward(self, x): + return self.F.apply(x) + + +# FReLU https://arxiv.org/abs/2007.11824 ------------------------------------------------------------------------------- +class FReLU(nn.Module): + def __init__(self, c1, k=3): # ch_in, kernel + super().__init__() + self.conv = nn.Conv2d(c1, c1, k, 1, 1, groups=c1, bias=False) + self.bn = nn.BatchNorm2d(c1) + + def forward(self, x): + return torch.max(x, self.bn(self.conv(x))) + + +# ACON https://arxiv.org/pdf/2009.04759.pdf ---------------------------------------------------------------------------- +class AconC(nn.Module): + r""" ACON activation (activate or not). + AconC: (p1*x-p2*x) * sigmoid(beta*(p1*x-p2*x)) + p2*x, beta is a learnable parameter + according to "Activate or Not: Learning Customized Activation" . + """ + + def __init__(self, c1): + super().__init__() + self.p1 = nn.Parameter(torch.randn(1, c1, 1, 1)) + self.p2 = nn.Parameter(torch.randn(1, c1, 1, 1)) + self.beta = nn.Parameter(torch.ones(1, c1, 1, 1)) + + def forward(self, x): + dpx = (self.p1 - self.p2) * x + return dpx * torch.sigmoid(self.beta * dpx) + self.p2 * x + + +class MetaAconC(nn.Module): + r""" ACON activation (activate or not). + MetaAconC: (p1*x-p2*x) * sigmoid(beta*(p1*x-p2*x)) + p2*x, beta is generated by a small network + according to "Activate or Not: Learning Customized Activation" . + """ + + def __init__(self, c1, k=1, s=1, r=16): # ch_in, kernel, stride, r + super().__init__() + c2 = max(r, c1 // r) + self.p1 = nn.Parameter(torch.randn(1, c1, 1, 1)) + self.p2 = nn.Parameter(torch.randn(1, c1, 1, 1)) + self.fc1 = nn.Conv2d(c1, c2, k, s, bias=True) + self.fc2 = nn.Conv2d(c2, c1, k, s, bias=True) + # self.bn1 = nn.BatchNorm2d(c2) + # self.bn2 = nn.BatchNorm2d(c1) + + def forward(self, x): + y = x.mean(dim=2, keepdims=True).mean(dim=3, keepdims=True) + # batch-size 1 bug/instabilities https://github.com/ultralytics/yolov5/issues/2891 + # beta = torch.sigmoid(self.bn2(self.fc2(self.bn1(self.fc1(y))))) # bug/unstable + beta = torch.sigmoid(self.fc2(self.fc1(y))) # bug patch BN layers removed + dpx = (self.p1 - self.p2) * x + return dpx * torch.sigmoid(beta * dpx) + self.p2 * x diff --git a/utils/augmentations.py b/utils/augmentations.py new file mode 100644 index 0000000..0311b97 --- /dev/null +++ b/utils/augmentations.py @@ -0,0 +1,277 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Image augmentation functions +""" + +import math +import random + +import cv2 +import numpy as np + +from utils.general import LOGGER, check_version, colorstr, resample_segments, segment2box +from utils.metrics import bbox_ioa + + +class Albumentations: + # YOLOv5 Albumentations class (optional, only used if package is installed) + def __init__(self): + self.transform = None + try: + import albumentations as A + check_version(A.__version__, '1.0.3', hard=True) # version requirement + + self.transform = A.Compose([ + A.Blur(p=0.01), + A.MedianBlur(p=0.01), + A.ToGray(p=0.01), + A.CLAHE(p=0.01), + A.RandomBrightnessContrast(p=0.0), + A.RandomGamma(p=0.0), + A.ImageCompression(quality_lower=75, p=0.0)], + bbox_params=A.BboxParams(format='yolo', label_fields=['class_labels'])) + + LOGGER.info(colorstr('albumentations: ') + ', '.join(f'{x}' for x in self.transform.transforms if x.p)) + except ImportError: # package not installed, skip + pass + except Exception as e: + LOGGER.info(colorstr('albumentations: ') + f'{e}') + + def __call__(self, im, labels, p=1.0): + if self.transform and random.random() < p: + new = self.transform(image=im, bboxes=labels[:, 1:], class_labels=labels[:, 0]) # transformed + im, labels = new['image'], np.array([[c, *b] for c, b in zip(new['class_labels'], new['bboxes'])]) + return im, labels + + +def augment_hsv(im, hgain=0.5, sgain=0.5, vgain=0.5): + # HSV color-space augmentation + if hgain or sgain or vgain: + r = np.random.uniform(-1, 1, 3) * [hgain, sgain, vgain] + 1 # random gains + hue, sat, val = cv2.split(cv2.cvtColor(im, cv2.COLOR_BGR2HSV)) + dtype = im.dtype # uint8 + + x = np.arange(0, 256, dtype=r.dtype) + lut_hue = ((x * r[0]) % 180).astype(dtype) + lut_sat = np.clip(x * r[1], 0, 255).astype(dtype) + lut_val = np.clip(x * r[2], 0, 255).astype(dtype) + + im_hsv = cv2.merge((cv2.LUT(hue, lut_hue), cv2.LUT(sat, lut_sat), cv2.LUT(val, lut_val))) + cv2.cvtColor(im_hsv, cv2.COLOR_HSV2BGR, dst=im) # no return needed + + +def hist_equalize(im, clahe=True, bgr=False): + # Equalize histogram on BGR image 'im' with im.shape(n,m,3) and range 0-255 + yuv = cv2.cvtColor(im, cv2.COLOR_BGR2YUV if bgr else cv2.COLOR_RGB2YUV) + if clahe: + c = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8)) + yuv[:, :, 0] = c.apply(yuv[:, :, 0]) + else: + yuv[:, :, 0] = cv2.equalizeHist(yuv[:, :, 0]) # equalize Y channel histogram + return cv2.cvtColor(yuv, cv2.COLOR_YUV2BGR if bgr else cv2.COLOR_YUV2RGB) # convert YUV image to RGB + + +def replicate(im, labels): + # Replicate labels + h, w = im.shape[:2] + boxes = labels[:, 1:].astype(int) + x1, y1, x2, y2 = boxes.T + s = ((x2 - x1) + (y2 - y1)) / 2 # side length (pixels) + for i in s.argsort()[:round(s.size * 0.5)]: # smallest indices + x1b, y1b, x2b, y2b = boxes[i] + bh, bw = y2b - y1b, x2b - x1b + yc, xc = int(random.uniform(0, h - bh)), int(random.uniform(0, w - bw)) # offset x, y + x1a, y1a, x2a, y2a = [xc, yc, xc + bw, yc + bh] + im[y1a:y2a, x1a:x2a] = im[y1b:y2b, x1b:x2b] # im4[ymin:ymax, xmin:xmax] + labels = np.append(labels, [[labels[i, 0], x1a, y1a, x2a, y2a]], axis=0) + + return im, labels + + +def letterbox(im, new_shape=(640, 640), color=(114, 114, 114), auto=True, scaleFill=False, scaleup=True, stride=32): + # Resize and pad image while meeting stride-multiple constraints + shape = im.shape[:2] # current shape [height, width] + if isinstance(new_shape, int): + new_shape = (new_shape, new_shape) + + # Scale ratio (new / old) + r = min(new_shape[0] / shape[0], new_shape[1] / shape[1]) + if not scaleup: # only scale down, do not scale up (for better val mAP) + r = min(r, 1.0) + + # Compute padding + ratio = r, r # width, height ratios + new_unpad = int(round(shape[1] * r)), int(round(shape[0] * r)) + dw, dh = new_shape[1] - new_unpad[0], new_shape[0] - new_unpad[1] # wh padding + if auto: # minimum rectangle + dw, dh = np.mod(dw, stride), np.mod(dh, stride) # wh padding + elif scaleFill: # stretch + dw, dh = 0.0, 0.0 + new_unpad = (new_shape[1], new_shape[0]) + ratio = new_shape[1] / shape[1], new_shape[0] / shape[0] # width, height ratios + + dw /= 2 # divide padding into 2 sides + dh /= 2 + + if shape[::-1] != new_unpad: # resize + im = cv2.resize(im, new_unpad, interpolation=cv2.INTER_LINEAR) + top, bottom = int(round(dh - 0.1)), int(round(dh + 0.1)) + left, right = int(round(dw - 0.1)), int(round(dw + 0.1)) + im = cv2.copyMakeBorder(im, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color) # add border + return im, ratio, (dw, dh) + + +def random_perspective(im, targets=(), segments=(), degrees=10, translate=.1, scale=.1, shear=10, perspective=0.0, + border=(0, 0)): + # torchvision.transforms.RandomAffine(degrees=(-10, 10), translate=(0.1, 0.1), scale=(0.9, 1.1), shear=(-10, 10)) + # targets = [cls, xyxy] + + height = im.shape[0] + border[0] * 2 # shape(h,w,c) + width = im.shape[1] + border[1] * 2 + + # Center + C = np.eye(3) + C[0, 2] = -im.shape[1] / 2 # x translation (pixels) + C[1, 2] = -im.shape[0] / 2 # y translation (pixels) + + # Perspective + P = np.eye(3) + P[2, 0] = random.uniform(-perspective, perspective) # x perspective (about y) + P[2, 1] = random.uniform(-perspective, perspective) # y perspective (about x) + + # Rotation and Scale + R = np.eye(3) + a = random.uniform(-degrees, degrees) + # a += random.choice([-180, -90, 0, 90]) # add 90deg rotations to small rotations + s = random.uniform(1 - scale, 1 + scale) + # s = 2 ** random.uniform(-scale, scale) + R[:2] = cv2.getRotationMatrix2D(angle=a, center=(0, 0), scale=s) + + # Shear + S = np.eye(3) + S[0, 1] = math.tan(random.uniform(-shear, shear) * math.pi / 180) # x shear (deg) + S[1, 0] = math.tan(random.uniform(-shear, shear) * math.pi / 180) # y shear (deg) + + # Translation + T = np.eye(3) + T[0, 2] = random.uniform(0.5 - translate, 0.5 + translate) * width # x translation (pixels) + T[1, 2] = random.uniform(0.5 - translate, 0.5 + translate) * height # y translation (pixels) + + # Combined rotation matrix + M = T @ S @ R @ P @ C # order of operations (right to left) is IMPORTANT + if (border[0] != 0) or (border[1] != 0) or (M != np.eye(3)).any(): # image changed + if perspective: + im = cv2.warpPerspective(im, M, dsize=(width, height), borderValue=(114, 114, 114)) + else: # affine + im = cv2.warpAffine(im, M[:2], dsize=(width, height), borderValue=(114, 114, 114)) + + # Visualize + # import matplotlib.pyplot as plt + # ax = plt.subplots(1, 2, figsize=(12, 6))[1].ravel() + # ax[0].imshow(im[:, :, ::-1]) # base + # ax[1].imshow(im2[:, :, ::-1]) # warped + + # Transform label coordinates + n = len(targets) + if n: + use_segments = any(x.any() for x in segments) + new = np.zeros((n, 4)) + if use_segments: # warp segments + segments = resample_segments(segments) # upsample + for i, segment in enumerate(segments): + xy = np.ones((len(segment), 3)) + xy[:, :2] = segment + xy = xy @ M.T # transform + xy = xy[:, :2] / xy[:, 2:3] if perspective else xy[:, :2] # perspective rescale or affine + + # clip + new[i] = segment2box(xy, width, height) + + else: # warp boxes + xy = np.ones((n * 4, 3)) + xy[:, :2] = targets[:, [1, 2, 3, 4, 1, 4, 3, 2]].reshape(n * 4, 2) # x1y1, x2y2, x1y2, x2y1 + xy = xy @ M.T # transform + xy = (xy[:, :2] / xy[:, 2:3] if perspective else xy[:, :2]).reshape(n, 8) # perspective rescale or affine + + # create new boxes + x = xy[:, [0, 2, 4, 6]] + y = xy[:, [1, 3, 5, 7]] + new = np.concatenate((x.min(1), y.min(1), x.max(1), y.max(1))).reshape(4, n).T + + # clip + new[:, [0, 2]] = new[:, [0, 2]].clip(0, width) + new[:, [1, 3]] = new[:, [1, 3]].clip(0, height) + + # filter candidates + i = box_candidates(box1=targets[:, 1:5].T * s, box2=new.T, area_thr=0.01 if use_segments else 0.10) + targets = targets[i] + targets[:, 1:5] = new[i] + + return im, targets + + +def copy_paste(im, labels, segments, p=0.5): + # Implement Copy-Paste augmentation https://arxiv.org/abs/2012.07177, labels as nx5 np.array(cls, xyxy) + n = len(segments) + if p and n: + h, w, c = im.shape # height, width, channels + im_new = np.zeros(im.shape, np.uint8) + for j in random.sample(range(n), k=round(p * n)): + l, s = labels[j], segments[j] + box = w - l[3], l[2], w - l[1], l[4] + ioa = bbox_ioa(box, labels[:, 1:5]) # intersection over area + if (ioa < 0.30).all(): # allow 30% obscuration of existing labels + labels = np.concatenate((labels, [[l[0], *box]]), 0) + segments.append(np.concatenate((w - s[:, 0:1], s[:, 1:2]), 1)) + cv2.drawContours(im_new, [segments[j].astype(np.int32)], -1, (255, 255, 255), cv2.FILLED) + + result = cv2.bitwise_and(src1=im, src2=im_new) + result = cv2.flip(result, 1) # augment segments (flip left-right) + i = result > 0 # pixels to replace + # i[:, :] = result.max(2).reshape(h, w, 1) # act over ch + im[i] = result[i] # cv2.imwrite('debug.jpg', im) # debug + + return im, labels, segments + + +def cutout(im, labels, p=0.5): + # Applies image cutout augmentation https://arxiv.org/abs/1708.04552 + if random.random() < p: + h, w = im.shape[:2] + scales = [0.5] * 1 + [0.25] * 2 + [0.125] * 4 + [0.0625] * 8 + [0.03125] * 16 # image size fraction + for s in scales: + mask_h = random.randint(1, int(h * s)) # create random masks + mask_w = random.randint(1, int(w * s)) + + # box + xmin = max(0, random.randint(0, w) - mask_w // 2) + ymin = max(0, random.randint(0, h) - mask_h // 2) + xmax = min(w, xmin + mask_w) + ymax = min(h, ymin + mask_h) + + # apply random color mask + im[ymin:ymax, xmin:xmax] = [random.randint(64, 191) for _ in range(3)] + + # return unobscured labels + if len(labels) and s > 0.03: + box = np.array([xmin, ymin, xmax, ymax], dtype=np.float32) + ioa = bbox_ioa(box, labels[:, 1:5]) # intersection over area + labels = labels[ioa < 0.60] # remove >60% obscured labels + + return labels + + +def mixup(im, labels, im2, labels2): + # Applies MixUp augmentation https://arxiv.org/pdf/1710.09412.pdf + r = np.random.beta(32.0, 32.0) # mixup ratio, alpha=beta=32.0 + im = (im * r + im2 * (1 - r)).astype(np.uint8) + labels = np.concatenate((labels, labels2), 0) + return im, labels + + +def box_candidates(box1, box2, wh_thr=2, ar_thr=100, area_thr=0.1, eps=1e-16): # box1(4,n), box2(4,n) + # Compute candidate boxes: box1 before augment, box2 after augment, wh_thr (pixels), aspect_ratio_thr, area_ratio + w1, h1 = box1[2] - box1[0], box1[3] - box1[1] + w2, h2 = box2[2] - box2[0], box2[3] - box2[1] + ar = np.maximum(w2 / (h2 + eps), h2 / (w2 + eps)) # aspect ratio + return (w2 > wh_thr) & (h2 > wh_thr) & (w2 * h2 / (w1 * h1 + eps) > area_thr) & (ar < ar_thr) # candidates diff --git a/utils/autoanchor.py b/utils/autoanchor.py new file mode 100644 index 0000000..fd73676 --- /dev/null +++ b/utils/autoanchor.py @@ -0,0 +1,166 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +AutoAnchor utils +""" + +import random + +import numpy as np +import torch +import yaml +from tqdm import tqdm + +from utils.general import LOGGER, colorstr, emojis + +PREFIX = colorstr('AutoAnchor: ') + + +def check_anchor_order(m): + # Check anchor order against stride order for YOLOv5 Detect() module m, and correct if necessary + a = m.anchors.prod(-1).view(-1) # anchor area + da = a[-1] - a[0] # delta a + ds = m.stride[-1] - m.stride[0] # delta s + if da.sign() != ds.sign(): # same order + LOGGER.info(f'{PREFIX}Reversing anchor order') + m.anchors[:] = m.anchors.flip(0) + + +def check_anchors(dataset, model, thr=4.0, imgsz=640): + # Check anchor fit to data, recompute if necessary + m = model.module.model[-1] if hasattr(model, 'module') else model.model[-1] # Detect() + shapes = imgsz * dataset.shapes / dataset.shapes.max(1, keepdims=True)#čŋ›čĄŒä¸€ä¸Ē整äŊ“拉äŧ¸ čŽŠæœ€å¤§čžšå˜æˆ640 å…ļäģ–įŧŠæ”ž + scale = np.random.uniform(0.9, 1.1, size=(shapes.shape[0], 1)) # augment scale + wh = torch.tensor(np.concatenate([l[:, 3:5] * s for s, l in zip(shapes * scale, dataset.labels)])).float() # wh + #[l[:, 3:5] * s for s, l in zip(shapes * scale, dataset.labels)] 大小 118287 + + def metric(k): # compute metric + r = wh[:, None] / k[None] + x = torch.min(r, 1 / r).min(2)[0] # ratio metric + best = x.max(1)[0] # best_x + aat = (x > 1 / thr).float().sum(1).mean() # anchors above threshold + bpr = (best > 1 / thr).float().mean() # best possible recall + return bpr, aat + + anchors = m.anchors.clone() * m.stride.to(m.anchors.device).view(-1, 1, 1) # current anchors + bpr, aat = metric(anchors.cpu().view(-1, 2)) + s = f'\n{PREFIX}{aat:.2f} anchors/target, {bpr:.3f} Best Possible Recall (BPR). ' + if bpr > 0.98: # threshold to recompute + LOGGER.info(emojis(f'{s}Current anchors are a good fit to dataset ✅')) + else: + LOGGER.info(emojis(f'{s}Anchors are a poor fit to dataset ⚠ī¸, attempting to improve...')) + na = m.anchors.numel() // 2 # number of anchors + try: + anchors = kmean_anchors(dataset, n=na, img_size=imgsz, thr=thr, gen=1000, verbose=False) + except Exception as e: + LOGGER.info(f'{PREFIX}ERROR: {e}') + new_bpr = metric(anchors)[0] + if new_bpr > bpr: # replace anchors + anchors = torch.tensor(anchors, device=m.anchors.device).type_as(m.anchors) + m.anchors[:] = anchors.clone().view_as(m.anchors) / m.stride.to(m.anchors.device).view(-1, 1, 1) # loss + check_anchor_order(m) + LOGGER.info(f'{PREFIX}New anchors saved to model. Update model *.yaml to use these anchors in the future.') + else: + LOGGER.info(f'{PREFIX}Original anchors better than new anchors. Proceeding with original anchors.') + + +def kmean_anchors(dataset='./data/coco128.yaml', n=9, img_size=640, thr=4.0, gen=1000, verbose=True): + """ Creates kmeans-evolved anchors from training dataset + + Arguments: + dataset: path to data.yaml, or a loaded dataset + n: number of anchors + img_size: image size used for training + thr: anchor-label wh ratio threshold hyperparameter hyp['anchor_t'] used for training, default=4.0 + gen: generations to evolve anchors using genetic algorithm + verbose: print all results + + Return: + k: kmeans evolved anchors + + Usage: + from utils.autoanchor import *; _ = kmean_anchors() + """ + from scipy.cluster.vq import kmeans + + npr = np.random + thr = 1 / thr + + def metric(k, wh): # compute metrics + r = wh[:, None] / k[None] + x = torch.min(r, 1 / r).min(2)[0] # ratio metric + # x = wh_iou(wh, torch.tensor(k)) # iou metric + return x, x.max(1)[0] # x, best_x + + def anchor_fitness(k): # mutation fitness + _, best = metric(torch.tensor(k, dtype=torch.float32), wh) + return (best * (best > thr).float()).mean() # fitness + + def print_results(k, verbose=True): + k = k[np.argsort(k.prod(1))] # sort small to large + x, best = metric(k, wh0) + bpr, aat = (best > thr).float().mean(), (x > thr).float().mean() * n # best possible recall, anch > thr + s = f'{PREFIX}thr={thr:.2f}: {bpr:.4f} best possible recall, {aat:.2f} anchors past thr\n' \ + f'{PREFIX}n={n}, img_size={img_size}, metric_all={x.mean():.3f}/{best.mean():.3f}-mean/best, ' \ + f'past_thr={x[x > thr].mean():.3f}-mean: ' + for i, x in enumerate(k): + s += '%i,%i, ' % (round(x[0]), round(x[1])) + if verbose: + LOGGER.info(s[:-2]) + return k + + if isinstance(dataset, str): # *.yaml file + with open(dataset, errors='ignore') as f: + data_dict = yaml.safe_load(f) # model dict + from utils.datasets import LoadImagesAndLabels + dataset = LoadImagesAndLabels(data_dict['train'], augment=True, rect=True) + + # Get label wh + shapes = img_size * dataset.shapes / dataset.shapes.max(1, keepdims=True) + wh0 = np.concatenate([l[:, 3:5] * s for s, l in zip(shapes, dataset.labels)]) # wh + + # Filter + i = (wh0 < 3.0).any(1).sum() + if i: + LOGGER.info(f'{PREFIX}WARNING: Extremely small objects found. {i} of {len(wh0)} labels are < 3 pixels in size.') + wh = wh0[(wh0 >= 2.0).any(1)] # filter > 2 pixels + # wh = wh * (npr.rand(wh.shape[0], 1) * 0.9 + 0.1) # multiply by random scale 0-1 + + # Kmeans calculation + LOGGER.info(f'{PREFIX}Running kmeans for {n} anchors on {len(wh)} points...') + s = wh.std(0) # sigmas for whitening + k = kmeans(wh / s, n, iter=30)[0] * s # points + if len(k) != n: # kmeans may return fewer points than requested if wh is insufficient or too similar + LOGGER.warning(f'{PREFIX}WARNING: scipy.cluster.vq.kmeans returned only {len(k)} of {n} requested points') + k = np.sort(npr.rand(n * 2)).reshape(n, 2) * img_size # random init + wh = torch.tensor(wh, dtype=torch.float32) # filtered + wh0 = torch.tensor(wh0, dtype=torch.float32) # unfiltered + k = print_results(k, verbose=False) + + # Plot + # k, d = [None] * 20, [None] * 20 + # for i in tqdm(range(1, 21)): + # k[i-1], d[i-1] = kmeans(wh / s, i) # points, mean distance + # fig, ax = plt.subplots(1, 2, figsize=(14, 7), tight_layout=True) + # ax = ax.ravel() + # ax[0].plot(np.arange(1, 21), np.array(d) ** 2, marker='.') + # fig, ax = plt.subplots(1, 2, figsize=(14, 7)) # plot wh + # ax[0].hist(wh[wh[:, 0]<100, 0],400) + # ax[1].hist(wh[wh[:, 1]<100, 1],400) + # fig.savefig('wh.png', dpi=200) + + # Evolve + f, sh, mp, s = anchor_fitness(k), k.shape, 0.9, 0.1 # fitness, generations, mutation prob, sigma + pbar = tqdm(range(gen), desc=f'{PREFIX}Evolving anchors with Genetic Algorithm:') # progress bar + for _ in pbar: + v = np.ones(sh) + while (v == 1).all(): # mutate until a change occurs (prevent duplicates) + v = ((npr.random(sh) < mp) * random.random() * npr.randn(*sh) * s + 1).clip(0.3, 3.0) + kg = (k.copy() * v).clip(min=2.0) + fg = anchor_fitness(kg) + if fg > f: + f, k = fg, kg.copy() + pbar.desc = f'{PREFIX}Evolving anchors with Genetic Algorithm: fitness = {f:.4f}' + if verbose: + print_results(k, verbose) + + return print_results(k) diff --git a/utils/autobatch.py b/utils/autobatch.py new file mode 100644 index 0000000..cb94f04 --- /dev/null +++ b/utils/autobatch.py @@ -0,0 +1,57 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Auto-batch utils +""" + +from copy import deepcopy + +import numpy as np +import torch +from torch.cuda import amp + +from utils.general import LOGGER, colorstr +from utils.torch_utils import profile + + +def check_train_batch_size(model, imgsz=640): + # Check YOLOv5 training batch size + with amp.autocast(): + return autobatch(deepcopy(model).train(), imgsz) # compute optimal batch size + + +def autobatch(model, imgsz=640, fraction=0.9, batch_size=16): + # Automatically estimate best batch size to use `fraction` of available CUDA memory + # Usage: + # import torch + # from utils.autobatch import autobatch + # model = torch.hub.load('ultralytics/yolov5', 'yolov5s', autoshape=False) + # print(autobatch(model)) + + prefix = colorstr('AutoBatch: ') + LOGGER.info(f'{prefix}Computing optimal batch size for --imgsz {imgsz}') + device = next(model.parameters()).device # get model device + if device.type == 'cpu': + LOGGER.info(f'{prefix}CUDA not detected, using default CPU batch-size {batch_size}') + return batch_size + + d = str(device).upper() # 'CUDA:0' + properties = torch.cuda.get_device_properties(device) # device properties + t = properties.total_memory / 1024 ** 3 # (GiB) + r = torch.cuda.memory_reserved(device) / 1024 ** 3 # (GiB) + a = torch.cuda.memory_allocated(device) / 1024 ** 3 # (GiB) + f = t - (r + a) # free inside reserved + LOGGER.info(f'{prefix}{d} ({properties.name}) {t:.2f}G total, {r:.2f}G reserved, {a:.2f}G allocated, {f:.2f}G free') + + batch_sizes = [1, 2, 4, 8, 16] + try: + img = [torch.zeros(b, 3, imgsz, imgsz) for b in batch_sizes] + y = profile(img, model, n=3, device=device) + except Exception as e: + LOGGER.warning(f'{prefix}{e}') + + y = [x[2] for x in y if x] # memory [2] + batch_sizes = batch_sizes[:len(y)] + p = np.polyfit(batch_sizes, y, deg=1) # first degree polynomial fit + b = int((f * fraction - p[1]) / p[0]) # y intercept (optimal batch size) + LOGGER.info(f'{prefix}Using batch-size {b} for {d} {t * fraction:.2f}G/{t:.2f}G ({fraction * 100:.0f}%)') + return b diff --git a/utils/aws/__init__.py b/utils/aws/__init__.py new file mode 100644 index 0000000..e69de29 diff --git a/utils/aws/mime.sh b/utils/aws/mime.sh new file mode 100644 index 0000000..c319a83 --- /dev/null +++ b/utils/aws/mime.sh @@ -0,0 +1,26 @@ +# AWS EC2 instance startup 'MIME' script https://aws.amazon.com/premiumsupport/knowledge-center/execute-user-data-ec2/ +# This script will run on every instance restart, not only on first start +# --- DO NOT COPY ABOVE COMMENTS WHEN PASTING INTO USERDATA --- + +Content-Type: multipart/mixed; boundary="//" +MIME-Version: 1.0 + +--// +Content-Type: text/cloud-config; charset="us-ascii" +MIME-Version: 1.0 +Content-Transfer-Encoding: 7bit +Content-Disposition: attachment; filename="cloud-config.txt" + +#cloud-config +cloud_final_modules: +- [scripts-user, always] + +--// +Content-Type: text/x-shellscript; charset="us-ascii" +MIME-Version: 1.0 +Content-Transfer-Encoding: 7bit +Content-Disposition: attachment; filename="userdata.txt" + +#!/bin/bash +# --- paste contents of userdata.sh here --- +--// diff --git a/utils/aws/resume.py b/utils/aws/resume.py new file mode 100644 index 0000000..b21731c --- /dev/null +++ b/utils/aws/resume.py @@ -0,0 +1,40 @@ +# Resume all interrupted trainings in yolov5/ dir including DDP trainings +# Usage: $ python utils/aws/resume.py + +import os +import sys +from pathlib import Path + +import torch +import yaml + +FILE = Path(__file__).resolve() +ROOT = FILE.parents[2] # YOLOv5 root directory +if str(ROOT) not in sys.path: + sys.path.append(str(ROOT)) # add ROOT to PATH + +port = 0 # --master_port +path = Path('').resolve() +for last in path.rglob('*/**/last.pt'): + ckpt = torch.load(last) + if ckpt['optimizer'] is None: + continue + + # Load opt.yaml + with open(last.parent.parent / 'opt.yaml', errors='ignore') as f: + opt = yaml.safe_load(f) + + # Get device count + d = opt['device'].split(',') # devices + nd = len(d) # number of devices + ddp = nd > 1 or (nd == 0 and torch.cuda.device_count() > 1) # distributed data parallel + + if ddp: # multi-GPU + port += 1 + cmd = f'python -m torch.distributed.run --nproc_per_node {nd} --master_port {port} train.py --resume {last}' + else: # single-GPU + cmd = f'python train.py --resume {last}' + + cmd += ' > /dev/null 2>&1 &' # redirect output to dev/null and run in daemon thread + print(cmd) + os.system(cmd) diff --git a/utils/aws/userdata.sh b/utils/aws/userdata.sh new file mode 100644 index 0000000..5fc1332 --- /dev/null +++ b/utils/aws/userdata.sh @@ -0,0 +1,27 @@ +#!/bin/bash +# AWS EC2 instance startup script https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/user-data.html +# This script will run only once on first instance start (for a re-start script see mime.sh) +# /home/ubuntu (ubuntu) or /home/ec2-user (amazon-linux) is working dir +# Use >300 GB SSD + +cd home/ubuntu +if [ ! -d yolov5 ]; then + echo "Running first-time script." # install dependencies, download COCO, pull Docker + git clone https://github.com/ultralytics/yolov5 -b master && sudo chmod -R 777 yolov5 + cd yolov5 + bash data/scripts/get_coco.sh && echo "COCO done." & + sudo docker pull ultralytics/yolov5:latest && echo "Docker done." & + python -m pip install --upgrade pip && pip install -r requirements.txt && python detect.py && echo "Requirements done." & + wait && echo "All tasks done." # finish background tasks +else + echo "Running re-start script." # resume interrupted runs + i=0 + list=$(sudo docker ps -qa) # container list i.e. $'one\ntwo\nthree\nfour' + while IFS= read -r id; do + ((i++)) + echo "restarting container $i: $id" + sudo docker start $id + # sudo docker exec -it $id python train.py --resume # single-GPU + sudo docker exec -d $id python utils/aws/resume.py # multi-scenario + done <<<"$list" +fi diff --git a/utils/benchmarks.py b/utils/benchmarks.py new file mode 100644 index 0000000..962df81 --- /dev/null +++ b/utils/benchmarks.py @@ -0,0 +1,92 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Run YOLOv5 benchmarks on all supported export formats + +Format | `export.py --include` | Model +--- | --- | --- +PyTorch | - | yolov5s.pt +TorchScript | `torchscript` | yolov5s.torchscript +ONNX | `onnx` | yolov5s.onnx +OpenVINO | `openvino` | yolov5s_openvino_model/ +TensorRT | `engine` | yolov5s.engine +CoreML | `coreml` | yolov5s.mlmodel +TensorFlow SavedModel | `saved_model` | yolov5s_saved_model/ +TensorFlow GraphDef | `pb` | yolov5s.pb +TensorFlow Lite | `tflite` | yolov5s.tflite +TensorFlow Edge TPU | `edgetpu` | yolov5s_edgetpu.tflite +TensorFlow.js | `tfjs` | yolov5s_web_model/ + +Requirements: + $ pip install -r requirements.txt coremltools onnx onnx-simplifier onnxruntime openvino-dev tensorflow-cpu # CPU + $ pip install -r requirements.txt coremltools onnx onnx-simplifier onnxruntime-gpu openvino-dev tensorflow # GPU + +Usage: + $ python utils/benchmarks.py --weights yolov5s.pt --img 640 +""" + +import argparse +import sys +import time +from pathlib import Path + +import pandas as pd + +FILE = Path(__file__).resolve() +ROOT = FILE.parents[1] # YOLOv5 root directory +if str(ROOT) not in sys.path: + sys.path.append(str(ROOT)) # add ROOT to PATH +# ROOT = ROOT.relative_to(Path.cwd()) # relative + +import export +import val +from utils import notebook_init +from utils.general import LOGGER, print_args + + +def run(weights=ROOT / 'yolov5s.pt', # weights path + imgsz=640, # inference size (pixels) + batch_size=1, # batch size + data=ROOT / 'data/coco128.yaml', # dataset.yaml path + ): + y, t = [], time.time() + formats = export.export_formats() + for i, (name, f, suffix) in formats.iterrows(): # index, (name, file, suffix) + try: + w = weights if f == '-' else export.run(weights=weights, imgsz=[imgsz], include=[f], device='cpu')[-1] + assert suffix in str(w), 'export failed' + result = val.run(data, w, batch_size, imgsz=imgsz, plots=False, device='cpu', task='benchmark') + metrics = result[0] # metrics (mp, mr, map50, map, *losses(box, obj, cls)) + speeds = result[2] # times (preprocess, inference, postprocess) + y.append([name, metrics[3], speeds[1]]) # mAP, t_inference + except Exception as e: + LOGGER.warning(f'WARNING: Benchmark failure for {name}: {e}') + y.append([name, None, None]) # mAP, t_inference + + # Print results + LOGGER.info('\n') + parse_opt() + notebook_init() # print system info + py = pd.DataFrame(y, columns=['Format', 'mAP@0.5:0.95', 'Inference time (ms)']) + LOGGER.info(f'\nBenchmarks complete ({time.time() - t:.2f}s)') + LOGGER.info(str(py)) + return py + + +def parse_opt(): + parser = argparse.ArgumentParser() + parser.add_argument('--weights', type=str, default=ROOT / 'yolov5s.pt', help='weights path') + parser.add_argument('--imgsz', '--img', '--img-size', type=int, default=640, help='inference size (pixels)') + parser.add_argument('--batch-size', type=int, default=1, help='batch size') + parser.add_argument('--data', type=str, default=ROOT / 'data/coco128.yaml', help='dataset.yaml path') + opt = parser.parse_args() + print_args(FILE.stem, opt) + return opt + + +def main(opt): + run(**vars(opt)) + + +if __name__ == "__main__": + opt = parse_opt() + main(opt) diff --git a/utils/callbacks.py b/utils/callbacks.py new file mode 100644 index 0000000..c51c268 --- /dev/null +++ b/utils/callbacks.py @@ -0,0 +1,78 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Callback utils +""" + + +class Callbacks: + """" + Handles all registered callbacks for YOLOv5 Hooks + """ + + def __init__(self): + # Define the available callbacks + self._callbacks = { + 'on_pretrain_routine_start': [], + 'on_pretrain_routine_end': [], + + 'on_train_start': [], + 'on_train_epoch_start': [], + 'on_train_batch_start': [], + 'optimizer_step': [], + 'on_before_zero_grad': [], + 'on_train_batch_end': [], + 'on_train_epoch_end': [], + + 'on_val_start': [], + 'on_val_batch_start': [], + 'on_val_image_end': [], + 'on_val_batch_end': [], + 'on_val_end': [], + + 'on_fit_epoch_end': [], # fit = train + val + 'on_model_save': [], + 'on_train_end': [], + 'on_params_update': [], + 'teardown': [], + } + self.stop_training = False # set True to interrupt training + + def register_action(self, hook, name='', callback=None): + """ + Register a new action to a callback hook + + Args: + hook The callback hook name to register the action to + name The name of the action for later reference + callback The callback to fire + """ + assert hook in self._callbacks, f"hook '{hook}' not found in callbacks {self._callbacks}" + assert callable(callback), f"callback '{callback}' is not callable" + self._callbacks[hook].append({'name': name, 'callback': callback}) + + def get_registered_actions(self, hook=None): + """" + Returns all the registered actions by callback hook + + Args: + hook The name of the hook to check, defaults to all + """ + if hook: + return self._callbacks[hook] + else: + return self._callbacks + + def run(self, hook, *args, **kwargs): + """ + Loop through the registered actions and fire all callbacks + + Args: + hook The name of the hook to check, defaults to all + args Arguments to receive from YOLOv5 + kwargs Keyword Arguments to receive from YOLOv5 + """ + + assert hook in self._callbacks, f"hook '{hook}' not found in callbacks {self._callbacks}" + + for logger in self._callbacks[hook]: + logger['callback'](*args, **kwargs) diff --git a/utils/datasets.py b/utils/datasets.py new file mode 100644 index 0000000..ca3332e --- /dev/null +++ b/utils/datasets.py @@ -0,0 +1,1040 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Dataloaders and dataset utils +""" + +import glob +import hashlib +import json +import math +import os +import random +import shutil +import time +from itertools import repeat +from multiprocessing.pool import Pool, ThreadPool +from pathlib import Path +from threading import Thread +from zipfile import ZipFile + +import cv2 +import numpy as np +import torch +import torch.nn.functional as F +import yaml +from PIL import ExifTags, Image, ImageOps +from torch.utils.data import DataLoader, Dataset, dataloader, distributed +from tqdm import tqdm + +from utils.augmentations import Albumentations, augment_hsv, copy_paste, letterbox, mixup, random_perspective +from utils.general import (DATASETS_DIR, LOGGER, NUM_THREADS, check_dataset, check_requirements, check_yaml, clean_str, + segments2boxes, xyn2xy, xywh2xyxy, xywhn2xyxy, xyxy2xywhn) +from utils.torch_utils import torch_distributed_zero_first + +# Parameters +HELP_URL = 'https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data' +IMG_FORMATS = ['bmp', 'dng', 'jpeg', 'jpg', 'mpo', 'png', 'tif', 'tiff', 'webp'] # include image suffixes +VID_FORMATS = ['asf', 'avi', 'gif', 'm4v', 'mkv', 'mov', 'mp4', 'mpeg', 'mpg', 'wmv'] # include video suffixes + +# Get orientation exif tag +for orientation in ExifTags.TAGS.keys(): + if ExifTags.TAGS[orientation] == 'Orientation': + break + + +def get_hash(paths): + # Returns a single hash value of a list of paths (files or dirs) + size = sum(os.path.getsize(p) for p in paths if os.path.exists(p)) # sizes + h = hashlib.md5(str(size).encode()) # hash sizes + h.update(''.join(paths).encode()) # hash paths + return h.hexdigest() # return hash + + +def exif_size(img): + # Returns exif-corrected PIL size + s = img.size # (width, height) + try: + rotation = dict(img._getexif().items())[orientation] + if rotation == 6: # rotation 270 + s = (s[1], s[0]) + elif rotation == 8: # rotation 90 + s = (s[1], s[0]) + except Exception: + pass + + return s + + +def exif_transpose(image): + """ + Transpose a PIL image accordingly if it has an EXIF Orientation tag. + Inplace version of https://github.com/python-pillow/Pillow/blob/master/src/PIL/ImageOps.py exif_transpose() + + :param image: The image to transpose. + :return: An image. + """ + exif = image.getexif() + orientation = exif.get(0x0112, 1) # default 1 + if orientation > 1: + method = {2: Image.FLIP_LEFT_RIGHT, + 3: Image.ROTATE_180, + 4: Image.FLIP_TOP_BOTTOM, + 5: Image.TRANSPOSE, + 6: Image.ROTATE_270, + 7: Image.TRANSVERSE, + 8: Image.ROTATE_90, + }.get(orientation) + if method is not None: + image = image.transpose(method) + del exif[0x0112] + image.info["exif"] = exif.tobytes() + return image + + +def create_dataloader(path, imgsz, batch_size, stride, single_cls=False, hyp=None, augment=False, cache=False, pad=0.0, + rect=False, rank=-1, workers=8, image_weights=False, quad=False, prefix='', shuffle=False): + if rect and shuffle: + LOGGER.warning('WARNING: --rect is incompatible with DataLoader shuffle, setting shuffle=False') + shuffle = False + with torch_distributed_zero_first(rank): # init dataset *.cache only once if DDP į”¨äēŽåŒæ­Ĩ不同čŋ›į¨‹å¯šæ•°æŽč¯ģ取įš„上下文įŽĄį†å™¨ + dataset = LoadImagesAndLabels(path, imgsz, batch_size, + augment=augment, # augmentation + hyp=hyp, # hyperparameters + rect=rect, # rectangular batches self.rect=Trueæ—ļself.batch_shapes莰čŊŊ每ä¸Ēbatchįš„shape(同一ä¸Ēbatchįš„回į‰‡shapeį›¸åŒ) #why1 + cache_images=cache, + single_cls=single_cls, + stride=int(stride), + pad=pad, + image_weights=image_weights, + prefix=prefix) + + batch_size = min(batch_size, len(dataset))#čĻæ˜¯åĒ有10åŧ  é‚Ŗč‚¯åŽšé€‰bs=10 + nd = torch.cuda.device_count() # number of CUDA devices + nw = min([os.cpu_count() // max(nd, 1), batch_size if batch_size > 1 else 0, workers]) # number of workers + sampler = None if rank == -1 else distributed.DistributedSampler(dataset, shuffle=shuffle) + loader = DataLoader if image_weights else InfiniteDataLoader # only DataLoader allows for attribute updates + return loader(dataset, + batch_size=batch_size, + shuffle=shuffle and sampler is None, + num_workers=nw, + sampler=sampler, + pin_memory=True, + collate_fn=LoadImagesAndLabels.collate_fn4 if quad else LoadImagesAndLabels.collate_fn), dataset + + +class InfiniteDataLoader(dataloader.DataLoader): + """ Dataloader that reuses workers + + Uses same syntax as vanilla DataLoader + """ + + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + object.__setattr__(self, 'batch_sampler', _RepeatSampler(self.batch_sampler)) + self.iterator = super().__iter__() + + def __len__(self): + return len(self.batch_sampler.sampler) + + def __iter__(self): + for i in range(len(self)): + yield next(self.iterator) + + +class _RepeatSampler: + """ Sampler that repeats forever + + Args: + sampler (Sampler) + """ + + def __init__(self, sampler): + self.sampler = sampler + + def __iter__(self): + while True: + yield from iter(self.sampler) + + +class LoadImages: + # YOLOv5 image/video dataloader, i.e. `python detect.py --source image.jpg/vid.mp4` + def __init__(self, path, img_size=640, stride=32, auto=True): + p = str(Path(path).resolve()) # os-agnostic absolute path + if '*' in p: + files = sorted(glob.glob(p, recursive=True)) # glob + elif os.path.isdir(p): + files = sorted(glob.glob(os.path.join(p, '*.*'))) # dir + elif os.path.isfile(p): + files = [p] # files + else: + raise Exception(f'ERROR: {p} does not exist') + + images = [x for x in files if x.split('.')[-1].lower() in IMG_FORMATS] + videos = [x for x in files if x.split('.')[-1].lower() in VID_FORMATS] + ni, nv = len(images), len(videos) + + self.img_size = img_size + self.stride = stride + self.files = images + videos + self.nf = ni + nv # number of files + self.video_flag = [False] * ni + [True] * nv + self.mode = 'image' + self.auto = auto + if any(videos): + self.new_video(videos[0]) # new video + else: + self.cap = None + assert self.nf > 0, f'No images or videos found in {p}. ' \ + f'Supported formats are:\nimages: {IMG_FORMATS}\nvideos: {VID_FORMATS}' + + def __iter__(self): + self.count = 0 + return self + + def __next__(self): + if self.count == self.nf: + raise StopIteration + path = self.files[self.count] + + if self.video_flag[self.count]: + # Read video + self.mode = 'video' + ret_val, img0 = self.cap.read() + while not ret_val: + self.count += 1 + self.cap.release() + if self.count == self.nf: # last video + raise StopIteration + else: + path = self.files[self.count] + self.new_video(path) + ret_val, img0 = self.cap.read() + + self.frame += 1 + s = f'video {self.count + 1}/{self.nf} ({self.frame}/{self.frames}) {path}: ' + + else: + # Read image + self.count += 1 + img0 = cv2.imread(path) # BGR + assert img0 is not None, f'Image Not Found {path}' + s = f'image {self.count}/{self.nf} {path}: ' + + # Padded resize + img = letterbox(img0, self.img_size, stride=self.stride, auto=self.auto)[0] + + # Convert + img = img.transpose((2, 0, 1))[::-1] # HWC to CHW, BGR to RGB + img = np.ascontiguousarray(img) + + return path, img, img0, self.cap, s + + def new_video(self, path): + self.frame = 0 + self.cap = cv2.VideoCapture(path) + self.frames = int(self.cap.get(cv2.CAP_PROP_FRAME_COUNT)) + + def __len__(self): + return self.nf # number of files + + +class LoadWebcam: # for inference + # YOLOv5 local webcam dataloader, i.e. `python detect.py --source 0` + def __init__(self, pipe='0', img_size=640, stride=32): + self.img_size = img_size + self.stride = stride + self.pipe = eval(pipe) if pipe.isnumeric() else pipe + self.cap = cv2.VideoCapture(self.pipe) # video capture object + self.cap.set(cv2.CAP_PROP_BUFFERSIZE, 3) # set buffer size + + def __iter__(self): + self.count = -1 + return self + + def __next__(self): + self.count += 1 + if cv2.waitKey(1) == ord('q'): # q to quit + self.cap.release() + cv2.destroyAllWindows() + raise StopIteration + + # Read frame + ret_val, img0 = self.cap.read() + img0 = cv2.flip(img0, 1) # flip left-right + + # Print + assert ret_val, f'Camera Error {self.pipe}' + img_path = 'webcam.jpg' + s = f'webcam {self.count}: ' + + # Padded resize + img = letterbox(img0, self.img_size, stride=self.stride)[0] + + # Convert + img = img.transpose((2, 0, 1))[::-1] # HWC to CHW, BGR to RGB + img = np.ascontiguousarray(img) + + return img_path, img, img0, None, s + + def __len__(self): + return 0 + + +class LoadStreams: + # YOLOv5 streamloader, i.e. `python detect.py --source 'rtsp://example.com/media.mp4' # RTSP, RTMP, HTTP streams` + def __init__(self, sources='streams.txt', img_size=640, stride=32, auto=True): + self.mode = 'stream' + self.img_size = img_size + self.stride = stride + + if os.path.isfile(sources): + with open(sources) as f: + sources = [x.strip() for x in f.read().strip().splitlines() if len(x.strip())] + else: + sources = [sources] + + n = len(sources) + self.imgs, self.fps, self.frames, self.threads = [None] * n, [0] * n, [0] * n, [None] * n + self.sources = [clean_str(x) for x in sources] # clean source names for later + self.auto = auto + for i, s in enumerate(sources): # index, source + # Start thread to read frames from video stream + st = f'{i + 1}/{n}: {s}... ' + if 'youtube.com/' in s or 'youtu.be/' in s: # if source is YouTube video + check_requirements(('pafy', 'youtube_dl==2020.12.2')) + import pafy + s = pafy.new(s).getbest(preftype="mp4").url # YouTube URL + s = eval(s) if s.isnumeric() else s # i.e. s = '0' local webcam + cap = cv2.VideoCapture(s) + assert cap.isOpened(), f'{st}Failed to open {s}' + w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)) + h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)) + fps = cap.get(cv2.CAP_PROP_FPS) # warning: may return 0 or nan + self.frames[i] = max(int(cap.get(cv2.CAP_PROP_FRAME_COUNT)), 0) or float('inf') # infinite stream fallback + self.fps[i] = max((fps if math.isfinite(fps) else 0) % 100, 0) or 30 # 30 FPS fallback + + _, self.imgs[i] = cap.read() # guarantee first frame + self.threads[i] = Thread(target=self.update, args=([i, cap, s]), daemon=True) + LOGGER.info(f"{st} Success ({self.frames[i]} frames {w}x{h} at {self.fps[i]:.2f} FPS)") + self.threads[i].start() + LOGGER.info('') # newline + + # check for common shapes + s = np.stack([letterbox(x, self.img_size, stride=self.stride, auto=self.auto)[0].shape for x in self.imgs]) + self.rect = np.unique(s, axis=0).shape[0] == 1 # rect inference if all shapes equal + if not self.rect: + LOGGER.warning('WARNING: Stream shapes differ. For optimal performance supply similarly-shaped streams.') + + def update(self, i, cap, stream): + # Read stream `i` frames in daemon thread + n, f, read = 0, self.frames[i], 1 # frame number, frame array, inference every 'read' frame + while cap.isOpened() and n < f: + n += 1 + # _, self.imgs[index] = cap.read() + cap.grab() + if n % read == 0: + success, im = cap.retrieve() + if success: + self.imgs[i] = im + else: + LOGGER.warning('WARNING: Video stream unresponsive, please check your IP camera connection.') + self.imgs[i] = np.zeros_like(self.imgs[i]) + cap.open(stream) # re-open stream if signal was lost + time.sleep(1 / self.fps[i]) # wait time + + def __iter__(self): + self.count = -1 + return self + + def __next__(self): + self.count += 1 + if not all(x.is_alive() for x in self.threads) or cv2.waitKey(1) == ord('q'): # q to quit + cv2.destroyAllWindows() + raise StopIteration + + # Letterbox + img0 = self.imgs.copy() + img = [letterbox(x, self.img_size, stride=self.stride, auto=self.rect and self.auto)[0] for x in img0] + + # Stack + img = np.stack(img, 0) + + # Convert + img = img[..., ::-1].transpose((0, 3, 1, 2)) # BGR to RGB, BHWC to BCHW + img = np.ascontiguousarray(img) + + return self.sources, img, img0, None, '' + + def __len__(self): + return len(self.sources) # 1E12 frames = 32 streams at 30 FPS for 30 years + + +def img2label_paths(img_paths): + # Define label paths as a function of image paths + sa, sb = os.sep + 'images' + os.sep, os.sep + 'labels' + os.sep # /images/, /labels/ substrings + return [sb.join(x.rsplit(sa, 1)).rsplit('.', 1)[0] + '.txt' for x in img_paths] + + +class LoadImagesAndLabels(Dataset): + # YOLOv5 train_loader/val_loader, loads images and labels for training and validation + cache_version = 0.6 # dataset labels *.cache version + + def __init__(self, path, img_size=640, batch_size=16, augment=False, hyp=None, rect=False, image_weights=False, + cache_images=False, single_cls=False, stride=32, pad=0.0, prefix=''): + self.img_size = img_size + self.augment = augment + self.hyp = hyp + self.image_weights = image_weights + self.rect = False if image_weights else rect + self.mosaic = self.augment and not self.rect # load 4 images at a time into a mosaic (only during training) + self.mosaic_border = [-img_size // 2, -img_size // 2] + self.stride = stride + self.path = path + self.albumentations = Albumentations() if augment else None + + try: + f = [] # image files + for p in path if isinstance(path, list) else [path]: + p = Path(p) # os-agnostic + if p.is_dir(): # dir + f += glob.glob(str(p / '**' / '*.*'), recursive=True) + # f = list(p.rglob('*.*')) # pathlib + elif p.is_file(): # file + with open(p) as t: + t = t.read().strip().splitlines() + parent = str(p.parent) + os.sep + f += [x.replace('./', parent) if x.startswith('./') else x for x in t] # local to global path + # f += [p.parent / x.lstrip(os.sep) for x in t] # local to global path (pathlib) + else: + raise Exception(f'{prefix}{p} does not exist') + self.img_files = sorted(x.replace('/', os.sep) for x in f if x.split('.')[-1].lower() in IMG_FORMATS) #数字éĄēåēæŽ’åē + # self.img_files = sorted([x for x in f if x.suffix[1:].lower() in IMG_FORMATS]) # pathlib + assert self.img_files, f'{prefix}No images found'#prefixčŽŠå­—äŊ“变成įēĸ色įš„æ ‡č¯†įŦĻ + except Exception as e: + raise Exception(f'{prefix}Error loading data from {path}: {e}\nSee {HELP_URL}') + + # Check cache + self.label_files = img2label_paths(self.img_files) # labels + cache_path = (p if p.is_file() else Path(self.label_files[0]).parent).with_suffix('.cache') + try: + cache, exists = np.load(cache_path, allow_pickle=True).item(), True # load dict įĨžåĨ‡įš„æ–šæŗ•å¯äģĨč¯ģå‡ēæĨ + assert cache['version'] == self.cache_version # same version + assert cache['hash'] == get_hash(self.label_files + self.img_files) # same hash + except Exception: + cache, exists = self.cache_labels(cache_path, prefix), False # cache + + # Display cache + nf, nm, ne, nc, n = cache.pop('results') #不įŸĨ道äģŽå“ĒæĨįš„ found, missing, empty, corrupt, total + if exists: + d = f"Scanning '{cache_path}' images and labels... {nf} found, {nm} missing, {ne} empty, {nc} corrupt" + tqdm(None, desc=prefix + d, total=n, initial=n) # display cache results + if cache['msgs']: + LOGGER.info('\n'.join(cache['msgs'])) # display warnings + assert nf > 0 or not augment, f'{prefix}No labels in {cache_path}. Can not train without labels. See {HELP_URL}' + + # Read cache + [cache.pop(k) for k in ('hash', 'version', 'msgs')] # remove items + labels, shapes, self.segments = zip(*cache.values()) + self.labels = list(labels) + self.shapes = np.array(shapes, dtype=np.float64) + self.img_files = list(cache.keys()) # update + self.label_files = img2label_paths(cache.keys()) # update + n = len(shapes) # number of images + bi = np.floor(np.arange(n) / batch_size).astype(np.int) # batch index + nb = bi[-1] + 1 # number of batches + self.batch = bi # batch index of image + self.n = n + self.indices = range(n) + + # Update labels + include_class = [] # filter labels to include only these classes (optional) + include_class_array = np.array(include_class).reshape(1, -1) + for i, (label, segment) in enumerate(zip(self.labels, self.segments)):#单įąģ情å†ĩ 暂æ—ļ不į”¨įœ‹äē† + if include_class: + j = (label[:, 0:1] == include_class_array).any(1) + self.labels[i] = label[j] + if segment: + self.segments[i] = segment[j] + if single_cls: # single-class training, merge all classes into 0 + self.labels[i][:, 0] = 0 + if segment: + self.segments[i][:, 0] = 0 + + # Rectangular Training + if self.rect: + # Sort by aspect ratio + s = self.shapes # wh + ar = s[:, 1] / s[:, 0] # aspect ratio + irect = ar.argsort() + self.img_files = [self.img_files[i] for i in irect] + self.label_files = [self.label_files[i] for i in irect] + self.labels = [self.labels[i] for i in irect] + self.shapes = s[irect] # wh + ar = ar[irect] + + # Set training image shapes + shapes = [[1, 1]] * nb + for i in range(nb): + ari = ar[bi == i] + mini, maxi = ari.min(), ari.max() + if maxi < 1: + shapes[i] = [maxi, 1] + elif mini > 1: + shapes[i] = [1, 1 / mini] + + self.batch_shapes = np.ceil(np.array(shapes) * img_size / stride + pad).astype(np.int) * stride + + # Cache images into RAM/disk for faster training (WARNING: large datasets may exceed system resources) + self.imgs, self.img_npy = [None] * n, [None] * n + if cache_images: + if cache_images == 'disk': + self.im_cache_dir = Path(Path(self.img_files[0]).parent.as_posix() + '_npy') + self.img_npy = [self.im_cache_dir / Path(f).with_suffix('.npy').name for f in self.img_files] + self.im_cache_dir.mkdir(parents=True, exist_ok=True) + gb = 0 # Gigabytes of cached images + self.img_hw0, self.img_hw = [None] * n, [None] * n + results = ThreadPool(NUM_THREADS).imap(self.load_image, range(n)) + pbar = tqdm(enumerate(results), total=n) + for i, x in pbar: + if cache_images == 'disk': + if not self.img_npy[i].exists(): + np.save(self.img_npy[i].as_posix(), x[0]) + gb += self.img_npy[i].stat().st_size + else: # 'ram' + self.imgs[i], self.img_hw0[i], self.img_hw[i] = x # im, hw_orig, hw_resized = load_image(self, i) + gb += self.imgs[i].nbytes + pbar.desc = f'{prefix}Caching images ({gb / 1E9:.1f}GB {cache_images})' + pbar.close() + + def cache_labels(self, path=Path('./labels.cache'), prefix=''): + # Cache dataset labels, check images and read shapes + x = {} # dict + nm, nf, ne, nc, msgs = 0, 0, 0, 0, [] # number missing, found, empty, corrupt, messages + desc = f"{prefix}Scanning '{path.parent / path.stem}' images and labels..." + with Pool(NUM_THREADS) as pool: + pbar = tqdm(pool.imap(verify_image_label, zip(self.img_files, self.label_files, repeat(prefix))), + desc=desc, total=len(self.img_files)) + for im_file, lb, shape, segments, nm_f, nf_f, ne_f, nc_f, msg in pbar: + nm += nm_f + nf += nf_f + ne += ne_f + nc += nc_f + if im_file: + x[im_file] = [lb, shape, segments] + if msg: + msgs.append(msg) + pbar.desc = f"{desc}{nf} found, {nm} missing, {ne} empty, {nc} corrupt" + + pbar.close() + if msgs: + LOGGER.info('\n'.join(msgs)) + if nf == 0: + LOGGER.warning(f'{prefix}WARNING: No labels found in {path}. See {HELP_URL}') + x['hash'] = get_hash(self.label_files + self.img_files) + x['results'] = nf, nm, ne, nc, len(self.img_files) + x['msgs'] = msgs # warnings + x['version'] = self.cache_version # cache version + try: + np.save(path, x) # save cache for next time + path.with_suffix('.cache.npy').rename(path) # remove .npy suffix + LOGGER.info(f'{prefix}New cache created: {path}') + except Exception as e: + LOGGER.warning(f'{prefix}WARNING: Cache directory {path.parent} is not writeable: {e}') # not writeable + return x + + def __len__(self): + return len(self.img_files) + + # def __iter__(self): + # self.count = -1 + # print('ran dataset iter') + # #self.shuffled_vector = np.random.permutation(self.nF) if self.augment else np.arange(self.nF) + # return self + + def __getitem__(self, index): + index = self.indices[index] # linear, shuffled, or image_weights + + hyp = self.hyp + mosaic = self.mosaic and random.random() < hyp['mosaic'] + if mosaic: + # Load mosaic + img, labels = self.load_mosaic(index) + shapes = None + + # MixUp augmentation + if random.random() < hyp['mixup']: + img, labels = mixup(img, labels, *self.load_mosaic(random.randint(0, self.n - 1))) + + else: + # Load image + img, (h0, w0), (h, w) = self.load_image(index) + + # Letterbox + shape = self.batch_shapes[self.batch[index]] if self.rect else self.img_size # final letterboxed shape + img, ratio, pad = letterbox(img, shape, auto=False, scaleup=self.augment) + shapes = (h0, w0), ((h / h0, w / w0), pad) # for COCO mAP rescaling + + labels = self.labels[index].copy() + if labels.size: # normalized xywh to pixel xyxy format + labels[:, 1:] = xywhn2xyxy(labels[:, 1:], ratio[0] * w, ratio[1] * h, padw=pad[0], padh=pad[1]) + + if self.augment: + img, labels = random_perspective(img, labels, + degrees=hyp['degrees'], + translate=hyp['translate'], + scale=hyp['scale'], + shear=hyp['shear'], + perspective=hyp['perspective']) + + nl = len(labels) # number of labels + if nl: + labels[:, 1:5] = xyxy2xywhn(labels[:, 1:5], w=img.shape[1], h=img.shape[0], clip=True, eps=1E-3) + + if self.augment: + # Albumentations + img, labels = self.albumentations(img, labels) + nl = len(labels) # update after albumentations + + # HSV color-space + augment_hsv(img, hgain=hyp['hsv_h'], sgain=hyp['hsv_s'], vgain=hyp['hsv_v']) + + # Flip up-down + if random.random() < hyp['flipud']: + img = np.flipud(img) + if nl: + labels[:, 2] = 1 - labels[:, 2] + + # Flip left-right + if random.random() < hyp['fliplr']: + img = np.fliplr(img) + if nl: + labels[:, 1] = 1 - labels[:, 1] + + # Cutouts + # labels = cutout(img, labels, p=0.5) + # nl = len(labels) # update after cutout + + labels_out = torch.zeros((nl, 6)) + if nl: + labels_out[:, 1:] = torch.from_numpy(labels) + + # Convert + img = img.transpose((2, 0, 1))[::-1] # HWC to CHW, BGR to RGB + img = np.ascontiguousarray(img) + + return torch.from_numpy(img), labels_out, self.img_files[index], shapes + + def load_image(self, i): + # loads 1 image from dataset index 'i', returns (im, original hw, resized hw) + im = self.imgs[i] + if im is None: # not cached in RAM + npy = self.img_npy[i] + if npy and npy.exists(): # load npy + im = np.load(npy) + else: # read image + f = self.img_files[i] + im = cv2.imread(f) # BGR + assert im is not None, f'Image Not Found {f}' + h0, w0 = im.shape[:2] # orig hw + r = self.img_size / max(h0, w0) # ratio + if r != 1: # if sizes are not equal + im = cv2.resize(im, + (int(w0 * r), int(h0 * r)), + interpolation=cv2.INTER_LINEAR if (self.augment or r > 1) else cv2.INTER_AREA) + return im, (h0, w0), im.shape[:2] # im, hw_original, hw_resized + else: + return self.imgs[i], self.img_hw0[i], self.img_hw[i] # im, hw_original, hw_resized + + def load_mosaic(self, index): + # YOLOv5 4-mosaic loader. Loads 1 image + 3 random images into a 4-image mosaic + labels4, segments4 = [], [] + s = self.img_size + yc, xc = (int(random.uniform(-x, 2 * s + x)) for x in self.mosaic_border) # mosaic center x, y + indices = [index] + random.choices(self.indices, k=3) # 3 additional image indices + random.shuffle(indices) + for i, index in enumerate(indices): + # Load image + img, _, (h, w) = self.load_image(index) + + # place img in img4 + if i == 0: # top left + img4 = np.full((s * 2, s * 2, img.shape[2]), 114, dtype=np.uint8) # base image with 4 tiles + x1a, y1a, x2a, y2a = max(xc - w, 0), max(yc - h, 0), xc, yc # xmin, ymin, xmax, ymax (large image) + x1b, y1b, x2b, y2b = w - (x2a - x1a), h - (y2a - y1a), w, h # xmin, ymin, xmax, ymax (small image) + elif i == 1: # top right + x1a, y1a, x2a, y2a = xc, max(yc - h, 0), min(xc + w, s * 2), yc + x1b, y1b, x2b, y2b = 0, h - (y2a - y1a), min(w, x2a - x1a), h + elif i == 2: # bottom left + x1a, y1a, x2a, y2a = max(xc - w, 0), yc, xc, min(s * 2, yc + h) + x1b, y1b, x2b, y2b = w - (x2a - x1a), 0, w, min(y2a - y1a, h) + elif i == 3: # bottom right + x1a, y1a, x2a, y2a = xc, yc, min(xc + w, s * 2), min(s * 2, yc + h) + x1b, y1b, x2b, y2b = 0, 0, min(w, x2a - x1a), min(y2a - y1a, h) + + img4[y1a:y2a, x1a:x2a] = img[y1b:y2b, x1b:x2b] # img4[ymin:ymax, xmin:xmax] + padw = x1a - x1b + padh = y1a - y1b + + # Labels + labels, segments = self.labels[index].copy(), self.segments[index].copy() + if labels.size: + labels[:, 1:] = xywhn2xyxy(labels[:, 1:], w, h, padw, padh) # normalized xywh to pixel xyxy format + segments = [xyn2xy(x, w, h, padw, padh) for x in segments] + labels4.append(labels) + segments4.extend(segments) + + # Concat/clip labels + labels4 = np.concatenate(labels4, 0) + for x in (labels4[:, 1:], *segments4): + np.clip(x, 0, 2 * s, out=x) # clip when using random_perspective() + # img4, labels4 = replicate(img4, labels4) # replicate + + # Augment + img4, labels4, segments4 = copy_paste(img4, labels4, segments4, p=self.hyp['copy_paste']) + img4, labels4 = random_perspective(img4, labels4, segments4, + degrees=self.hyp['degrees'], + translate=self.hyp['translate'], + scale=self.hyp['scale'], + shear=self.hyp['shear'], + perspective=self.hyp['perspective'], + border=self.mosaic_border) # border to remove + + return img4, labels4 + + def load_mosaic9(self, index): + # YOLOv5 9-mosaic loader. Loads 1 image + 8 random images into a 9-image mosaic + labels9, segments9 = [], [] + s = self.img_size + indices = [index] + random.choices(self.indices, k=8) # 8 additional image indices + random.shuffle(indices) + hp, wp = -1, -1 # height, width previous + for i, index in enumerate(indices): + # Load image + img, _, (h, w) = self.load_image(index) + + # place img in img9 + if i == 0: # center + img9 = np.full((s * 3, s * 3, img.shape[2]), 114, dtype=np.uint8) # base image with 4 tiles + h0, w0 = h, w + c = s, s, s + w, s + h # xmin, ymin, xmax, ymax (base) coordinates + elif i == 1: # top + c = s, s - h, s + w, s + elif i == 2: # top right + c = s + wp, s - h, s + wp + w, s + elif i == 3: # right + c = s + w0, s, s + w0 + w, s + h + elif i == 4: # bottom right + c = s + w0, s + hp, s + w0 + w, s + hp + h + elif i == 5: # bottom + c = s + w0 - w, s + h0, s + w0, s + h0 + h + elif i == 6: # bottom left + c = s + w0 - wp - w, s + h0, s + w0 - wp, s + h0 + h + elif i == 7: # left + c = s - w, s + h0 - h, s, s + h0 + elif i == 8: # top left + c = s - w, s + h0 - hp - h, s, s + h0 - hp + + padx, pady = c[:2] + x1, y1, x2, y2 = (max(x, 0) for x in c) # allocate coords + + # Labels + labels, segments = self.labels[index].copy(), self.segments[index].copy() + if labels.size: + labels[:, 1:] = xywhn2xyxy(labels[:, 1:], w, h, padx, pady) # normalized xywh to pixel xyxy format + segments = [xyn2xy(x, w, h, padx, pady) for x in segments] + labels9.append(labels) + segments9.extend(segments) + + # Image + img9[y1:y2, x1:x2] = img[y1 - pady:, x1 - padx:] # img9[ymin:ymax, xmin:xmax] + hp, wp = h, w # height, width previous + + # Offset + yc, xc = (int(random.uniform(0, s)) for _ in self.mosaic_border) # mosaic center x, y + img9 = img9[yc:yc + 2 * s, xc:xc + 2 * s] + + # Concat/clip labels + labels9 = np.concatenate(labels9, 0) + labels9[:, [1, 3]] -= xc + labels9[:, [2, 4]] -= yc + c = np.array([xc, yc]) # centers + segments9 = [x - c for x in segments9] + + for x in (labels9[:, 1:], *segments9): + np.clip(x, 0, 2 * s, out=x) # clip when using random_perspective() + # img9, labels9 = replicate(img9, labels9) # replicate + + # Augment + img9, labels9 = random_perspective(img9, labels9, segments9, + degrees=self.hyp['degrees'], + translate=self.hyp['translate'], + scale=self.hyp['scale'], + shear=self.hyp['shear'], + perspective=self.hyp['perspective'], + border=self.mosaic_border) # border to remove + + return img9, labels9 + + @staticmethod + def collate_fn(batch): + img, label, path, shapes = zip(*batch) # transposed + for i, lb in enumerate(label): + lb[:, 0] = i # add target image index for build_targets() + return torch.stack(img, 0), torch.cat(label, 0), path, shapes + + @staticmethod + def collate_fn4(batch): + img, label, path, shapes = zip(*batch) # transposed + n = len(shapes) // 4 + img4, label4, path4, shapes4 = [], [], path[:n], shapes[:n] + + ho = torch.tensor([[0.0, 0, 0, 1, 0, 0]]) + wo = torch.tensor([[0.0, 0, 1, 0, 0, 0]]) + s = torch.tensor([[1, 1, 0.5, 0.5, 0.5, 0.5]]) # scale + for i in range(n): # zidane torch.zeros(16,3,720,1280) # BCHW + i *= 4 + if random.random() < 0.5: + im = F.interpolate(img[i].unsqueeze(0).float(), scale_factor=2.0, mode='bilinear', align_corners=False)[ + 0].type(img[i].type()) + lb = label[i] + else: + im = torch.cat((torch.cat((img[i], img[i + 1]), 1), torch.cat((img[i + 2], img[i + 3]), 1)), 2) + lb = torch.cat((label[i], label[i + 1] + ho, label[i + 2] + wo, label[i + 3] + ho + wo), 0) * s + img4.append(im) + label4.append(lb) + + for i, lb in enumerate(label4): + lb[:, 0] = i # add target image index for build_targets() + + return torch.stack(img4, 0), torch.cat(label4, 0), path4, shapes4 + + +# Ancillary functions -------------------------------------------------------------------------------------------------- +def create_folder(path='./new'): + # Create folder + if os.path.exists(path): + shutil.rmtree(path) # delete output folder + os.makedirs(path) # make new output folder + + +def flatten_recursive(path=DATASETS_DIR / 'coco128'): + # Flatten a recursive directory by bringing all files to top level + new_path = Path(str(path) + '_flat') + create_folder(new_path) + for file in tqdm(glob.glob(str(Path(path)) + '/**/*.*', recursive=True)): + shutil.copyfile(file, new_path / Path(file).name) + + +def extract_boxes(path=DATASETS_DIR / 'coco128'): # from utils.datasets import *; extract_boxes() + # Convert detection dataset into classification dataset, with one directory per class + path = Path(path) # images dir + shutil.rmtree(path / 'classifier') if (path / 'classifier').is_dir() else None # remove existing + files = list(path.rglob('*.*')) + n = len(files) # number of files + for im_file in tqdm(files, total=n): + if im_file.suffix[1:] in IMG_FORMATS: + # image + im = cv2.imread(str(im_file))[..., ::-1] # BGR to RGB + h, w = im.shape[:2] + + # labels + lb_file = Path(img2label_paths([str(im_file)])[0]) + if Path(lb_file).exists(): + with open(lb_file) as f: + lb = np.array([x.split() for x in f.read().strip().splitlines()], dtype=np.float32) # labels + + for j, x in enumerate(lb): + c = int(x[0]) # class + f = (path / 'classifier') / f'{c}' / f'{path.stem}_{im_file.stem}_{j}.jpg' # new filename + if not f.parent.is_dir(): + f.parent.mkdir(parents=True) + + b = x[1:] * [w, h, w, h] # box + # b[2:] = b[2:].max() # rectangle to square + b[2:] = b[2:] * 1.2 + 3 # pad + b = xywh2xyxy(b.reshape(-1, 4)).ravel().astype(np.int) + + b[[0, 2]] = np.clip(b[[0, 2]], 0, w) # clip boxes outside of image + b[[1, 3]] = np.clip(b[[1, 3]], 0, h) + assert cv2.imwrite(str(f), im[b[1]:b[3], b[0]:b[2]]), f'box failure in {f}' + + +def autosplit(path=DATASETS_DIR / 'coco128/images', weights=(0.9, 0.1, 0.0), annotated_only=False): + """ Autosplit a dataset into train/val/test splits and save path/autosplit_*.txt files + Usage: from utils.datasets import *; autosplit() + Arguments + path: Path to images directory + weights: Train, val, test weights (list, tuple) + annotated_only: Only use images with an annotated txt file + """ + path = Path(path) # images dir + files = sorted(x for x in path.rglob('*.*') if x.suffix[1:].lower() in IMG_FORMATS) # image files only + n = len(files) # number of files + random.seed(0) # for reproducibility + indices = random.choices([0, 1, 2], weights=weights, k=n) # assign each image to a split + + txt = ['autosplit_train.txt', 'autosplit_val.txt', 'autosplit_test.txt'] # 3 txt files + [(path.parent / x).unlink(missing_ok=True) for x in txt] # remove existing + + print(f'Autosplitting images from {path}' + ', using *.txt labeled images only' * annotated_only) + for i, img in tqdm(zip(indices, files), total=n): + if not annotated_only or Path(img2label_paths([str(img)])[0]).exists(): # check label + with open(path.parent / txt[i], 'a') as f: + f.write('./' + img.relative_to(path.parent).as_posix() + '\n') # add image to txt file + + +def verify_image_label(args): + # Verify one image-label pair + im_file, lb_file, prefix = args + nm, nf, ne, nc, msg, segments = 0, 0, 0, 0, '', [] # number (missing, found, empty, corrupt), message, segments + try: + # verify images + im = Image.open(im_file) + im.verify() # PIL verify + shape = exif_size(im) # image size + assert (shape[0] > 9) & (shape[1] > 9), f'image size {shape} <10 pixels' + assert im.format.lower() in IMG_FORMATS, f'invalid image format {im.format}' + if im.format.lower() in ('jpg', 'jpeg'): + with open(im_file, 'rb') as f: + f.seek(-2, 2) + if f.read() != b'\xff\xd9': # corrupt JPEG + ImageOps.exif_transpose(Image.open(im_file)).save(im_file, 'JPEG', subsampling=0, quality=100) + msg = f'{prefix}WARNING: {im_file}: corrupt JPEG restored and saved' + + # verify labels + if os.path.isfile(lb_file): + nf = 1 # label found + with open(lb_file) as f: + lb = [x.split() for x in f.read().strip().splitlines() if len(x)] + if any([len(x) > 8 for x in lb]): # is segment + classes = np.array([x[0] for x in lb], dtype=np.float32) + segments = [np.array(x[1:], dtype=np.float32).reshape(-1, 2) for x in lb] # (cls, xy1...) + lb = np.concatenate((classes.reshape(-1, 1), segments2boxes(segments)), 1) # (cls, xywh) + lb = np.array(lb, dtype=np.float32) + nl = len(lb) + #莊lb少äēŽ0大äēŽ1éƒŊ变成0和1 + #lb[lb<0.]=0. + #lb[lb>1.]=1. + if nl: + assert lb.shape[1] == 5, f'labels require 5 columns, {lb.shape[1]} columns detected' + assert (lb >= 0).all(), f'negative label values {lb[lb < 0]}' + assert (lb[:, 1:] <= 1).all(), f'non-normalized or out of bounds coordinates {lb[:, 1:][lb[:, 1:] > 1]}' + _, i = np.unique(lb, axis=0, return_index=True) + if len(i) < nl: # duplicate row check + lb = lb[i] # remove duplicates + if segments: + segments = segments[i] + msg = f'{prefix}WARNING: {im_file}: {nl - len(i)} duplicate labels removed' + else: + ne = 1 # label empty + lb = np.zeros((0, 5), dtype=np.float32) + else: + nm = 1 # label missing + lb = np.zeros((0, 5), dtype=np.float32) + return im_file, lb, shape, segments, nm, nf, ne, nc, msg + except Exception as e: + nc = 1 + msg = f'{prefix}WARNING: {im_file}: ignoring corrupt image/label: {e}' + return [None, None, None, None, nm, nf, ne, nc, msg] + + +def dataset_stats(path='coco128.yaml', autodownload=False, verbose=False, profile=False, hub=False): + """ Return dataset statistics dictionary with images and instances counts per split per class + To run in parent directory: export PYTHONPATH="$PWD/yolov5" + Usage1: from utils.datasets import *; dataset_stats('coco128.yaml', autodownload=True) + Usage2: from utils.datasets import *; dataset_stats('path/to/coco128_with_yaml.zip') + Arguments + path: Path to data.yaml or data.zip (with data.yaml inside data.zip) + autodownload: Attempt to download dataset if not found locally + verbose: Print stats dictionary + """ + + def round_labels(labels): + # Update labels to integer class and 6 decimal place floats + return [[int(c), *(round(x, 4) for x in points)] for c, *points in labels] + + def unzip(path): + # Unzip data.zip TODO: CONSTRAINT: path/to/abc.zip MUST unzip to 'path/to/abc/' + if str(path).endswith('.zip'): # path is data.zip + assert Path(path).is_file(), f'Error unzipping {path}, file not found' + ZipFile(path).extractall(path=path.parent) # unzip + dir = path.with_suffix('') # dataset directory == zip name + return True, str(dir), next(dir.rglob('*.yaml')) # zipped, data_dir, yaml_path + else: # path is data.yaml + return False, None, path + + def hub_ops(f, max_dim=1920): + # HUB ops for 1 image 'f': resize and save at reduced quality in /dataset-hub for web/app viewing + f_new = im_dir / Path(f).name # dataset-hub image filename + try: # use PIL + im = Image.open(f) + r = max_dim / max(im.height, im.width) # ratio + if r < 1.0: # image too large + im = im.resize((int(im.width * r), int(im.height * r))) + im.save(f_new, 'JPEG', quality=75, optimize=True) # save + except Exception as e: # use OpenCV + print(f'WARNING: HUB ops PIL failure {f}: {e}') + im = cv2.imread(f) + im_height, im_width = im.shape[:2] + r = max_dim / max(im_height, im_width) # ratio + if r < 1.0: # image too large + im = cv2.resize(im, (int(im_width * r), int(im_height * r)), interpolation=cv2.INTER_AREA) + cv2.imwrite(str(f_new), im) + + zipped, data_dir, yaml_path = unzip(Path(path)) + with open(check_yaml(yaml_path), errors='ignore') as f: + data = yaml.safe_load(f) # data dict + if zipped: + data['path'] = data_dir # TODO: should this be dir.resolve()? + check_dataset(data, autodownload) # download dataset if missing + hub_dir = Path(data['path'] + ('-hub' if hub else '')) + stats = {'nc': data['nc'], 'names': data['names']} # statistics dictionary + for split in 'train', 'val', 'test': + if data.get(split) is None: + stats[split] = None # i.e. no test set + continue + x = [] + dataset = LoadImagesAndLabels(data[split]) # load dataset + for label in tqdm(dataset.labels, total=dataset.n, desc='Statistics'): + x.append(np.bincount(label[:, 0].astype(int), minlength=data['nc'])) + x = np.array(x) # shape(128x80) + stats[split] = {'instance_stats': {'total': int(x.sum()), 'per_class': x.sum(0).tolist()}, + 'image_stats': {'total': dataset.n, 'unlabelled': int(np.all(x == 0, 1).sum()), + 'per_class': (x > 0).sum(0).tolist()}, + 'labels': [{str(Path(k).name): round_labels(v.tolist())} for k, v in + zip(dataset.img_files, dataset.labels)]} + + if hub: + im_dir = hub_dir / 'images' + im_dir.mkdir(parents=True, exist_ok=True) + for _ in tqdm(ThreadPool(NUM_THREADS).imap(hub_ops, dataset.img_files), total=dataset.n, desc='HUB Ops'): + pass + + # Profile + stats_path = hub_dir / 'stats.json' + if profile: + for _ in range(1): + file = stats_path.with_suffix('.npy') + t1 = time.time() + np.save(file, stats) + t2 = time.time() + x = np.load(file, allow_pickle=True) + print(f'stats.npy times: {time.time() - t2:.3f}s read, {t2 - t1:.3f}s write') + + file = stats_path.with_suffix('.json') + t1 = time.time() + with open(file, 'w') as f: + json.dump(stats, f) # save stats *.json + t2 = time.time() + with open(file) as f: + x = json.load(f) # load hyps dict + print(f'stats.json times: {time.time() - t2:.3f}s read, {t2 - t1:.3f}s write') + + # Save, print and return + if hub: + print(f'Saving {stats_path.resolve()}...') + with open(stats_path, 'w') as f: + json.dump(stats, f) # save stats.json + if verbose: + print(json.dumps(stats, indent=2, sort_keys=False)) + return stats diff --git a/utils/downloads.py b/utils/downloads.py new file mode 100644 index 0000000..d7b87cb --- /dev/null +++ b/utils/downloads.py @@ -0,0 +1,153 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Download utils +""" + +import os +import platform +import subprocess +import time +import urllib +from pathlib import Path +from zipfile import ZipFile + +import requests +import torch + + +def gsutil_getsize(url=''): + # gs://bucket/file size https://cloud.google.com/storage/docs/gsutil/commands/du + s = subprocess.check_output(f'gsutil du {url}', shell=True).decode('utf-8') + return eval(s.split(' ')[0]) if len(s) else 0 # bytes + + +def safe_download(file, url, url2=None, min_bytes=1E0, error_msg=''): + # Attempts to download file from url or url2, checks and removes incomplete downloads < min_bytes + file = Path(file) + assert_msg = f"Downloaded file '{file}' does not exist or size is < min_bytes={min_bytes}" + try: # url1 + print(f'Downloading {url} to {file}...') + torch.hub.download_url_to_file(url, str(file)) + assert file.exists() and file.stat().st_size > min_bytes, assert_msg # check + except Exception as e: # url2 + file.unlink(missing_ok=True) # remove partial downloads + print(f'ERROR: {e}\nRe-attempting {url2 or url} to {file}...') + os.system(f"curl -L '{url2 or url}' -o '{file}' --retry 3 -C -") # curl download, retry and resume on fail + finally: + if not file.exists() or file.stat().st_size < min_bytes: # check + file.unlink(missing_ok=True) # remove partial downloads + print(f"ERROR: {assert_msg}\n{error_msg}") + print('') + + +def attempt_download(file, repo='ultralytics/yolov5'): # from utils.downloads import *; attempt_download() + # Attempt file download if does not exist + file = Path(str(file).strip().replace("'", '')) + + if not file.exists(): + # URL specified + name = Path(urllib.parse.unquote(str(file))).name # decode '%2F' to '/' etc. + if str(file).startswith(('http:/', 'https:/')): # download + url = str(file).replace(':/', '://') # Pathlib turns :// -> :/ + file = name.split('?')[0] # parse authentication https://url.com/file.txt?auth... + if Path(file).is_file(): + print(f'Found {url} locally at {file}') # file already exists + else: + safe_download(file=file, url=url, min_bytes=1E5) + return file + + # GitHub assets + file.parent.mkdir(parents=True, exist_ok=True) # make parent dir (if required) + try: + response = requests.get(f'https://api.github.com/repos/{repo}/releases/latest').json() # github api + assets = [x['name'] for x in response['assets']] # release assets, i.e. ['yolov5s.pt', 'yolov5m.pt', ...] + tag = response['tag_name'] # i.e. 'v1.0' + except Exception: # fallback plan + assets = ['yolov5n.pt', 'yolov5s.pt', 'yolov5m.pt', 'yolov5l.pt', 'yolov5x.pt', + 'yolov5n6.pt', 'yolov5s6.pt', 'yolov5m6.pt', 'yolov5l6.pt', 'yolov5x6.pt'] + try: + tag = subprocess.check_output('git tag', shell=True, stderr=subprocess.STDOUT).decode().split()[-1] + except Exception: + tag = 'v6.0' # current release + + if name in assets: + safe_download(file, + url=f'https://github.com/{repo}/releases/download/{tag}/{name}', + # url2=f'https://storage.googleapis.com/{repo}/ckpt/{name}', # backup url (optional) + min_bytes=1E5, + error_msg=f'{file} missing, try downloading from https://github.com/{repo}/releases/') + + return str(file) + + +def gdrive_download(id='16TiPfZj7htmTyhntwcZyEEAejOUxuT6m', file='tmp.zip'): + # Downloads a file from Google Drive. from yolov5.utils.downloads import *; gdrive_download() + t = time.time() + file = Path(file) + cookie = Path('cookie') # gdrive cookie + print(f'Downloading https://drive.google.com/uc?export=download&id={id} as {file}... ', end='') + file.unlink(missing_ok=True) # remove existing file + cookie.unlink(missing_ok=True) # remove existing cookie + + # Attempt file download + out = "NUL" if platform.system() == "Windows" else "/dev/null" + os.system(f'curl -c ./cookie -s -L "drive.google.com/uc?export=download&id={id}" > {out}') + if os.path.exists('cookie'): # large file + s = f'curl -Lb ./cookie "drive.google.com/uc?export=download&confirm={get_token()}&id={id}" -o {file}' + else: # small file + s = f'curl -s -L -o {file} "drive.google.com/uc?export=download&id={id}"' + r = os.system(s) # execute, capture return + cookie.unlink(missing_ok=True) # remove existing cookie + + # Error check + if r != 0: + file.unlink(missing_ok=True) # remove partial + print('Download error ') # raise Exception('Download error') + return r + + # Unzip if archive + if file.suffix == '.zip': + print('unzipping... ', end='') + ZipFile(file).extractall(path=file.parent) # unzip + file.unlink() # remove zip + + print(f'Done ({time.time() - t:.1f}s)') + return r + + +def get_token(cookie="./cookie"): + with open(cookie) as f: + for line in f: + if "download" in line: + return line.split()[-1] + return "" + +# Google utils: https://cloud.google.com/storage/docs/reference/libraries ---------------------------------------------- +# +# +# def upload_blob(bucket_name, source_file_name, destination_blob_name): +# # Uploads a file to a bucket +# # https://cloud.google.com/storage/docs/uploading-objects#storage-upload-object-python +# +# storage_client = storage.Client() +# bucket = storage_client.get_bucket(bucket_name) +# blob = bucket.blob(destination_blob_name) +# +# blob.upload_from_filename(source_file_name) +# +# print('File {} uploaded to {}.'.format( +# source_file_name, +# destination_blob_name)) +# +# +# def download_blob(bucket_name, source_blob_name, destination_file_name): +# # Uploads a blob from a bucket +# storage_client = storage.Client() +# bucket = storage_client.get_bucket(bucket_name) +# blob = bucket.blob(source_blob_name) +# +# blob.download_to_filename(destination_file_name) +# +# print('Blob {} downloaded to {}.'.format( +# source_blob_name, +# destination_file_name)) diff --git a/utils/flask_rest_api/README.md b/utils/flask_rest_api/README.md new file mode 100644 index 0000000..a726acb --- /dev/null +++ b/utils/flask_rest_api/README.md @@ -0,0 +1,73 @@ +# Flask REST API + +[REST](https://en.wikipedia.org/wiki/Representational_state_transfer) [API](https://en.wikipedia.org/wiki/API)s are +commonly used to expose Machine Learning (ML) models to other services. This folder contains an example REST API +created using Flask to expose the YOLOv5s model from [PyTorch Hub](https://pytorch.org/hub/ultralytics_yolov5/). + +## Requirements + +[Flask](https://palletsprojects.com/p/flask/) is required. Install with: + +```shell +$ pip install Flask +``` + +## Run + +After Flask installation run: + +```shell +$ python3 restapi.py --port 5000 +``` + +Then use [curl](https://curl.se/) to perform a request: + +```shell +$ curl -X POST -F image=@zidane.jpg 'http://localhost:5000/v1/object-detection/yolov5s' +``` + +The model inference results are returned as a JSON response: + +```json +[ + { + "class": 0, + "confidence": 0.8900438547, + "height": 0.9318675399, + "name": "person", + "width": 0.3264600933, + "xcenter": 0.7438579798, + "ycenter": 0.5207948685 + }, + { + "class": 0, + "confidence": 0.8440024257, + "height": 0.7155083418, + "name": "person", + "width": 0.6546785235, + "xcenter": 0.427829951, + "ycenter": 0.6334488392 + }, + { + "class": 27, + "confidence": 0.3771208823, + "height": 0.3902671337, + "name": "tie", + "width": 0.0696444362, + "xcenter": 0.3675483763, + "ycenter": 0.7991207838 + }, + { + "class": 27, + "confidence": 0.3527112305, + "height": 0.1540903747, + "name": "tie", + "width": 0.0336618312, + "xcenter": 0.7814827561, + "ycenter": 0.5065554976 + } +] +``` + +An example python script to perform inference using [requests](https://docs.python-requests.org/en/master/) is given +in `example_request.py` diff --git a/utils/flask_rest_api/example_request.py b/utils/flask_rest_api/example_request.py new file mode 100644 index 0000000..ff21f30 --- /dev/null +++ b/utils/flask_rest_api/example_request.py @@ -0,0 +1,13 @@ +"""Perform test request""" +import pprint + +import requests + +DETECTION_URL = "http://localhost:5000/v1/object-detection/yolov5s" +TEST_IMAGE = "zidane.jpg" + +image_data = open(TEST_IMAGE, "rb").read() + +response = requests.post(DETECTION_URL, files={"image": image_data}).json() + +pprint.pprint(response) diff --git a/utils/flask_rest_api/restapi.py b/utils/flask_rest_api/restapi.py new file mode 100644 index 0000000..b93ad16 --- /dev/null +++ b/utils/flask_rest_api/restapi.py @@ -0,0 +1,37 @@ +""" +Run a rest API exposing the yolov5s object detection model +""" +import argparse +import io + +import torch +from flask import Flask, request +from PIL import Image + +app = Flask(__name__) + +DETECTION_URL = "/v1/object-detection/yolov5s" + + +@app.route(DETECTION_URL, methods=["POST"]) +def predict(): + if not request.method == "POST": + return + + if request.files.get("image"): + image_file = request.files["image"] + image_bytes = image_file.read() + + img = Image.open(io.BytesIO(image_bytes)) + + results = model(img, size=640) # reduce size=320 for faster inference + return results.pandas().xyxy[0].to_json(orient="records") + + +if __name__ == "__main__": + parser = argparse.ArgumentParser(description="Flask API exposing YOLOv5 model") + parser.add_argument("--port", default=5000, type=int, help="port number") + args = parser.parse_args() + + model = torch.hub.load("ultralytics/yolov5", "yolov5s", force_reload=True) # force_reload to recache + app.run(host="0.0.0.0", port=args.port) # debug=True causes Restarting with stat diff --git a/utils/general.py b/utils/general.py new file mode 100644 index 0000000..3044b9c --- /dev/null +++ b/utils/general.py @@ -0,0 +1,880 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +General utils +""" + +import contextlib +import glob +import logging +import math +import os +import platform +import random +import re +import shutil +import signal +import time +import urllib +from itertools import repeat +from multiprocessing.pool import ThreadPool +from pathlib import Path +from subprocess import check_output +from zipfile import ZipFile + +import cv2 +import numpy as np +import pandas as pd +import pkg_resources as pkg +import torch +import torchvision +import yaml + +from utils.downloads import gsutil_getsize +from utils.metrics import box_iou, fitness + +# Settings +FILE = Path(__file__).resolve() +ROOT = FILE.parents[1] # YOLOv5 root directory +DATASETS_DIR = ROOT.parent / 'datasets' # YOLOv5 datasets directory +NUM_THREADS = min(8, max(1, os.cpu_count() - 1)) # number of YOLOv5 multiprocessing threads +VERBOSE = str(os.getenv('YOLOv5_VERBOSE', True)).lower() == 'true' # global verbose mode +FONT = 'Arial.ttf' # https://ultralytics.com/assets/Arial.ttf + +torch.set_printoptions(linewidth=320, precision=5, profile='long') +np.set_printoptions(linewidth=320, formatter={'float_kind': '{:11.5g}'.format}) # format short g, %precision=5 +pd.options.display.max_columns = 10 +cv2.setNumThreads(0) # prevent OpenCV from multithreading (incompatible with PyTorch DataLoader) +os.environ['NUMEXPR_MAX_THREADS'] = str(NUM_THREADS) # NumExpr max threads + + +def is_kaggle(): + # Is environment a Kaggle Notebook? + try: + assert os.environ.get('PWD') == '/kaggle/working' + assert os.environ.get('KAGGLE_URL_BASE') == 'https://www.kaggle.com' + return True + except AssertionError: + return False + + +def is_writeable(dir, test=False): + # Return True if directory has write permissions, test opening a file with write permissions if test=True + if test: # method 1 + file = Path(dir) / 'tmp.txt' + try: + with open(file, 'w'): # open file with write permissions + pass + file.unlink() # remove file + return True + except OSError: + return False + else: # method 2 + return os.access(dir, os.R_OK) # possible issues on Windows + + +def set_logging(name=None, verbose=VERBOSE): + # Sets level and returns logger + if is_kaggle(): + for h in logging.root.handlers: + logging.root.removeHandler(h) # remove all handlers associated with the root logger object + rank = int(os.getenv('RANK', -1)) # rank in world for Multi-GPU trainings + logging.basicConfig(format="%(message)s", level=logging.INFO if (verbose and rank in (-1, 0)) else logging.WARNING) + return logging.getLogger(name) + + +LOGGER = set_logging('yolov5') # define globally (used in train.py, val.py, detect.py, etc.) + + +def user_config_dir(dir='Ultralytics', env_var='YOLOV5_CONFIG_DIR'): + # Return path of user configuration directory. Prefer environment variable if exists. Make dir if required. + env = os.getenv(env_var) + if env: + path = Path(env) # use environment variable + else: + cfg = {'Windows': 'AppData/Roaming', 'Linux': '.config', 'Darwin': 'Library/Application Support'} # 3 OS dirs + path = Path.home() / cfg.get(platform.system(), '') # OS-specific config dir + path = (path if is_writeable(path) else Path('/tmp')) / dir # GCP and AWS lambda fix, only /tmp is writeable + path.mkdir(exist_ok=True) # make if required + return path + + +CONFIG_DIR = user_config_dir() # Ultralytics settings dir + + +class Profile(contextlib.ContextDecorator): + # Usage: @Profile() decorator or 'with Profile():' context manager + def __enter__(self): + self.start = time.time() + + def __exit__(self, type, value, traceback): + print(f'Profile results: {time.time() - self.start:.5f}s') + + +class Timeout(contextlib.ContextDecorator): + # Usage: @Timeout(seconds) decorator or 'with Timeout(seconds):' context manager + def __init__(self, seconds, *, timeout_msg='', suppress_timeout_errors=True): + self.seconds = int(seconds) + self.timeout_message = timeout_msg + self.suppress = bool(suppress_timeout_errors) + + def _timeout_handler(self, signum, frame): + raise TimeoutError(self.timeout_message) + + def __enter__(self): + signal.signal(signal.SIGALRM, self._timeout_handler) # Set handler for SIGALRM + signal.alarm(self.seconds) # start countdown for SIGALRM to be raised + + def __exit__(self, exc_type, exc_val, exc_tb): + signal.alarm(0) # Cancel SIGALRM if it's scheduled + if self.suppress and exc_type is TimeoutError: # Suppress TimeoutError + return True + + +class WorkingDirectory(contextlib.ContextDecorator): + # Usage: @WorkingDirectory(dir) decorator or 'with WorkingDirectory(dir):' context manager + def __init__(self, new_dir): + self.dir = new_dir # new dir + self.cwd = Path.cwd().resolve() # current dir + + def __enter__(self): + os.chdir(self.dir) + + def __exit__(self, exc_type, exc_val, exc_tb): + os.chdir(self.cwd) + + +def try_except(func): + # try-except function. Usage: @try_except decorator + def handler(*args, **kwargs): + try: + func(*args, **kwargs) + except Exception as e: + print(e) + + return handler + + +def methods(instance): + # Get class/instance methods + return [f for f in dir(instance) if callable(getattr(instance, f)) and not f.startswith("__")] + + +def print_args(name, opt): + # Print argparser arguments + LOGGER.info(colorstr(f'{name}: ') + ', '.join(f'{k}={v}' for k, v in vars(opt).items())) + + +def init_seeds(seed=0): + # Initialize random number generator (RNG) seeds https://pytorch.org/docs/stable/notes/randomness.html + # cudnn seed 0 settings are slower and more reproducible, else faster and less reproducible + import torch.backends.cudnn as cudnn + random.seed(seed) + np.random.seed(seed) + torch.manual_seed(seed) + cudnn.benchmark, cudnn.deterministic = (False, True) if seed == 0 else (True, False) + + +def intersect_dicts(da, db, exclude=()): + # Dictionary intersection of matching keys and shapes, omitting 'exclude' keys, using da values + return {k: v for k, v in da.items() if k in db and not any(x in k for x in exclude) and v.shape == db[k].shape} + + +def get_latest_run(search_dir='.'): + # Return path to most recent 'last.pt' in /runs (i.e. to --resume from) + last_list = glob.glob(f'{search_dir}/**/last*.pt', recursive=True) + return max(last_list, key=os.path.getctime) if last_list else '' + + +def is_docker(): + # Is environment a Docker container? + return Path('/workspace').exists() # or Path('/.dockerenv').exists() + + +def is_colab(): + # Is environment a Google Colab instance? + try: + import google.colab + return True + except ImportError: + return False + + +def is_pip(): + # Is file in a pip package? + return 'site-packages' in Path(__file__).resolve().parts + + +def is_ascii(s=''): + # Is string composed of all ASCII (no UTF) characters? (note str().isascii() introduced in python 3.7) + s = str(s) # convert list, tuple, None, etc. to str + return len(s.encode().decode('ascii', 'ignore')) == len(s) + + +def is_chinese(s='äēēåˇĨæ™ēčƒŊ'): + # Is string composed of any Chinese characters? + return True if re.search('[\u4e00-\u9fff]', str(s)) else False + + +def emojis(str=''): + # Return platform-dependent emoji-safe version of string + return str.encode().decode('ascii', 'ignore') if platform.system() == 'Windows' else str + + +def file_size(path): + # Return file/dir size (MB) + path = Path(path) + if path.is_file(): + return path.stat().st_size / 1E6 + elif path.is_dir(): + return sum(f.stat().st_size for f in path.glob('**/*') if f.is_file()) / 1E6 + else: + return 0.0 + + +def check_online(): + # Check internet connectivity + import socket + try: + socket.create_connection(("1.1.1.1", 443), 5) # check host accessibility + return True + except OSError: + return False + + +@try_except +@WorkingDirectory(ROOT) +def check_git_status(): + # Recommend 'git pull' if code is out of date + msg = ', for updates see https://github.com/ultralytics/yolov5' + s = colorstr('github: ') # string + assert Path('.git').exists(), s + 'skipping check (not a git repository)' + msg + assert not is_docker(), s + 'skipping check (Docker image)' + msg + assert check_online(), s + 'skipping check (offline)' + msg + + cmd = 'git fetch && git config --get remote.origin.url' + url = check_output(cmd, shell=True, timeout=5).decode().strip().rstrip('.git') # git fetch + branch = check_output('git rev-parse --abbrev-ref HEAD', shell=True).decode().strip() # checked out + n = int(check_output(f'git rev-list {branch}..origin/master --count', shell=True)) # commits behind + if n > 0: + s += f"⚠ī¸ YOLOv5 is out of date by {n} commit{'s' * (n > 1)}. Use `git pull` or `git clone {url}` to update." + else: + s += f'up to date with {url} ✅' + LOGGER.info(emojis(s)) # emoji-safe + + +def check_python(minimum='3.6.2'): + # Check current python version vs. required python version + check_version(platform.python_version(), minimum, name='Python ', hard=True) + + +def check_version(current='0.0.0', minimum='0.0.0', name='version ', pinned=False, hard=False, verbose=False): + # Check version vs. required version + current, minimum = (pkg.parse_version(x) for x in (current, minimum)) + result = (current == minimum) if pinned else (current >= minimum) # bool + s = f'{name}{minimum} required by YOLOv5, but {name}{current} is currently installed' # string + if hard: + assert result, s # assert min requirements met + if verbose and not result: + LOGGER.warning(s) + return result + + +@try_except +def check_requirements(requirements=ROOT / 'requirements.txt', exclude=(), install=True): + # Check installed dependencies meet requirements (pass *.txt file or list of packages) + prefix = colorstr('red', 'bold', 'requirements:') + check_python() # check python version + if isinstance(requirements, (str, Path)): # requirements.txt file + file = Path(requirements) + assert file.exists(), f"{prefix} {file.resolve()} not found, check failed." + with file.open() as f: + requirements = [f'{x.name}{x.specifier}' for x in pkg.parse_requirements(f) if x.name not in exclude] + else: # list or tuple of packages + requirements = [x for x in requirements if x not in exclude] + + n = 0 # number of packages updates + for r in requirements: + try: + pkg.require(r) + except Exception: # DistributionNotFound or VersionConflict if requirements not met + s = f"{prefix} {r} not found and is required by YOLOv5" + if install: + LOGGER.info(f"{s}, attempting auto-update...") + try: + assert check_online(), f"'pip install {r}' skipped (offline)" + LOGGER.info(check_output(f"pip install '{r}'", shell=True).decode()) + n += 1 + except Exception as e: + LOGGER.warning(f'{prefix} {e}') + else: + LOGGER.info(f'{s}. Please install and rerun your command.') + + if n: # if packages updated + source = file.resolve() if 'file' in locals() else requirements + s = f"{prefix} {n} package{'s' * (n > 1)} updated per {source}\n" \ + f"{prefix} ⚠ī¸ {colorstr('bold', 'Restart runtime or rerun command for updates to take effect')}\n" + LOGGER.info(emojis(s)) + + +def check_img_size(imgsz, s=32, floor=0): + # Verify image size is a multiple of stride s in each dimension + if isinstance(imgsz, int): # integer i.e. img_size=640 + new_size = max(make_divisible(imgsz, int(s)), floor) + else: # list i.e. img_size=[640, 480] + new_size = [max(make_divisible(x, int(s)), floor) for x in imgsz] + if new_size != imgsz: + LOGGER.warning(f'WARNING: --img-size {imgsz} must be multiple of max stride {s}, updating to {new_size}') + return new_size + + +def check_imshow(): + # Check if environment supports image displays + try: + assert not is_docker(), 'cv2.imshow() is disabled in Docker environments' + assert not is_colab(), 'cv2.imshow() is disabled in Google Colab environments' + cv2.imshow('test', np.zeros((1, 1, 3))) + cv2.waitKey(1) + cv2.destroyAllWindows() + cv2.waitKey(1) + return True + except Exception as e: + LOGGER.warning(f'WARNING: Environment does not support cv2.imshow() or PIL Image.show() image displays\n{e}') + return False + + +def check_suffix(file='yolov5s.pt', suffix=('.pt',), msg=''): + # Check file(s) for acceptable suffix + if file and suffix: + if isinstance(suffix, str): + suffix = [suffix] + for f in file if isinstance(file, (list, tuple)) else [file]: + s = Path(f).suffix.lower() # file suffix + if len(s): + assert s in suffix, f"{msg}{f} acceptable suffix is {suffix}" + + +def check_yaml(file, suffix=('.yaml', '.yml')): + # Search/download YAML file (if necessary) and return path, checking suffix + return check_file(file, suffix) + + +def check_file(file, suffix=''): + # Search/download file (if necessary) and return path + check_suffix(file, suffix) # optional + file = str(file) # convert to str() + if Path(file).is_file() or file == '': # exists + return file + elif file.startswith(('http:/', 'https:/')): # download + url = str(Path(file)).replace(':/', '://') # Pathlib turns :// -> :/ + file = Path(urllib.parse.unquote(file).split('?')[0]).name # '%2F' to '/', split https://url.com/file.txt?auth + if Path(file).is_file(): + LOGGER.info(f'Found {url} locally at {file}') # file already exists + else: + LOGGER.info(f'Downloading {url} to {file}...') + torch.hub.download_url_to_file(url, file) + assert Path(file).exists() and Path(file).stat().st_size > 0, f'File download failed: {url}' # check + return file + else: # search + files = [] + for d in 'data', 'models', 'utils': # search directories + files.extend(glob.glob(str(ROOT / d / '**' / file), recursive=True)) # find file + assert len(files), f'File not found: {file}' # assert file was found + assert len(files) == 1, f"Multiple files match '{file}', specify exact path: {files}" # assert unique + return files[0] # return file + + +def check_font(font=FONT): + # Download font to CONFIG_DIR if necessary + font = Path(font) + if not font.exists() and not (CONFIG_DIR / font.name).exists(): + url = "https://ultralytics.com/assets/" + font.name + LOGGER.info(f'Downloading {url} to {CONFIG_DIR / font.name}...') + torch.hub.download_url_to_file(url, str(font), progress=False) + + +def check_dataset(data, autodownload=True): + # Download and/or unzip dataset if not found locally + # Usage: https://github.com/ultralytics/yolov5/releases/download/v1.0/coco128_with_yaml.zip + + # Download (optional) + extract_dir = '' + if isinstance(data, (str, Path)) and str(data).endswith('.zip'): # i.e. gs://bucket/dir/coco128.zip + download(data, dir=DATASETS_DIR, unzip=True, delete=False, curl=False, threads=1) + data = next((DATASETS_DIR / Path(data).stem).rglob('*.yaml')) + extract_dir, autodownload = data.parent, False + + # Read yaml (optional) + if isinstance(data, (str, Path)): + with open(data, errors='ignore') as f: + data = yaml.safe_load(f) # dictionary + + # Resolve paths + path = Path(extract_dir or data.get('path') or '') # optional 'path' default to '.' + if not path.is_absolute(): + path = (ROOT / path).resolve() + for k in 'train', 'val', 'test': + if data.get(k): # prepend path + data[k] = str(path / data[k]) if isinstance(data[k], str) else [str(path / x) for x in data[k]] + + # Parse yaml + assert 'nc' in data, "Dataset 'nc' key missing." + if 'names' not in data: + data['names'] = [f'class{i}' for i in range(data['nc'])] # assign class names if missing + train, val, test, s = (data.get(x) for x in ('train', 'val', 'test', 'download')) + if val: + val = [Path(x).resolve() for x in (val if isinstance(val, list) else [val])] # val path + if not all(x.exists() for x in val): + LOGGER.info('\nDataset not found, missing paths: %s' % [str(x) for x in val if not x.exists()]) + if s and autodownload: # download script + root = path.parent if 'path' in data else '..' # unzip directory i.e. '../' + if s.startswith('http') and s.endswith('.zip'): # URL + f = Path(s).name # filename + LOGGER.info(f'Downloading {s} to {f}...') + torch.hub.download_url_to_file(s, f) + Path(root).mkdir(parents=True, exist_ok=True) # create root + ZipFile(f).extractall(path=root) # unzip + Path(f).unlink() # remove zip + r = None # success + elif s.startswith('bash '): # bash script + LOGGER.info(f'Running {s} ...') + r = os.system(s) + else: # python script + r = exec(s, {'yaml': data}) # return None + LOGGER.info(f"Dataset autodownload {f'success, saved to {root}' if r in (0, None) else 'failure'}\n") + else: + raise Exception('Dataset not found.') + + return data # dictionary + + +def url2file(url): + # Convert URL to filename, i.e. https://url.com/file.txt?auth -> file.txt + url = str(Path(url)).replace(':/', '://') # Pathlib turns :// -> :/ + file = Path(urllib.parse.unquote(url)).name.split('?')[0] # '%2F' to '/', split https://url.com/file.txt?auth + return file + + +def download(url, dir='.', unzip=True, delete=True, curl=False, threads=1): + # Multi-threaded file download and unzip function, used in data.yaml for autodownload + def download_one(url, dir): + # Download 1 file + f = dir / Path(url).name # filename + if Path(url).is_file(): # exists in current path + Path(url).rename(f) # move to dir + elif not f.exists(): + LOGGER.info(f'Downloading {url} to {f}...') + if curl: + os.system(f"curl -L '{url}' -o '{f}' --retry 9 -C -") # curl download, retry and resume on fail + else: + torch.hub.download_url_to_file(url, f, progress=True) # torch download + if unzip and f.suffix in ('.zip', '.gz'): + LOGGER.info(f'Unzipping {f}...') + if f.suffix == '.zip': + ZipFile(f).extractall(path=dir) # unzip + elif f.suffix == '.gz': + os.system(f'tar xfz {f} --directory {f.parent}') # unzip + if delete: + f.unlink() # remove zip + + dir = Path(dir) + dir.mkdir(parents=True, exist_ok=True) # make directory + if threads > 1: + pool = ThreadPool(threads) + pool.imap(lambda x: download_one(*x), zip(url, repeat(dir))) # multi-threaded + pool.close() + pool.join() + else: + for u in [url] if isinstance(url, (str, Path)) else url: + download_one(u, dir) + + +def make_divisible(x, divisor): + # Returns nearest x divisible by divisor + if isinstance(divisor, torch.Tensor): + divisor = int(divisor.max()) # to int + return math.ceil(x / divisor) * divisor + + +def clean_str(s): + # Cleans a string by replacing special characters with underscore _ + return re.sub(pattern="[|@#!¥¡$â‚Ŧ%&()=?Âŋ^*;:,¨´><+]", repl="_", string=s) + + +def one_cycle(y1=0.0, y2=1.0, steps=100): + # lambda function for sinusoidal ramp from y1 to y2 https://arxiv.org/pdf/1812.01187.pdf + return lambda x: ((1 - math.cos(x * math.pi / steps)) / 2) * (y2 - y1) + y1 + + +def colorstr(*input): + # Colors a string https://en.wikipedia.org/wiki/ANSI_escape_code, i.e. colorstr('blue', 'hello world') + *args, string = input if len(input) > 1 else ('blue', 'bold', input[0]) # color arguments, string + colors = {'black': '\033[30m', # basic colors + 'red': '\033[31m', + 'green': '\033[32m', + 'yellow': '\033[33m', + 'blue': '\033[34m', + 'magenta': '\033[35m', + 'cyan': '\033[36m', + 'white': '\033[37m', + 'bright_black': '\033[90m', # bright colors + 'bright_red': '\033[91m', + 'bright_green': '\033[92m', + 'bright_yellow': '\033[93m', + 'bright_blue': '\033[94m', + 'bright_magenta': '\033[95m', + 'bright_cyan': '\033[96m', + 'bright_white': '\033[97m', + 'end': '\033[0m', # misc + 'bold': '\033[1m', + 'underline': '\033[4m'} + return ''.join(colors[x] for x in args) + f'{string}' + colors['end'] + + +def labels_to_class_weights(labels, nc=80): + # Get class weights (inverse frequency) from training labels + if labels[0] is None: # no labels loaded + return torch.Tensor() + + labels = np.concatenate(labels, 0) # labels.shape = (866643, 5) for COCO + classes = labels[:, 0].astype(np.int) # labels = [class xywh] + weights = np.bincount(classes, minlength=nc) # occurrences per class + + # Prepend gridpoint count (for uCE training) + # gpi = ((320 / 32 * np.array([1, 2, 4])) ** 2 * 3).sum() # gridpoints per image + # weights = np.hstack([gpi * len(labels) - weights.sum() * 9, weights * 9]) ** 0.5 # prepend gridpoints to start + + weights[weights == 0] = 1 # replace empty bins with 1 + weights = 1 / weights # number of targets per class + weights /= weights.sum() # normalize + return torch.from_numpy(weights) + + +def labels_to_image_weights(labels, nc=80, class_weights=np.ones(80)): + # Produces image weights based on class_weights and image contents + class_counts = np.array([np.bincount(x[:, 0].astype(np.int), minlength=nc) for x in labels]) + image_weights = (class_weights.reshape(1, nc) * class_counts).sum(1) + # index = random.choices(range(n), weights=image_weights, k=1) # weight image sample + return image_weights + + +def coco80_to_coco91_class(): # converts 80-index (val2014) to 91-index (paper) + # https://tech.amikelive.com/node-718/what-object-categories-labels-are-in-coco-dataset/ + # a = np.loadtxt('data/coco.names', dtype='str', delimiter='\n') + # b = np.loadtxt('data/coco_paper.names', dtype='str', delimiter='\n') + # x1 = [list(a[i] == b).index(True) + 1 for i in range(80)] # darknet to coco + # x2 = [list(b[i] == a).index(True) if any(b[i] == a) else None for i in range(91)] # coco to darknet + x = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 27, 28, 31, 32, 33, 34, + 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, + 64, 65, 67, 70, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 84, 85, 86, 87, 88, 89, 90] + return x + + +def xyxy2xywh(x): + # Convert nx4 boxes from [x1, y1, x2, y2] to [x, y, w, h] where xy1=top-left, xy2=bottom-right + y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x) + y[:, 0] = (x[:, 0] + x[:, 2]) / 2 # x center + y[:, 1] = (x[:, 1] + x[:, 3]) / 2 # y center + y[:, 2] = x[:, 2] - x[:, 0] # width + y[:, 3] = x[:, 3] - x[:, 1] # height + return y + + +def xywh2xyxy(x): + # Convert nx4 boxes from [x, y, w, h] to [x1, y1, x2, y2] where xy1=top-left, xy2=bottom-right + y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x) + y[:, 0] = x[:, 0] - x[:, 2] / 2 # top left x + y[:, 1] = x[:, 1] - x[:, 3] / 2 # top left y + y[:, 2] = x[:, 0] + x[:, 2] / 2 # bottom right x + y[:, 3] = x[:, 1] + x[:, 3] / 2 # bottom right y + return y + + +def xywhn2xyxy(x, w=640, h=640, padw=0, padh=0): + # Convert nx4 boxes from [x, y, w, h] normalized to [x1, y1, x2, y2] where xy1=top-left, xy2=bottom-right + y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x) + y[:, 0] = w * (x[:, 0] - x[:, 2] / 2) + padw # top left x + y[:, 1] = h * (x[:, 1] - x[:, 3] / 2) + padh # top left y + y[:, 2] = w * (x[:, 0] + x[:, 2] / 2) + padw # bottom right x + y[:, 3] = h * (x[:, 1] + x[:, 3] / 2) + padh # bottom right y + return y + + +def xyxy2xywhn(x, w=640, h=640, clip=False, eps=0.0): + # Convert nx4 boxes from [x1, y1, x2, y2] to [x, y, w, h] normalized where xy1=top-left, xy2=bottom-right + if clip: + clip_coords(x, (h - eps, w - eps)) # warning: inplace clip + y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x) + y[:, 0] = ((x[:, 0] + x[:, 2]) / 2) / w # x center + y[:, 1] = ((x[:, 1] + x[:, 3]) / 2) / h # y center + y[:, 2] = (x[:, 2] - x[:, 0]) / w # width + y[:, 3] = (x[:, 3] - x[:, 1]) / h # height + return y + + +def xyn2xy(x, w=640, h=640, padw=0, padh=0): + # Convert normalized segments into pixel segments, shape (n,2) + y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x) + y[:, 0] = w * x[:, 0] + padw # top left x + y[:, 1] = h * x[:, 1] + padh # top left y + return y + + +def segment2box(segment, width=640, height=640): + # Convert 1 segment label to 1 box label, applying inside-image constraint, i.e. (xy1, xy2, ...) to (xyxy) + x, y = segment.T # segment xy + inside = (x >= 0) & (y >= 0) & (x <= width) & (y <= height) + x, y, = x[inside], y[inside] + return np.array([x.min(), y.min(), x.max(), y.max()]) if any(x) else np.zeros((1, 4)) # xyxy + + +def segments2boxes(segments): + # Convert segment labels to box labels, i.e. (cls, xy1, xy2, ...) to (cls, xywh) + boxes = [] + for s in segments: + x, y = s.T # segment xy + boxes.append([x.min(), y.min(), x.max(), y.max()]) # cls, xyxy + return xyxy2xywh(np.array(boxes)) # cls, xywh + + +def resample_segments(segments, n=1000): + # Up-sample an (n,2) segment + for i, s in enumerate(segments): + x = np.linspace(0, len(s) - 1, n) + xp = np.arange(len(s)) + segments[i] = np.concatenate([np.interp(x, xp, s[:, i]) for i in range(2)]).reshape(2, -1).T # segment xy + return segments + + +def scale_coords(img1_shape, coords, img0_shape, ratio_pad=None): + # Rescale coords (xyxy) from img1_shape to img0_shape + if ratio_pad is None: # calculate from img0_shape + gain = min(img1_shape[0] / img0_shape[0], img1_shape[1] / img0_shape[1]) # gain = old / new + pad = (img1_shape[1] - img0_shape[1] * gain) / 2, (img1_shape[0] - img0_shape[0] * gain) / 2 # wh padding + else: + gain = ratio_pad[0][0] + pad = ratio_pad[1] + + coords[:, [0, 2]] -= pad[0] # x padding + coords[:, [1, 3]] -= pad[1] # y padding + coords[:, :4] /= gain + clip_coords(coords, img0_shape) + return coords + + +def clip_coords(boxes, shape): + # Clip bounding xyxy bounding boxes to image shape (height, width) + if isinstance(boxes, torch.Tensor): # faster individually + boxes[:, 0].clamp_(0, shape[1]) # x1 + boxes[:, 1].clamp_(0, shape[0]) # y1 + boxes[:, 2].clamp_(0, shape[1]) # x2 + boxes[:, 3].clamp_(0, shape[0]) # y2 + else: # np.array (faster grouped) + boxes[:, [0, 2]] = boxes[:, [0, 2]].clip(0, shape[1]) # x1, x2 + boxes[:, [1, 3]] = boxes[:, [1, 3]].clip(0, shape[0]) # y1, y2 + + +def non_max_suppression(prediction, conf_thres=0.25, iou_thres=0.45, classes=None, agnostic=False, multi_label=False, + labels=(), max_det=300): + """Runs Non-Maximum Suppression (NMS) on inference results + + Returns: + list of detections, on (n,6) tensor per image [xyxy, conf, cls] + """ + + nc = prediction.shape[2] - 5 # number of classes + xc = prediction[..., 4] > conf_thres # candidates + + # Checks + assert 0 <= conf_thres <= 1, f'Invalid Confidence threshold {conf_thres}, valid values are between 0.0 and 1.0' + assert 0 <= iou_thres <= 1, f'Invalid IoU {iou_thres}, valid values are between 0.0 and 1.0' + + # Settings + min_wh, max_wh = 2, 7680 # (pixels) minimum and maximum box width and height + max_nms = 30000 # maximum number of boxes into torchvision.ops.nms() + time_limit = 10.0 # seconds to quit after + redundant = True # require redundant detections + multi_label &= nc > 1 # multiple labels per box (adds 0.5ms/img) + merge = False # use merge-NMS + + t = time.time() + output = [torch.zeros((0, 6), device=prediction.device)] * prediction.shape[0] + for xi, x in enumerate(prediction): # image index, image inference + # Apply constraints + x[((x[..., 2:4] < min_wh) | (x[..., 2:4] > max_wh)).any(1), 4] = 0 # width-height + x = x[xc[xi]] # confidence + + # Cat apriori labels if autolabelling + if labels and len(labels[xi]): + lb = labels[xi] + v = torch.zeros((len(lb), nc + 5), device=x.device) + v[:, :4] = lb[:, 1:5] # box + v[:, 4] = 1.0 # conf + v[range(len(lb)), lb[:, 0].long() + 5] = 1.0 # cls + x = torch.cat((x, v), 0) + + # If none remain process next image + if not x.shape[0]: + continue + + # Compute conf + x[:, 5:] *= x[:, 4:5] # conf = obj_conf * cls_conf + + # Box (center x, center y, width, height) to (x1, y1, x2, y2) + box = xywh2xyxy(x[:, :4]) + + # Detections matrix nx6 (xyxy, conf, cls) + if multi_label: + i, j = (x[:, 5:] > conf_thres).nonzero(as_tuple=False).T + x = torch.cat((box[i], x[i, j + 5, None], j[:, None].float()), 1) + else: # best class only + conf, j = x[:, 5:].max(1, keepdim=True) + x = torch.cat((box, conf, j.float()), 1)[conf.view(-1) > conf_thres] + + # Filter by class + if classes is not None: + x = x[(x[:, 5:6] == torch.tensor(classes, device=x.device)).any(1)] + + # Apply finite constraint + # if not torch.isfinite(x).all(): + # x = x[torch.isfinite(x).all(1)] + + # Check shape + n = x.shape[0] # number of boxes + if not n: # no boxes + continue + elif n > max_nms: # excess boxes + x = x[x[:, 4].argsort(descending=True)[:max_nms]] # sort by confidence + + # Batched NMS + c = x[:, 5:6] * (0 if agnostic else max_wh) # classes + boxes, scores = x[:, :4] + c, x[:, 4] # boxes (offset by class), scores + i = torchvision.ops.nms(boxes, scores, iou_thres) # NMS + if i.shape[0] > max_det: # limit detections + i = i[:max_det] + if merge and (1 < n < 3E3): # Merge NMS (boxes merged using weighted mean) + # update boxes as boxes(i,4) = weights(i,n) * boxes(n,4) + iou = box_iou(boxes[i], boxes) > iou_thres # iou matrix + weights = iou * scores[None] # box weights + x[i, :4] = torch.mm(weights, x[:, :4]).float() / weights.sum(1, keepdim=True) # merged boxes + if redundant: + i = i[iou.sum(1) > 1] # require redundancy + + output[xi] = x[i] + if (time.time() - t) > time_limit: + LOGGER.warning(f'WARNING: NMS time limit {time_limit}s exceeded') + break # time limit exceeded + + return output + + +def strip_optimizer(f='best.pt', s=''): # from utils.general import *; strip_optimizer() + # Strip optimizer from 'f' to finalize training, optionally save as 's' + x = torch.load(f, map_location=torch.device('cpu')) + if x.get('ema'): + x['model'] = x['ema'] # replace model with ema + for k in 'optimizer', 'best_fitness', 'wandb_id', 'ema', 'updates': # keys + x[k] = None + x['epoch'] = -1 + x['model'].half() # to FP16 + for p in x['model'].parameters(): + p.requires_grad = False + torch.save(x, s or f) + mb = os.path.getsize(s or f) / 1E6 # filesize + LOGGER.info(f"Optimizer stripped from {f},{(' saved as %s,' % s) if s else ''} {mb:.1f}MB") + + +def print_mutation(results, hyp, save_dir, bucket, prefix=colorstr('evolve: ')): + evolve_csv = save_dir / 'evolve.csv' + evolve_yaml = save_dir / 'hyp_evolve.yaml' + keys = ('metrics/precision', 'metrics/recall', 'metrics/mAP_0.5', 'metrics/mAP_0.5:0.95', + 'val/box_loss', 'val/obj_loss', 'val/cls_loss') + tuple(hyp.keys()) # [results + hyps] + keys = tuple(x.strip() for x in keys) + vals = results + tuple(hyp.values()) + n = len(keys) + + # Download (optional) + if bucket: + url = f'gs://{bucket}/evolve.csv' + if gsutil_getsize(url) > (evolve_csv.stat().st_size if evolve_csv.exists() else 0): + os.system(f'gsutil cp {url} {save_dir}') # download evolve.csv if larger than local + + # Log to evolve.csv + s = '' if evolve_csv.exists() else (('%20s,' * n % keys).rstrip(',') + '\n') # add header + with open(evolve_csv, 'a') as f: + f.write(s + ('%20.5g,' * n % vals).rstrip(',') + '\n') + + # Save yaml + with open(evolve_yaml, 'w') as f: + data = pd.read_csv(evolve_csv) + data = data.rename(columns=lambda x: x.strip()) # strip keys + i = np.argmax(fitness(data.values[:, :4])) # + generations = len(data) + f.write('# YOLOv5 Hyperparameter Evolution Results\n' + + f'# Best generation: {i}\n' + + f'# Last generation: {generations - 1}\n' + + '# ' + ', '.join(f'{x.strip():>20s}' for x in keys[:7]) + '\n' + + '# ' + ', '.join(f'{x:>20.5g}' for x in data.values[i, :7]) + '\n\n') + yaml.safe_dump(data.loc[i][7:].to_dict(), f, sort_keys=False) + + # Print to screen + LOGGER.info(prefix + f'{generations} generations finished, current result:\n' + + prefix + ', '.join(f'{x.strip():>20s}' for x in keys) + '\n' + + prefix + ', '.join(f'{x:20.5g}' for x in vals) + '\n\n') + + if bucket: + os.system(f'gsutil cp {evolve_csv} {evolve_yaml} gs://{bucket}') # upload + + +def apply_classifier(x, model, img, im0): + # Apply a second stage classifier to YOLO outputs + # Example model = torchvision.models.__dict__['efficientnet_b0'](pretrained=True).to(device).eval() + im0 = [im0] if isinstance(im0, np.ndarray) else im0 + for i, d in enumerate(x): # per image + if d is not None and len(d): + d = d.clone() + + # Reshape and pad cutouts + b = xyxy2xywh(d[:, :4]) # boxes + b[:, 2:] = b[:, 2:].max(1)[0].unsqueeze(1) # rectangle to square + b[:, 2:] = b[:, 2:] * 1.3 + 30 # pad + d[:, :4] = xywh2xyxy(b).long() + + # Rescale boxes from img_size to im0 size + scale_coords(img.shape[2:], d[:, :4], im0[i].shape) + + # Classes + pred_cls1 = d[:, 5].long() + ims = [] + for j, a in enumerate(d): # per item + cutout = im0[i][int(a[1]):int(a[3]), int(a[0]):int(a[2])] + im = cv2.resize(cutout, (224, 224)) # BGR + # cv2.imwrite('example%i.jpg' % j, cutout) + + im = im[:, :, ::-1].transpose(2, 0, 1) # BGR to RGB, to 3x416x416 + im = np.ascontiguousarray(im, dtype=np.float32) # uint8 to float32 + im /= 255 # 0 - 255 to 0.0 - 1.0 + ims.append(im) + + pred_cls2 = model(torch.Tensor(ims).to(d.device)).argmax(1) # classifier prediction + x[i] = x[i][pred_cls1 == pred_cls2] # retain matching class detections + + return x + + +def increment_path(path, exist_ok=False, sep='', mkdir=False): + # Increment file or directory path, i.e. runs/exp --> runs/exp{sep}2, runs/exp{sep}3, ... etc. + path = Path(path) # os-agnostic + if path.exists() and not exist_ok: + path, suffix = (path.with_suffix(''), path.suffix) if path.is_file() else (path, '') + dirs = glob.glob(f"{path}{sep}*") # similar paths + matches = [re.search(rf"%s{sep}(\d+)" % path.stem, d) for d in dirs] + i = [int(m.groups()[0]) for m in matches if m] # indices + n = max(i) + 1 if i else 2 # increment number + path = Path(f"{path}{sep}{n}{suffix}") # increment path + if mkdir: + path.mkdir(parents=True, exist_ok=True) # make directory + return path + + +# Variables +NCOLS = 0 if is_docker() else shutil.get_terminal_size().columns # terminal window size for tqdm diff --git a/utils/google_app_engine/Dockerfile b/utils/google_app_engine/Dockerfile new file mode 100644 index 0000000..0155618 --- /dev/null +++ b/utils/google_app_engine/Dockerfile @@ -0,0 +1,25 @@ +FROM gcr.io/google-appengine/python + +# Create a virtualenv for dependencies. This isolates these packages from +# system-level packages. +# Use -p python3 or -p python3.7 to select python version. Default is version 2. +RUN virtualenv /env -p python3 + +# Setting these environment variables are the same as running +# source /env/bin/activate. +ENV VIRTUAL_ENV /env +ENV PATH /env/bin:$PATH + +RUN apt-get update && apt-get install -y python-opencv + +# Copy the application's requirements.txt and run pip to install all +# dependencies into the virtualenv. +ADD requirements.txt /app/requirements.txt +RUN pip install -r /app/requirements.txt + +# Add the application source code. +ADD . /app + +# Run a WSGI server to serve the application. gunicorn must be declared as +# a dependency in requirements.txt. +CMD gunicorn -b :$PORT main:app diff --git a/utils/google_app_engine/additional_requirements.txt b/utils/google_app_engine/additional_requirements.txt new file mode 100644 index 0000000..42d7ffc --- /dev/null +++ b/utils/google_app_engine/additional_requirements.txt @@ -0,0 +1,4 @@ +# add these requirements in your app on top of the existing ones +pip==21.1 +Flask==1.0.2 +gunicorn==19.9.0 diff --git a/utils/google_app_engine/app.yaml b/utils/google_app_engine/app.yaml new file mode 100644 index 0000000..5056b7c --- /dev/null +++ b/utils/google_app_engine/app.yaml @@ -0,0 +1,14 @@ +runtime: custom +env: flex + +service: yolov5app + +liveness_check: + initial_delay_sec: 600 + +manual_scaling: + instances: 1 +resources: + cpu: 1 + memory_gb: 4 + disk_size_gb: 20 diff --git a/utils/loggers/__init__.py b/utils/loggers/__init__.py new file mode 100644 index 0000000..86ccf38 --- /dev/null +++ b/utils/loggers/__init__.py @@ -0,0 +1,168 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Logging utils +""" + +import os +import warnings +from threading import Thread + +import pkg_resources as pkg +import torch +from torch.utils.tensorboard import SummaryWriter + +from utils.general import colorstr, emojis +from utils.loggers.wandb.wandb_utils import WandbLogger +from utils.plots import plot_images, plot_results +from utils.torch_utils import de_parallel + +LOGGERS = ('csv', 'tb', 'wandb') # text-file, TensorBoard, Weights & Biases +RANK = int(os.getenv('RANK', -1)) + +try: + import wandb + + assert hasattr(wandb, '__version__') # verify package import not local dir + if pkg.parse_version(wandb.__version__) >= pkg.parse_version('0.12.2') and RANK in [0, -1]: + try: + wandb_login_success = wandb.login(timeout=30) + except wandb.errors.UsageError: # known non-TTY terminal issue + wandb_login_success = False + if not wandb_login_success: + wandb = None +except (ImportError, AssertionError): + wandb = None + + +class Loggers(): + # YOLOv5 Loggers class + def __init__(self, save_dir=None, weights=None, opt=None, hyp=None, logger=None, include=LOGGERS): + self.save_dir = save_dir + self.weights = weights + self.opt = opt + self.hyp = hyp + self.logger = logger # for printing results to console + self.include = include + self.keys = ['train/box_loss', 'train/obj_loss', 'train/cls_loss', # train loss + 'metrics/precision', 'metrics/recall', 'metrics/mAP_0.5', 'metrics/mAP_0.5:0.95', # metrics + 'val/box_loss', 'val/obj_loss', 'val/cls_loss', # val loss + 'x/lr0', 'x/lr1', 'x/lr2'] # params + self.best_keys = ['best/epoch', 'best/precision', 'best/recall', 'best/mAP_0.5', 'best/mAP_0.5:0.95',] + for k in LOGGERS: + setattr(self, k, None) # init empty logger dictionary + self.csv = True # always log to csv + + # Message + if not wandb: + prefix = colorstr('Weights & Biases: ') + s = f"{prefix}run 'pip install wandb' to automatically track and visualize YOLOv5 🚀 runs (RECOMMENDED)" + print(emojis(s)) + + # TensorBoard + s = self.save_dir + if 'tb' in self.include and not self.opt.evolve: + prefix = colorstr('TensorBoard: ') + self.logger.info(f"{prefix}Start with 'tensorboard --logdir {s.parent}', view at http://localhost:6006/") + self.tb = SummaryWriter(str(s)) + + # W&B + if wandb and 'wandb' in self.include: + wandb_artifact_resume = isinstance(self.opt.resume, str) and self.opt.resume.startswith('wandb-artifact://') + run_id = torch.load(self.weights).get('wandb_id') if self.opt.resume and not wandb_artifact_resume else None + self.opt.hyp = self.hyp # add hyperparameters + self.wandb = WandbLogger(self.opt, run_id) + else: + self.wandb = None + + def on_pretrain_routine_end(self): + # Callback runs on pre-train routine end + paths = self.save_dir.glob('*labels*.jpg') # training labels + if self.wandb: + self.wandb.log({"Labels": [wandb.Image(str(x), caption=x.name) for x in paths]}) + + def on_train_batch_end(self, ni, model, imgs, targets, paths, plots, sync_bn): + # Callback runs on train batch end + if plots: + if ni == 0: + if not sync_bn: # tb.add_graph() --sync known issue https://github.com/ultralytics/yolov5/issues/3754 + with warnings.catch_warnings(): + warnings.simplefilter('ignore') # suppress jit trace warning + self.tb.add_graph(torch.jit.trace(de_parallel(model), imgs[0:1], strict=False), []) + if ni < 3: + f = self.save_dir / f'train_batch{ni}.jpg' # filename + Thread(target=plot_images, args=(imgs, targets, paths, f), daemon=True).start() + if self.wandb and ni == 10: + files = sorted(self.save_dir.glob('train*.jpg')) + self.wandb.log({'Mosaics': [wandb.Image(str(f), caption=f.name) for f in files if f.exists()]}) + + def on_train_epoch_end(self, epoch): + # Callback runs on train epoch end + if self.wandb: + self.wandb.current_epoch = epoch + 1 + + def on_val_image_end(self, pred, predn, path, names, im): + # Callback runs on val image end + if self.wandb: + self.wandb.val_one_image(pred, predn, path, names, im) + + def on_val_end(self): + # Callback runs on val end + if self.wandb: + files = sorted(self.save_dir.glob('val*.jpg')) + self.wandb.log({"Validation": [wandb.Image(str(f), caption=f.name) for f in files]}) + + def on_fit_epoch_end(self, vals, epoch, best_fitness, fi): + # Callback runs at the end of each fit (train+val) epoch + x = {k: v for k, v in zip(self.keys, vals)} # dict + if self.csv: + file = self.save_dir / 'results.csv' + n = len(x) + 1 # number of cols + s = '' if file.exists() else (('%20s,' * n % tuple(['epoch'] + self.keys)).rstrip(',') + '\n') # add header + with open(file, 'a') as f: + f.write(s + ('%20.5g,' * n % tuple([epoch] + vals)).rstrip(',') + '\n') + + if self.tb: + for k, v in x.items(): + self.tb.add_scalar(k, v, epoch) + + if self.wandb: + if best_fitness == fi: + best_results = [epoch] + vals[3:7] + for i, name in enumerate(self.best_keys): + self.wandb.wandb_run.summary[name] = best_results[i] # log best results in the summary + self.wandb.log(x) + self.wandb.end_epoch(best_result=best_fitness == fi) + + def on_model_save(self, last, epoch, final_epoch, best_fitness, fi): + # Callback runs on model save event + if self.wandb: + if ((epoch + 1) % self.opt.save_period == 0 and not final_epoch) and self.opt.save_period != -1: + self.wandb.log_model(last.parent, self.opt, epoch, fi, best_model=best_fitness == fi) + + def on_train_end(self, last, best, plots, epoch, results): + # Callback runs on training end + if plots: + plot_results(file=self.save_dir / 'results.csv') # save results.png + files = ['results.png', 'confusion_matrix.png', *(f'{x}_curve.png' for x in ('F1', 'PR', 'P', 'R'))] + files = [(self.save_dir / f) for f in files if (self.save_dir / f).exists()] # filter + + if self.tb: + import cv2 + for f in files: + self.tb.add_image(f.stem, cv2.imread(str(f))[..., ::-1], epoch, dataformats='HWC') + + if self.wandb: + self.wandb.log({k: v for k, v in zip(self.keys[3:10], results)}) # log best.pt val results + self.wandb.log({"Results": [wandb.Image(str(f), caption=f.name) for f in files]}) + # Calling wandb.log. TODO: Refactor this into WandbLogger.log_model + if not self.opt.evolve: + wandb.log_artifact(str(best if best.exists() else last), type='model', + name='run_' + self.wandb.wandb_run.id + '_model', + aliases=['latest', 'best', 'stripped']) + self.wandb.finish_run() + + def on_params_update(self, params): + # Update hyperparams or configs of the experiment + # params: A dict containing {param: value} pairs + if self.wandb: + self.wandb.wandb_run.config.update(params, allow_val_change=True) diff --git a/utils/loggers/wandb/README.md b/utils/loggers/wandb/README.md new file mode 100644 index 0000000..63d9998 --- /dev/null +++ b/utils/loggers/wandb/README.md @@ -0,0 +1,152 @@ +📚 This guide explains how to use **Weights & Biases** (W&B) with YOLOv5 🚀. UPDATED 29 September 2021. +* [About Weights & Biases](#about-weights-&-biases) +* [First-Time Setup](#first-time-setup) +* [Viewing runs](#viewing-runs) +* [Disabling wandb](#disabling-wandb) +* [Advanced Usage: Dataset Versioning and Evaluation](#advanced-usage) +* [Reports: Share your work with the world!](#reports) + +## About Weights & Biases +Think of [W&B](https://wandb.ai/site?utm_campaign=repo_yolo_wandbtutorial) like GitHub for machine learning models. With a few lines of code, save everything you need to debug, compare and reproduce your models — architecture, hyperparameters, git commits, model weights, GPU usage, and even datasets and predictions. + +Used by top researchers including teams at OpenAI, Lyft, Github, and MILA, W&B is part of the new standard of best practices for machine learning. How W&B can help you optimize your machine learning workflows: + + * [Debug](https://wandb.ai/wandb/getting-started/reports/Visualize-Debug-Machine-Learning-Models--VmlldzoyNzY5MDk#Free-2) model performance in real time + * [GPU usage](https://wandb.ai/wandb/getting-started/reports/Visualize-Debug-Machine-Learning-Models--VmlldzoyNzY5MDk#System-4) visualized automatically + * [Custom charts](https://wandb.ai/wandb/customizable-charts/reports/Powerful-Custom-Charts-To-Debug-Model-Peformance--VmlldzoyNzY4ODI) for powerful, extensible visualization + * [Share insights](https://wandb.ai/wandb/getting-started/reports/Visualize-Debug-Machine-Learning-Models--VmlldzoyNzY5MDk#Share-8) interactively with collaborators + * [Optimize hyperparameters](https://docs.wandb.com/sweeps) efficiently + * [Track](https://docs.wandb.com/artifacts) datasets, pipelines, and production models + +## First-Time Setup +
+ Toggle Details +When you first train, W&B will prompt you to create a new account and will generate an **API key** for you. If you are an existing user you can retrieve your key from https://wandb.ai/authorize. This key is used to tell W&B where to log your data. You only need to supply your key once, and then it is remembered on the same device. + +W&B will create a cloud **project** (default is 'YOLOv5') for your training runs, and each new training run will be provided a unique run **name** within that project as project/name. You can also manually set your project and run name as: + + ```shell + $ python train.py --project ... --name ... + ``` + +YOLOv5 notebook example: Open In Colab Open In Kaggle +Screen Shot 2021-09-29 at 10 23 13 PM + + +
+ +## Viewing Runs +
+ Toggle Details +Run information streams from your environment to the W&B cloud console as you train. This allows you to monitor and even cancel runs in realtime . All important information is logged: + + * Training & Validation losses + * Metrics: Precision, Recall, mAP@0.5, mAP@0.5:0.95 + * Learning Rate over time + * A bounding box debugging panel, showing the training progress over time + * GPU: Type, **GPU Utilization**, power, temperature, **CUDA memory usage** + * System: Disk I/0, CPU utilization, RAM memory usage + * Your trained model as W&B Artifact + * Environment: OS and Python types, Git repository and state, **training command** + +

Weights & Biases dashboard

+
+ + ## Disabling wandb +* training after running `wandb disabled` inside that directory creates no wandb run +![Screenshot (84)](https://user-images.githubusercontent.com/15766192/143441777-c780bdd7-7cb4-4404-9559-b4316030a985.png) + +* To enable wandb again, run `wandb online` +![Screenshot (85)](https://user-images.githubusercontent.com/15766192/143441866-7191b2cb-22f0-4e0f-ae64-2dc47dc13078.png) + +## Advanced Usage +You can leverage W&B artifacts and Tables integration to easily visualize and manage your datasets, models and training evaluations. Here are some quick examples to get you started. +
+

1: Train and Log Evaluation simultaneousy

+ This is an extension of the previous section, but it'll also training after uploading the dataset. This also evaluation Table + Evaluation table compares your predictions and ground truths across the validation set for each epoch. It uses the references to the already uploaded datasets, + so no images will be uploaded from your system more than once. +
+ Usage + Code $ python train.py --upload_data val + +![Screenshot from 2021-11-21 17-40-06](https://user-images.githubusercontent.com/15766192/142761183-c1696d8c-3f38-45ab-991a-bb0dfd98ae7d.png) +
+ +

2. Visualize and Version Datasets

+ Log, visualize, dynamically query, and understand your data with W&B Tables. You can use the following command to log your dataset as a W&B Table. This will generate a {dataset}_wandb.yaml file which can be used to train from dataset artifact. +
+ Usage + Code $ python utils/logger/wandb/log_dataset.py --project ... --name ... --data .. + + ![Screenshot (64)](https://user-images.githubusercontent.com/15766192/128486078-d8433890-98a3-4d12-8986-b6c0e3fc64b9.png) +
+ +

3: Train using dataset artifact

+ When you upload a dataset as described in the first section, you get a new config file with an added `_wandb` to its name. This file contains the information that + can be used to train a model directly from the dataset artifact. This also logs evaluation +
+ Usage + Code $ python train.py --data {data}_wandb.yaml + +![Screenshot (72)](https://user-images.githubusercontent.com/15766192/128979739-4cf63aeb-a76f-483f-8861-1c0100b938a5.png) +
+ +

4: Save model checkpoints as artifacts

+ To enable saving and versioning checkpoints of your experiment, pass `--save_period n` with the base cammand, where `n` represents checkpoint interval. + You can also log both the dataset and model checkpoints simultaneously. If not passed, only the final model will be logged + +
+ Usage + Code $ python train.py --save_period 1 + +![Screenshot (68)](https://user-images.githubusercontent.com/15766192/128726138-ec6c1f60-639d-437d-b4ee-3acd9de47ef3.png) +
+ +
+ +

5: Resume runs from checkpoint artifacts.

+Any run can be resumed using artifacts if the --resume argument starts with wandb-artifact:// prefix followed by the run path, i.e, wandb-artifact://username/project/runid . This doesn't require the model checkpoint to be present on the local system. + +
+ Usage + Code $ python train.py --resume wandb-artifact://{run_path} + +![Screenshot (70)](https://user-images.githubusercontent.com/15766192/128728988-4e84b355-6c87-41ae-a591-14aecf45343e.png) +
+ +

6: Resume runs from dataset artifact & checkpoint artifacts.

+ Local dataset or model checkpoints are not required. This can be used to resume runs directly on a different device + The syntax is same as the previous section, but you'll need to lof both the dataset and model checkpoints as artifacts, i.e, set bot --upload_dataset or + train from _wandb.yaml file and set --save_period + +
+ Usage + Code $ python train.py --resume wandb-artifact://{run_path} + +![Screenshot (70)](https://user-images.githubusercontent.com/15766192/128728988-4e84b355-6c87-41ae-a591-14aecf45343e.png) +
+ + + +

Reports

+W&B Reports can be created from your saved runs for sharing online. Once a report is created you will receive a link you can use to publically share your results. Here is an example report created from the COCO128 tutorial trainings of all four YOLOv5 models ([link](https://wandb.ai/glenn-jocher/yolov5_tutorial/reports/YOLOv5-COCO128-Tutorial-Results--VmlldzozMDI5OTY)). + +Weights & Biases Reports + + +## Environments + +YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including [CUDA](https://developer.nvidia.com/cuda)/[CUDNN](https://developer.nvidia.com/cudnn), [Python](https://www.python.org/) and [PyTorch](https://pytorch.org/) preinstalled): + +- **Google Colab and Kaggle** notebooks with free GPU: Open In Colab Open In Kaggle +- **Google Cloud** Deep Learning VM. See [GCP Quickstart Guide](https://github.com/ultralytics/yolov5/wiki/GCP-Quickstart) +- **Amazon** Deep Learning AMI. See [AWS Quickstart Guide](https://github.com/ultralytics/yolov5/wiki/AWS-Quickstart) +- **Docker Image**. See [Docker Quickstart Guide](https://github.com/ultralytics/yolov5/wiki/Docker-Quickstart) Docker Pulls + + +## Status + +![CI CPU testing](https://github.com/ultralytics/yolov5/workflows/CI%20CPU%20testing/badge.svg) + +If this badge is green, all [YOLOv5 GitHub Actions](https://github.com/ultralytics/yolov5/actions) Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training ([train.py](https://github.com/ultralytics/yolov5/blob/master/train.py)), validation ([val.py](https://github.com/ultralytics/yolov5/blob/master/val.py)), inference ([detect.py](https://github.com/ultralytics/yolov5/blob/master/detect.py)) and export ([export.py](https://github.com/ultralytics/yolov5/blob/master/export.py)) on MacOS, Windows, and Ubuntu every 24 hours and on every commit. diff --git a/utils/loggers/wandb/__init__.py b/utils/loggers/wandb/__init__.py new file mode 100644 index 0000000..e69de29 diff --git a/utils/loggers/wandb/log_dataset.py b/utils/loggers/wandb/log_dataset.py new file mode 100644 index 0000000..06e81fb --- /dev/null +++ b/utils/loggers/wandb/log_dataset.py @@ -0,0 +1,27 @@ +import argparse + +from wandb_utils import WandbLogger + +from utils.general import LOGGER + +WANDB_ARTIFACT_PREFIX = 'wandb-artifact://' + + +def create_dataset_artifact(opt): + logger = WandbLogger(opt, None, job_type='Dataset Creation') # TODO: return value unused + if not logger.wandb: + LOGGER.info("install wandb using `pip install wandb` to log the dataset") + + +if __name__ == '__main__': + parser = argparse.ArgumentParser() + parser.add_argument('--data', type=str, default='data/coco128.yaml', help='data.yaml path') + parser.add_argument('--single-cls', action='store_true', help='train as single-class dataset') + parser.add_argument('--project', type=str, default='YOLOv5', help='name of W&B Project') + parser.add_argument('--entity', default=None, help='W&B entity') + parser.add_argument('--name', type=str, default='log dataset', help='name of W&B run') + + opt = parser.parse_args() + opt.resume = False # Explicitly disallow resume check for dataset upload job + + create_dataset_artifact(opt) diff --git a/utils/loggers/wandb/sweep.py b/utils/loggers/wandb/sweep.py new file mode 100644 index 0000000..206059b --- /dev/null +++ b/utils/loggers/wandb/sweep.py @@ -0,0 +1,41 @@ +import sys +from pathlib import Path + +import wandb + +FILE = Path(__file__).resolve() +ROOT = FILE.parents[3] # YOLOv5 root directory +if str(ROOT) not in sys.path: + sys.path.append(str(ROOT)) # add ROOT to PATH + +from train import parse_opt, train +from utils.callbacks import Callbacks +from utils.general import increment_path +from utils.torch_utils import select_device + + +def sweep(): + wandb.init() + # Get hyp dict from sweep agent + hyp_dict = vars(wandb.config).get("_items") + + # Workaround: get necessary opt args + opt = parse_opt(known=True) + opt.batch_size = hyp_dict.get("batch_size") + opt.save_dir = str(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok or opt.evolve)) + opt.epochs = hyp_dict.get("epochs") + opt.nosave = True + opt.data = hyp_dict.get("data") + opt.weights = str(opt.weights) + opt.cfg = str(opt.cfg) + opt.data = str(opt.data) + opt.hyp = str(opt.hyp) + opt.project = str(opt.project) + device = select_device(opt.device, batch_size=opt.batch_size) + + # train + train(hyp_dict, opt, device, callbacks=Callbacks()) + + +if __name__ == "__main__": + sweep() diff --git a/utils/loggers/wandb/sweep.yaml b/utils/loggers/wandb/sweep.yaml new file mode 100644 index 0000000..c7790d7 --- /dev/null +++ b/utils/loggers/wandb/sweep.yaml @@ -0,0 +1,143 @@ +# Hyperparameters for training +# To set range- +# Provide min and max values as: +# parameter: +# +# min: scalar +# max: scalar +# OR +# +# Set a specific list of search space- +# parameter: +# values: [scalar1, scalar2, scalar3...] +# +# You can use grid, bayesian and hyperopt search strategy +# For more info on configuring sweeps visit - https://docs.wandb.ai/guides/sweeps/configuration + +program: utils/loggers/wandb/sweep.py +method: random +metric: + name: metrics/mAP_0.5 + goal: maximize + +parameters: + # hyperparameters: set either min, max range or values list + data: + value: "data/coco128.yaml" + batch_size: + values: [64] + epochs: + values: [10] + + lr0: + distribution: uniform + min: 1e-5 + max: 1e-1 + lrf: + distribution: uniform + min: 0.01 + max: 1.0 + momentum: + distribution: uniform + min: 0.6 + max: 0.98 + weight_decay: + distribution: uniform + min: 0.0 + max: 0.001 + warmup_epochs: + distribution: uniform + min: 0.0 + max: 5.0 + warmup_momentum: + distribution: uniform + min: 0.0 + max: 0.95 + warmup_bias_lr: + distribution: uniform + min: 0.0 + max: 0.2 + box: + distribution: uniform + min: 0.02 + max: 0.2 + cls: + distribution: uniform + min: 0.2 + max: 4.0 + cls_pw: + distribution: uniform + min: 0.5 + max: 2.0 + obj: + distribution: uniform + min: 0.2 + max: 4.0 + obj_pw: + distribution: uniform + min: 0.5 + max: 2.0 + iou_t: + distribution: uniform + min: 0.1 + max: 0.7 + anchor_t: + distribution: uniform + min: 2.0 + max: 8.0 + fl_gamma: + distribution: uniform + min: 0.0 + max: 0.1 + hsv_h: + distribution: uniform + min: 0.0 + max: 0.1 + hsv_s: + distribution: uniform + min: 0.0 + max: 0.9 + hsv_v: + distribution: uniform + min: 0.0 + max: 0.9 + degrees: + distribution: uniform + min: 0.0 + max: 45.0 + translate: + distribution: uniform + min: 0.0 + max: 0.9 + scale: + distribution: uniform + min: 0.0 + max: 0.9 + shear: + distribution: uniform + min: 0.0 + max: 10.0 + perspective: + distribution: uniform + min: 0.0 + max: 0.001 + flipud: + distribution: uniform + min: 0.0 + max: 1.0 + fliplr: + distribution: uniform + min: 0.0 + max: 1.0 + mosaic: + distribution: uniform + min: 0.0 + max: 1.0 + mixup: + distribution: uniform + min: 0.0 + max: 1.0 + copy_paste: + distribution: uniform + min: 0.0 + max: 1.0 diff --git a/utils/loggers/wandb/wandb_utils.py b/utils/loggers/wandb/wandb_utils.py new file mode 100644 index 0000000..3835436 --- /dev/null +++ b/utils/loggers/wandb/wandb_utils.py @@ -0,0 +1,562 @@ +"""Utilities and tools for tracking runs with Weights & Biases.""" + +import logging +import os +import sys +from contextlib import contextmanager +from pathlib import Path +from typing import Dict + +import yaml +from tqdm import tqdm + +FILE = Path(__file__).resolve() +ROOT = FILE.parents[3] # YOLOv5 root directory +if str(ROOT) not in sys.path: + sys.path.append(str(ROOT)) # add ROOT to PATH + +from utils.datasets import LoadImagesAndLabels, img2label_paths +from utils.general import LOGGER, check_dataset, check_file + +try: + import wandb + + assert hasattr(wandb, '__version__') # verify package import not local dir +except (ImportError, AssertionError): + wandb = None + +RANK = int(os.getenv('RANK', -1)) +WANDB_ARTIFACT_PREFIX = 'wandb-artifact://' + + +def remove_prefix(from_string, prefix=WANDB_ARTIFACT_PREFIX): + return from_string[len(prefix):] + + +def check_wandb_config_file(data_config_file): + wandb_config = '_wandb.'.join(data_config_file.rsplit('.', 1)) # updated data.yaml path + if Path(wandb_config).is_file(): + return wandb_config + return data_config_file + + +def check_wandb_dataset(data_file): + is_trainset_wandb_artifact = False + is_valset_wandb_artifact = False + if check_file(data_file) and data_file.endswith('.yaml'): + with open(data_file, errors='ignore') as f: + data_dict = yaml.safe_load(f) + is_trainset_wandb_artifact = (isinstance(data_dict['train'], str) and + data_dict['train'].startswith(WANDB_ARTIFACT_PREFIX)) + is_valset_wandb_artifact = (isinstance(data_dict['val'], str) and + data_dict['val'].startswith(WANDB_ARTIFACT_PREFIX)) + if is_trainset_wandb_artifact or is_valset_wandb_artifact: + return data_dict + else: + return check_dataset(data_file) + + +def get_run_info(run_path): + run_path = Path(remove_prefix(run_path, WANDB_ARTIFACT_PREFIX)) + run_id = run_path.stem + project = run_path.parent.stem + entity = run_path.parent.parent.stem + model_artifact_name = 'run_' + run_id + '_model' + return entity, project, run_id, model_artifact_name + + +def check_wandb_resume(opt): + process_wandb_config_ddp_mode(opt) if RANK not in [-1, 0] else None + if isinstance(opt.resume, str): + if opt.resume.startswith(WANDB_ARTIFACT_PREFIX): + if RANK not in [-1, 0]: # For resuming DDP runs + entity, project, run_id, model_artifact_name = get_run_info(opt.resume) + api = wandb.Api() + artifact = api.artifact(entity + '/' + project + '/' + model_artifact_name + ':latest') + modeldir = artifact.download() + opt.weights = str(Path(modeldir) / "last.pt") + return True + return None + + +def process_wandb_config_ddp_mode(opt): + with open(check_file(opt.data), errors='ignore') as f: + data_dict = yaml.safe_load(f) # data dict + train_dir, val_dir = None, None + if isinstance(data_dict['train'], str) and data_dict['train'].startswith(WANDB_ARTIFACT_PREFIX): + api = wandb.Api() + train_artifact = api.artifact(remove_prefix(data_dict['train']) + ':' + opt.artifact_alias) + train_dir = train_artifact.download() + train_path = Path(train_dir) / 'data/images/' + data_dict['train'] = str(train_path) + + if isinstance(data_dict['val'], str) and data_dict['val'].startswith(WANDB_ARTIFACT_PREFIX): + api = wandb.Api() + val_artifact = api.artifact(remove_prefix(data_dict['val']) + ':' + opt.artifact_alias) + val_dir = val_artifact.download() + val_path = Path(val_dir) / 'data/images/' + data_dict['val'] = str(val_path) + if train_dir or val_dir: + ddp_data_path = str(Path(val_dir) / 'wandb_local_data.yaml') + with open(ddp_data_path, 'w') as f: + yaml.safe_dump(data_dict, f) + opt.data = ddp_data_path + + +class WandbLogger(): + """Log training runs, datasets, models, and predictions to Weights & Biases. + + This logger sends information to W&B at wandb.ai. By default, this information + includes hyperparameters, system configuration and metrics, model metrics, + and basic data metrics and analyses. + + By providing additional command line arguments to train.py, datasets, + models and predictions can also be logged. + + For more on how this logger is used, see the Weights & Biases documentation: + https://docs.wandb.com/guides/integrations/yolov5 + """ + + def __init__(self, opt, run_id=None, job_type='Training'): + """ + - Initialize WandbLogger instance + - Upload dataset if opt.upload_dataset is True + - Setup trainig processes if job_type is 'Training' + + arguments: + opt (namespace) -- Commandline arguments for this run + run_id (str) -- Run ID of W&B run to be resumed + job_type (str) -- To set the job_type for this run + + """ + # Pre-training routine -- + self.job_type = job_type + self.wandb, self.wandb_run = wandb, None if not wandb else wandb.run + self.val_artifact, self.train_artifact = None, None + self.train_artifact_path, self.val_artifact_path = None, None + self.result_artifact = None + self.val_table, self.result_table = None, None + self.bbox_media_panel_images = [] + self.val_table_path_map = None + self.max_imgs_to_log = 16 + self.wandb_artifact_data_dict = None + self.data_dict = None + # It's more elegant to stick to 1 wandb.init call, + # but useful config data is overwritten in the WandbLogger's wandb.init call + if isinstance(opt.resume, str): # checks resume from artifact + if opt.resume.startswith(WANDB_ARTIFACT_PREFIX): + entity, project, run_id, model_artifact_name = get_run_info(opt.resume) + model_artifact_name = WANDB_ARTIFACT_PREFIX + model_artifact_name + assert wandb, 'install wandb to resume wandb runs' + # Resume wandb-artifact:// runs here| workaround for not overwriting wandb.config + self.wandb_run = wandb.init(id=run_id, + project=project, + entity=entity, + resume='allow', + allow_val_change=True) + opt.resume = model_artifact_name + elif self.wandb: + self.wandb_run = wandb.init(config=opt, + resume="allow", + project='YOLOv5' if opt.project == 'runs/train' else Path(opt.project).stem, + entity=opt.entity, + name=opt.name if opt.name != 'exp' else None, + job_type=job_type, + id=run_id, + allow_val_change=True) if not wandb.run else wandb.run + if self.wandb_run: + if self.job_type == 'Training': + if opt.upload_dataset: + if not opt.resume: + self.wandb_artifact_data_dict = self.check_and_upload_dataset(opt) + + if opt.resume: + # resume from artifact + if isinstance(opt.resume, str) and opt.resume.startswith(WANDB_ARTIFACT_PREFIX): + self.data_dict = dict(self.wandb_run.config.data_dict) + else: # local resume + self.data_dict = check_wandb_dataset(opt.data) + else: + self.data_dict = check_wandb_dataset(opt.data) + self.wandb_artifact_data_dict = self.wandb_artifact_data_dict or self.data_dict + + # write data_dict to config. useful for resuming from artifacts. Do this only when not resuming. + self.wandb_run.config.update({'data_dict': self.wandb_artifact_data_dict}, + allow_val_change=True) + self.setup_training(opt) + + if self.job_type == 'Dataset Creation': + self.wandb_run.config.update({"upload_dataset": True}) + self.data_dict = self.check_and_upload_dataset(opt) + + def check_and_upload_dataset(self, opt): + """ + Check if the dataset format is compatible and upload it as W&B artifact + + arguments: + opt (namespace)-- Commandline arguments for current run + + returns: + Updated dataset info dictionary where local dataset paths are replaced by WAND_ARFACT_PREFIX links. + """ + assert wandb, 'Install wandb to upload dataset' + config_path = self.log_dataset_artifact(opt.data, + opt.single_cls, + 'YOLOv5' if opt.project == 'runs/train' else Path(opt.project).stem) + with open(config_path, errors='ignore') as f: + wandb_data_dict = yaml.safe_load(f) + return wandb_data_dict + + def setup_training(self, opt): + """ + Setup the necessary processes for training YOLO models: + - Attempt to download model checkpoint and dataset artifacts if opt.resume stats with WANDB_ARTIFACT_PREFIX + - Update data_dict, to contain info of previous run if resumed and the paths of dataset artifact if downloaded + - Setup log_dict, initialize bbox_interval + + arguments: + opt (namespace) -- commandline arguments for this run + + """ + self.log_dict, self.current_epoch = {}, 0 + self.bbox_interval = opt.bbox_interval + if isinstance(opt.resume, str): + modeldir, _ = self.download_model_artifact(opt) + if modeldir: + self.weights = Path(modeldir) / "last.pt" + config = self.wandb_run.config + opt.weights, opt.save_period, opt.batch_size, opt.bbox_interval, opt.epochs, opt.hyp, opt.imgsz = str( + self.weights), config.save_period, config.batch_size, config.bbox_interval, config.epochs,\ + config.hyp, config.imgsz + data_dict = self.data_dict + if self.val_artifact is None: # If --upload_dataset is set, use the existing artifact, don't download + self.train_artifact_path, self.train_artifact = self.download_dataset_artifact(data_dict.get('train'), + opt.artifact_alias) + self.val_artifact_path, self.val_artifact = self.download_dataset_artifact(data_dict.get('val'), + opt.artifact_alias) + + if self.train_artifact_path is not None: + train_path = Path(self.train_artifact_path) / 'data/images/' + data_dict['train'] = str(train_path) + if self.val_artifact_path is not None: + val_path = Path(self.val_artifact_path) / 'data/images/' + data_dict['val'] = str(val_path) + + if self.val_artifact is not None: + self.result_artifact = wandb.Artifact("run_" + wandb.run.id + "_progress", "evaluation") + columns = ["epoch", "id", "ground truth", "prediction"] + columns.extend(self.data_dict['names']) + self.result_table = wandb.Table(columns) + self.val_table = self.val_artifact.get("val") + if self.val_table_path_map is None: + self.map_val_table_path() + if opt.bbox_interval == -1: + self.bbox_interval = opt.bbox_interval = (opt.epochs // 10) if opt.epochs > 10 else 1 + if opt.evolve: + self.bbox_interval = opt.bbox_interval = opt.epochs + 1 + train_from_artifact = self.train_artifact_path is not None and self.val_artifact_path is not None + # Update the the data_dict to point to local artifacts dir + if train_from_artifact: + self.data_dict = data_dict + + def download_dataset_artifact(self, path, alias): + """ + download the model checkpoint artifact if the path starts with WANDB_ARTIFACT_PREFIX + + arguments: + path -- path of the dataset to be used for training + alias (str)-- alias of the artifact to be download/used for training + + returns: + (str, wandb.Artifact) -- path of the downladed dataset and it's corresponding artifact object if dataset + is found otherwise returns (None, None) + """ + if isinstance(path, str) and path.startswith(WANDB_ARTIFACT_PREFIX): + artifact_path = Path(remove_prefix(path, WANDB_ARTIFACT_PREFIX) + ":" + alias) + dataset_artifact = wandb.use_artifact(artifact_path.as_posix().replace("\\", "/")) + assert dataset_artifact is not None, "'Error: W&B dataset artifact doesn\'t exist'" + datadir = dataset_artifact.download() + return datadir, dataset_artifact + return None, None + + def download_model_artifact(self, opt): + """ + download the model checkpoint artifact if the resume path starts with WANDB_ARTIFACT_PREFIX + + arguments: + opt (namespace) -- Commandline arguments for this run + """ + if opt.resume.startswith(WANDB_ARTIFACT_PREFIX): + model_artifact = wandb.use_artifact(remove_prefix(opt.resume, WANDB_ARTIFACT_PREFIX) + ":latest") + assert model_artifact is not None, 'Error: W&B model artifact doesn\'t exist' + modeldir = model_artifact.download() + # epochs_trained = model_artifact.metadata.get('epochs_trained') + total_epochs = model_artifact.metadata.get('total_epochs') + is_finished = total_epochs is None + assert not is_finished, 'training is finished, can only resume incomplete runs.' + return modeldir, model_artifact + return None, None + + def log_model(self, path, opt, epoch, fitness_score, best_model=False): + """ + Log the model checkpoint as W&B artifact + + arguments: + path (Path) -- Path of directory containing the checkpoints + opt (namespace) -- Command line arguments for this run + epoch (int) -- Current epoch number + fitness_score (float) -- fitness score for current epoch + best_model (boolean) -- Boolean representing if the current checkpoint is the best yet. + """ + model_artifact = wandb.Artifact('run_' + wandb.run.id + '_model', type='model', metadata={ + 'original_url': str(path), + 'epochs_trained': epoch + 1, + 'save period': opt.save_period, + 'project': opt.project, + 'total_epochs': opt.epochs, + 'fitness_score': fitness_score + }) + model_artifact.add_file(str(path / 'last.pt'), name='last.pt') + wandb.log_artifact(model_artifact, + aliases=['latest', 'last', 'epoch ' + str(self.current_epoch), 'best' if best_model else '']) + LOGGER.info(f"Saving model artifact on epoch {epoch + 1}") + + def log_dataset_artifact(self, data_file, single_cls, project, overwrite_config=False): + """ + Log the dataset as W&B artifact and return the new data file with W&B links + + arguments: + data_file (str) -- the .yaml file with information about the dataset like - path, classes etc. + single_class (boolean) -- train multi-class data as single-class + project (str) -- project name. Used to construct the artifact path + overwrite_config (boolean) -- overwrites the data.yaml file if set to true otherwise creates a new + file with _wandb postfix. Eg -> data_wandb.yaml + + returns: + the new .yaml file with artifact links. it can be used to start training directly from artifacts + """ + upload_dataset = self.wandb_run.config.upload_dataset + log_val_only = isinstance(upload_dataset, str) and upload_dataset == 'val' + self.data_dict = check_dataset(data_file) # parse and check + data = dict(self.data_dict) + nc, names = (1, ['item']) if single_cls else (int(data['nc']), data['names']) + names = {k: v for k, v in enumerate(names)} # to index dictionary + + # log train set + if not log_val_only: + self.train_artifact = self.create_dataset_table(LoadImagesAndLabels( + data['train'], rect=True, batch_size=1), names, name='train') if data.get('train') else None + if data.get('train'): + data['train'] = WANDB_ARTIFACT_PREFIX + str(Path(project) / 'train') + + self.val_artifact = self.create_dataset_table(LoadImagesAndLabels( + data['val'], rect=True, batch_size=1), names, name='val') if data.get('val') else None + if data.get('val'): + data['val'] = WANDB_ARTIFACT_PREFIX + str(Path(project) / 'val') + + path = Path(data_file) + # create a _wandb.yaml file with artifacts links if both train and test set are logged + if not log_val_only: + path = (path.stem if overwrite_config else path.stem + '_wandb') + '.yaml' # updated data.yaml path + path = ROOT / 'data' / path + data.pop('download', None) + data.pop('path', None) + with open(path, 'w') as f: + yaml.safe_dump(data, f) + LOGGER.info(f"Created dataset config file {path}") + + if self.job_type == 'Training': # builds correct artifact pipeline graph + if not log_val_only: + self.wandb_run.log_artifact( + self.train_artifact) # calling use_artifact downloads the dataset. NOT NEEDED! + self.wandb_run.use_artifact(self.val_artifact) + self.val_artifact.wait() + self.val_table = self.val_artifact.get('val') + self.map_val_table_path() + else: + self.wandb_run.log_artifact(self.train_artifact) + self.wandb_run.log_artifact(self.val_artifact) + return path + + def map_val_table_path(self): + """ + Map the validation dataset Table like name of file -> it's id in the W&B Table. + Useful for - referencing artifacts for evaluation. + """ + self.val_table_path_map = {} + LOGGER.info("Mapping dataset") + for i, data in enumerate(tqdm(self.val_table.data)): + self.val_table_path_map[data[3]] = data[0] + + def create_dataset_table(self, dataset: LoadImagesAndLabels, class_to_id: Dict[int, str], name: str = 'dataset'): + """ + Create and return W&B artifact containing W&B Table of the dataset. + + arguments: + dataset -- instance of LoadImagesAndLabels class used to iterate over the data to build Table + class_to_id -- hash map that maps class ids to labels + name -- name of the artifact + + returns: + dataset artifact to be logged or used + """ + # TODO: Explore multiprocessing to slpit this loop parallely| This is essential for speeding up the the logging + artifact = wandb.Artifact(name=name, type="dataset") + img_files = tqdm([dataset.path]) if isinstance(dataset.path, str) and Path(dataset.path).is_dir() else None + img_files = tqdm(dataset.img_files) if not img_files else img_files + for img_file in img_files: + if Path(img_file).is_dir(): + artifact.add_dir(img_file, name='data/images') + labels_path = 'labels'.join(dataset.path.rsplit('images', 1)) + artifact.add_dir(labels_path, name='data/labels') + else: + artifact.add_file(img_file, name='data/images/' + Path(img_file).name) + label_file = Path(img2label_paths([img_file])[0]) + artifact.add_file(str(label_file), + name='data/labels/' + label_file.name) if label_file.exists() else None + table = wandb.Table(columns=["id", "train_image", "Classes", "name"]) + class_set = wandb.Classes([{'id': id, 'name': name} for id, name in class_to_id.items()]) + for si, (img, labels, paths, shapes) in enumerate(tqdm(dataset)): + box_data, img_classes = [], {} + for cls, *xywh in labels[:, 1:].tolist(): + cls = int(cls) + box_data.append({"position": {"middle": [xywh[0], xywh[1]], "width": xywh[2], "height": xywh[3]}, + "class_id": cls, + "box_caption": "%s" % (class_to_id[cls])}) + img_classes[cls] = class_to_id[cls] + boxes = {"ground_truth": {"box_data": box_data, "class_labels": class_to_id}} # inference-space + table.add_data(si, wandb.Image(paths, classes=class_set, boxes=boxes), list(img_classes.values()), + Path(paths).name) + artifact.add(table, name) + return artifact + + def log_training_progress(self, predn, path, names): + """ + Build evaluation Table. Uses reference from validation dataset table. + + arguments: + predn (list): list of predictions in the native space in the format - [xmin, ymin, xmax, ymax, confidence, class] + path (str): local path of the current evaluation image + names (dict(int, str)): hash map that maps class ids to labels + """ + class_set = wandb.Classes([{'id': id, 'name': name} for id, name in names.items()]) + box_data = [] + avg_conf_per_class = [0] * len(self.data_dict['names']) + pred_class_count = {} + for *xyxy, conf, cls in predn.tolist(): + if conf >= 0.25: + cls = int(cls) + box_data.append( + {"position": {"minX": xyxy[0], "minY": xyxy[1], "maxX": xyxy[2], "maxY": xyxy[3]}, + "class_id": cls, + "box_caption": f"{names[cls]} {conf:.3f}", + "scores": {"class_score": conf}, + "domain": "pixel"}) + avg_conf_per_class[cls] += conf + + if cls in pred_class_count: + pred_class_count[cls] += 1 + else: + pred_class_count[cls] = 1 + + for pred_class in pred_class_count.keys(): + avg_conf_per_class[pred_class] = avg_conf_per_class[pred_class] / pred_class_count[pred_class] + + boxes = {"predictions": {"box_data": box_data, "class_labels": names}} # inference-space + id = self.val_table_path_map[Path(path).name] + self.result_table.add_data(self.current_epoch, + id, + self.val_table.data[id][1], + wandb.Image(self.val_table.data[id][1], boxes=boxes, classes=class_set), + *avg_conf_per_class + ) + + def val_one_image(self, pred, predn, path, names, im): + """ + Log validation data for one image. updates the result Table if validation dataset is uploaded and log bbox media panel + + arguments: + pred (list): list of scaled predictions in the format - [xmin, ymin, xmax, ymax, confidence, class] + predn (list): list of predictions in the native space - [xmin, ymin, xmax, ymax, confidence, class] + path (str): local path of the current evaluation image + """ + if self.val_table and self.result_table: # Log Table if Val dataset is uploaded as artifact + self.log_training_progress(predn, path, names) + + if len(self.bbox_media_panel_images) < self.max_imgs_to_log and self.current_epoch > 0: + if self.current_epoch % self.bbox_interval == 0: + box_data = [{"position": {"minX": xyxy[0], "minY": xyxy[1], "maxX": xyxy[2], "maxY": xyxy[3]}, + "class_id": int(cls), + "box_caption": f"{names[int(cls)]} {conf:.3f}", + "scores": {"class_score": conf}, + "domain": "pixel"} for *xyxy, conf, cls in pred.tolist()] + boxes = {"predictions": {"box_data": box_data, "class_labels": names}} # inference-space + self.bbox_media_panel_images.append(wandb.Image(im, boxes=boxes, caption=path.name)) + + def log(self, log_dict): + """ + save the metrics to the logging dictionary + + arguments: + log_dict (Dict) -- metrics/media to be logged in current step + """ + if self.wandb_run: + for key, value in log_dict.items(): + self.log_dict[key] = value + + def end_epoch(self, best_result=False): + """ + commit the log_dict, model artifacts and Tables to W&B and flush the log_dict. + + arguments: + best_result (boolean): Boolean representing if the result of this evaluation is best or not + """ + if self.wandb_run: + with all_logging_disabled(): + if self.bbox_media_panel_images: + self.log_dict["BoundingBoxDebugger"] = self.bbox_media_panel_images + try: + wandb.log(self.log_dict) + except BaseException as e: + LOGGER.info( + f"An error occurred in wandb logger. The training will proceed without interruption. More info\n{e}") + self.wandb_run.finish() + self.wandb_run = None + + self.log_dict = {} + self.bbox_media_panel_images = [] + if self.result_artifact: + self.result_artifact.add(self.result_table, 'result') + wandb.log_artifact(self.result_artifact, aliases=['latest', 'last', 'epoch ' + str(self.current_epoch), + ('best' if best_result else '')]) + + wandb.log({"evaluation": self.result_table}) + columns = ["epoch", "id", "ground truth", "prediction"] + columns.extend(self.data_dict['names']) + self.result_table = wandb.Table(columns) + self.result_artifact = wandb.Artifact("run_" + wandb.run.id + "_progress", "evaluation") + + def finish_run(self): + """ + Log metrics if any and finish the current W&B run + """ + if self.wandb_run: + if self.log_dict: + with all_logging_disabled(): + wandb.log(self.log_dict) + wandb.run.finish() + + +@contextmanager +def all_logging_disabled(highest_level=logging.CRITICAL): + """ source - https://gist.github.com/simon-weber/7853144 + A context manager that will prevent any logging messages triggered during the body from being processed. + :param highest_level: the maximum logging level in use. + This would only need to be changed if a custom level greater than CRITICAL is defined. + """ + previous_level = logging.root.manager.disable + logging.disable(highest_level) + try: + yield + finally: + logging.disable(previous_level) diff --git a/utils/loss.py b/utils/loss.py new file mode 100644 index 0000000..d29084c --- /dev/null +++ b/utils/loss.py @@ -0,0 +1,224 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Loss functions +""" + +import torch +import torch.nn as nn + +from utils.metrics import bbox_iou +from utils.torch_utils import de_parallel + + +def smooth_BCE(eps=0.1): # https://github.com/ultralytics/yolov3/issues/238#issuecomment-598028441 + # return positive, negative label smoothing BCE targets + return 1.0 - 0.5 * eps, 0.5 * eps + + +class BCEBlurWithLogitsLoss(nn.Module): + # BCEwithLogitLoss() with reduced missing label effects. + def __init__(self, alpha=0.05): + super().__init__() + self.loss_fcn = nn.BCEWithLogitsLoss(reduction='none') # must be nn.BCEWithLogitsLoss() + self.alpha = alpha + + def forward(self, pred, true): + loss = self.loss_fcn(pred, true) + pred = torch.sigmoid(pred) # prob from logits + dx = pred - true # reduce only missing label effects + # dx = (pred - true).abs() # reduce missing label and false label effects + alpha_factor = 1 - torch.exp((dx - 1) / (self.alpha + 1e-4)) + loss *= alpha_factor + return loss.mean() + + +class FocalLoss(nn.Module): + # Wraps focal loss around existing loss_fcn(), i.e. criteria = FocalLoss(nn.BCEWithLogitsLoss(), gamma=1.5) + def __init__(self, loss_fcn, gamma=1.5, alpha=0.25): + super().__init__() + self.loss_fcn = loss_fcn # must be nn.BCEWithLogitsLoss() + self.gamma = gamma + self.alpha = alpha + self.reduction = loss_fcn.reduction + self.loss_fcn.reduction = 'none' # required to apply FL to each element + + def forward(self, pred, true): + loss = self.loss_fcn(pred, true) + # p_t = torch.exp(-loss) + # loss *= self.alpha * (1.000001 - p_t) ** self.gamma # non-zero power for gradient stability + + # TF implementation https://github.com/tensorflow/addons/blob/v0.7.1/tensorflow_addons/losses/focal_loss.py + pred_prob = torch.sigmoid(pred) # prob from logits + p_t = true * pred_prob + (1 - true) * (1 - pred_prob) + alpha_factor = true * self.alpha + (1 - true) * (1 - self.alpha) + modulating_factor = (1.0 - p_t) ** self.gamma + loss *= alpha_factor * modulating_factor + + if self.reduction == 'mean': + return loss.mean() + elif self.reduction == 'sum': + return loss.sum() + else: # 'none' + return loss + + +class QFocalLoss(nn.Module): + # Wraps Quality focal loss around existing loss_fcn(), i.e. criteria = FocalLoss(nn.BCEWithLogitsLoss(), gamma=1.5) + def __init__(self, loss_fcn, gamma=1.5, alpha=0.25): + super().__init__() + self.loss_fcn = loss_fcn # must be nn.BCEWithLogitsLoss() + self.gamma = gamma + self.alpha = alpha + self.reduction = loss_fcn.reduction + self.loss_fcn.reduction = 'none' # required to apply FL to each element + + def forward(self, pred, true): + loss = self.loss_fcn(pred, true) + + pred_prob = torch.sigmoid(pred) # prob from logits + alpha_factor = true * self.alpha + (1 - true) * (1 - self.alpha) + modulating_factor = torch.abs(true - pred_prob) ** self.gamma + loss *= alpha_factor * modulating_factor + + if self.reduction == 'mean': + return loss.mean() + elif self.reduction == 'sum': + return loss.sum() + else: # 'none' + return loss + + +class ComputeLoss: + # Compute losses + def __init__(self, model, autobalance=False): + self.sort_obj_iou = False + device = next(model.parameters()).device # get model device + h = model.hyp # hyperparameters + + # Define criteria + BCEcls = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['cls_pw']], device=device)) + BCEobj = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['obj_pw']], device=device)) + + # Class label smoothing https://arxiv.org/pdf/1902.04103.pdf eqn 3 + self.cp, self.cn = smooth_BCE(eps=h.get('label_smoothing', 0.0)) # positive, negative BCE targets + + # Focal loss + g = h['fl_gamma'] # focal loss gamma + if g > 0: + BCEcls, BCEobj = FocalLoss(BCEcls, g), FocalLoss(BCEobj, g) + + det = de_parallel(model).model[-1] # Detect() module + self.balance = {3: [4.0, 1.0, 0.4]}.get(det.nl, [4.0, 1.0, 0.25, 0.06, 0.02]) # P3-P7 balanceį”¨æĨ莞įŊŽä¸‰ä¸Ēfeature map寚åē”čž“å‡ēįš„įŊŽäŋĄåēĻæŸå¤ąįŗģ数 + self.ssi = list(det.stride).index(16) if autobalance else 0 # stride 16 index + self.BCEcls, self.BCEobj, self.gr, self.hyp, self.autobalance = BCEcls, BCEobj, 1.0, h, autobalance + for k in 'na', 'nc', 'nl', 'anchors': + setattr(self, k, getattr(det, k)) + + def __call__(self, p, targets): # predictions, targets, model + device = targets.device + lcls, lbox, lobj = torch.zeros(1, device=device), torch.zeros(1, device=device), torch.zeros(1, device=device) + tcls, tbox, indices, anchors = self.build_targets(p, targets) # targets + + # Losses + for i, pi in enumerate(p): # layer index, layer predictions + b, a, gj, gi = indices[i] # image, anchor, gridy, gridx + tobj = torch.zeros_like(pi[..., 0], device=device) # target obj + + n = b.shape[0] # number of targets + if n: + ps = pi[b, a, gj, gi] # prediction subset corresponding to targets + + # Regression + pxy = ps[:, :2].sigmoid() * 2 - 0.5 + pwh = (ps[:, 2:4].sigmoid() * 2) ** 2 * anchors[i] + pbox = torch.cat((pxy, pwh), 1) # predicted box + iou = bbox_iou(pbox.T, tbox[i], x1y1x2y2=False, CIoU=True) # iou(prediction, target) + lbox += (1.0 - iou).mean() # iou loss + + # Objectness + score_iou = iou.detach().clamp(0).type(tobj.dtype) + if self.sort_obj_iou: + sort_id = torch.argsort(score_iou) + b, a, gj, gi, score_iou = b[sort_id], a[sort_id], gj[sort_id], gi[sort_id], score_iou[sort_id] + tobj[b, a, gj, gi] = (1.0 - self.gr) + self.gr * score_iou # iou ratio + + # Classification + if self.nc > 1: # cls loss (only if multiple classes) + t = torch.full_like(ps[:, 5:], self.cn, device=device) # targets + t[range(n), tcls[i]] = self.cp + lcls += self.BCEcls(ps[:, 5:], t) # BCE + + # Append targets to text file + # with open('targets.txt', 'a') as file: + # [file.write('%11.5g ' * 4 % tuple(x) + '\n') for x in torch.cat((txy[i], twh[i]), 1)] + + obji = self.BCEobj(pi[..., 4], tobj) + lobj += obji * self.balance[i] # obj loss + if self.autobalance: + self.balance[i] = self.balance[i] * 0.9999 + 0.0001 / obji.detach().item() + + if self.autobalance: + self.balance = [x / self.balance[self.ssi] for x in self.balance] + lbox *= self.hyp['box'] + lobj *= self.hyp['obj'] + lcls *= self.hyp['cls'] + bs = tobj.shape[0] # batch size + + return (lbox + lobj + lcls) * bs, torch.cat((lbox, lobj, lcls)).detach() + + def build_targets(self, p, targets): + # Build targets for compute_loss(), input targets(image,class,x,y,w,h) + na, nt = self.na, targets.shape[0] # number of anchors, targets + tcls, tbox, indices, anch = [], [], [], [] + gain = torch.ones(7, device=targets.device) # normalized to gridspace gain + ai = torch.arange(na, device=targets.device).float().view(na, 1).repeat(1, nt) # same as .repeat_interleave(nt) + targets = torch.cat((targets.repeat(na, 1, 1), ai[:, :, None]), 2) # append anchor indices + + g = 0.5 # bias + off = torch.tensor([[0, 0], + [1, 0], [0, 1], [-1, 0], [0, -1], # j,k,l,m + # [1, 1], [1, -1], [-1, 1], [-1, -1], # jk,jm,lk,lm + ], device=targets.device).float() * g # offsets + + for i in range(self.nl): + #anchors = self.anchors[i] + anchors, shape = self.anchors[i], p[i].shape + gain[2:6] = torch.tensor(p[i].shape)[[3, 2, 3, 2]] # xyxy gain + + # Match targets to anchors + t = targets * gain + if nt: + # Matches + r = t[:, :, 4:6] / anchors[:, None] # wh ratio + j = torch.max(r, 1 / r).max(2)[0] < self.hyp['anchor_t'] # compare + # j = wh_iou(anchors, t[:, 4:6]) > model.hyp['iou_t'] # iou(3,n)=wh_iou(anchors(3,2), gwh(n,2)) + t = t[j] # filter + + # Offsets + gxy = t[:, 2:4] # grid xy + gxi = gain[[2, 3]] - gxy # inverse + j, k = ((gxy % 1 < g) & (gxy > 1)).T + l, m = ((gxi % 1 < g) & (gxi > 1)).T + j = torch.stack((torch.ones_like(j), j, k, l, m)) + t = t.repeat((5, 1, 1))[j] + offsets = (torch.zeros_like(gxy)[None] + off[:, None])[j] + else: + t = targets[0] + offsets = 0 + + # Define + b, c = t[:, :2].long().T # image, class + gxy = t[:, 2:4] # grid xy + gwh = t[:, 4:6] # grid wh + gij = (gxy - offsets).long() + gi, gj = gij.T # grid xy indices + + # Append + a = t[:, 6].long() # anchor indices + #indices.append((b, a, gj.clamp_(0, gain[3] - 1), gi.clamp_(0, gain[2] - 1))) # image, anchor, grid indices + indices.append((b, a, gj.clamp_(0, shape[2] - 1), gi.clamp_(0, shape[3] - 1))) # image, anchor, grid + tbox.append(torch.cat((gxy - gij, gwh), 1)) # box + anch.append(anchors[a]) # anchors + tcls.append(c) # class + + return tcls, tbox, indices, anch diff --git a/utils/metrics.py b/utils/metrics.py new file mode 100644 index 0000000..857fa5d --- /dev/null +++ b/utils/metrics.py @@ -0,0 +1,342 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Model validation metrics +""" + +import math +import warnings +from pathlib import Path + +import matplotlib.pyplot as plt +import numpy as np +import torch + + +def fitness(x): + # Model fitness as a weighted combination of metrics + w = [0.0, 0.0, 0.1, 0.9] # weights for [P, R, mAP@0.5, mAP@0.5:0.95] + return (x[:, :4] * w).sum(1) + + +def ap_per_class(tp, conf, pred_cls, target_cls, plot=False, save_dir='.', names=(), eps=1e-16): + """ Compute the average precision, given the recall and precision curves. + Source: https://github.com/rafaelpadilla/Object-Detection-Metrics. + # Arguments + tp: True positives (nparray, nx1 or nx10). + conf: Objectness value from 0-1 (nparray). + pred_cls: Predicted object classes (nparray). + target_cls: True object classes (nparray). + plot: Plot precision-recall curve at mAP@0.5 + save_dir: Plot save directory + # Returns + The average precision as computed in py-faster-rcnn. + """ + + # Sort by objectness + i = np.argsort(-conf) + tp, conf, pred_cls = tp[i], conf[i], pred_cls[i] + + # Find unique classes + unique_classes, nt = np.unique(target_cls, return_counts=True) + nc = unique_classes.shape[0] # number of classes, number of detections + + # Create Precision-Recall curve and compute AP for each class + px, py = np.linspace(0, 1, 1000), [] # for plotting + ap, p, r = np.zeros((nc, tp.shape[1])), np.zeros((nc, 1000)), np.zeros((nc, 1000)) + for ci, c in enumerate(unique_classes): + i = pred_cls == c + n_l = nt[ci] # number of labels + n_p = i.sum() # number of predictions + + if n_p == 0 or n_l == 0: + continue + else: + # Accumulate FPs and TPs + fpc = (1 - tp[i]).cumsum(0) + tpc = tp[i].cumsum(0) + + # Recall + recall = tpc / (n_l + eps) # recall curve + r[ci] = np.interp(-px, -conf[i], recall[:, 0], left=0) # negative x, xp because xp decreases + + # Precision + precision = tpc / (tpc + fpc) # precision curve + p[ci] = np.interp(-px, -conf[i], precision[:, 0], left=1) # p at pr_score + + # AP from recall-precision curve + for j in range(tp.shape[1]): + ap[ci, j], mpre, mrec = compute_ap(recall[:, j], precision[:, j]) + if plot and j == 0: + py.append(np.interp(px, mrec, mpre)) # precision at mAP@0.5 + + # Compute F1 (harmonic mean of precision and recall) + f1 = 2 * p * r / (p + r + eps) + names = [v for k, v in names.items() if k in unique_classes] # list: only classes that have data + names = {i: v for i, v in enumerate(names)} # to dict + if plot: + plot_pr_curve(px, py, ap, Path(save_dir) / 'PR_curve.png', names) + plot_mc_curve(px, f1, Path(save_dir) / 'F1_curve.png', names, ylabel='F1') + plot_mc_curve(px, p, Path(save_dir) / 'P_curve.png', names, ylabel='Precision') + plot_mc_curve(px, r, Path(save_dir) / 'R_curve.png', names, ylabel='Recall') + + i = f1.mean(0).argmax() # max F1 index + p, r, f1 = p[:, i], r[:, i], f1[:, i] + tp = (r * nt).round() # true positives + fp = (tp / (p + eps) - tp).round() # false positives + return tp, fp, p, r, f1, ap, unique_classes.astype('int32') + + +def compute_ap(recall, precision): + """ Compute the average precision, given the recall and precision curves + # Arguments + recall: The recall curve (list) + precision: The precision curve (list) + # Returns + Average precision, precision curve, recall curve + """ + + # Append sentinel values to beginning and end + mrec = np.concatenate(([0.0], recall, [1.0])) + mpre = np.concatenate(([1.0], precision, [0.0])) + + # Compute the precision envelope + mpre = np.flip(np.maximum.accumulate(np.flip(mpre))) + + # Integrate area under curve + method = 'interp' # methods: 'continuous', 'interp' + if method == 'interp': + x = np.linspace(0, 1, 101) # 101-point interp (COCO) + ap = np.trapz(np.interp(x, mrec, mpre), x) # integrate + else: # 'continuous' + i = np.where(mrec[1:] != mrec[:-1])[0] # points where x axis (recall) changes + ap = np.sum((mrec[i + 1] - mrec[i]) * mpre[i + 1]) # area under curve + + return ap, mpre, mrec + + +class ConfusionMatrix: + # Updated version of https://github.com/kaanakan/object_detection_confusion_matrix + def __init__(self, nc, conf=0.25, iou_thres=0.45): + self.matrix = np.zeros((nc + 1, nc + 1)) + self.nc = nc # number of classes + self.conf = conf + self.iou_thres = iou_thres + + def process_batch(self, detections, labels): + """ + Return intersection-over-union (Jaccard index) of boxes. + Both sets of boxes are expected to be in (x1, y1, x2, y2) format. + Arguments: + detections (Array[N, 6]), x1, y1, x2, y2, conf, class + labels (Array[M, 5]), class, x1, y1, x2, y2 + Returns: + None, updates confusion matrix accordingly + """ + detections = detections[detections[:, 4] > self.conf] + gt_classes = labels[:, 0].int() + detection_classes = detections[:, 5].int() + iou = box_iou(labels[:, 1:], detections[:, :4]) + + x = torch.where(iou > self.iou_thres) + if x[0].shape[0]: + matches = torch.cat((torch.stack(x, 1), iou[x[0], x[1]][:, None]), 1).cpu().numpy() + if x[0].shape[0] > 1: + matches = matches[matches[:, 2].argsort()[::-1]] + matches = matches[np.unique(matches[:, 1], return_index=True)[1]] + matches = matches[matches[:, 2].argsort()[::-1]] + matches = matches[np.unique(matches[:, 0], return_index=True)[1]] + else: + matches = np.zeros((0, 3)) + + n = matches.shape[0] > 0 + m0, m1, _ = matches.transpose().astype(np.int16) + for i, gc in enumerate(gt_classes): + j = m0 == i + if n and sum(j) == 1: + self.matrix[detection_classes[m1[j]], gc] += 1 # correct + else: + self.matrix[self.nc, gc] += 1 # background FP + + if n: + for i, dc in enumerate(detection_classes): + if not any(m1 == i): + self.matrix[dc, self.nc] += 1 # background FN + + def matrix(self): + return self.matrix + + def tp_fp(self): + tp = self.matrix.diagonal() # true positives + fp = self.matrix.sum(1) - tp # false positives + # fn = self.matrix.sum(0) - tp # false negatives (missed detections) + return tp[:-1], fp[:-1] # remove background class + + def plot(self, normalize=True, save_dir='', names=()): + try: + import seaborn as sn + + array = self.matrix / ((self.matrix.sum(0).reshape(1, -1) + 1E-9) if normalize else 1) # normalize columns + array[array < 0.005] = np.nan # don't annotate (would appear as 0.00) + + fig = plt.figure(figsize=(12, 9), tight_layout=True) + nc, nn = self.nc, len(names) # number of classes, names + sn.set(font_scale=1.0 if nc < 50 else 0.8) # for label size + labels = (0 < nn < 99) and (nn == nc) # apply names to ticklabels + with warnings.catch_warnings(): + warnings.simplefilter('ignore') # suppress empty matrix RuntimeWarning: All-NaN slice encountered + sn.heatmap(array, annot=nc < 30, annot_kws={"size": 8}, cmap='Blues', fmt='.2f', square=True, vmin=0.0, + xticklabels=names + ['background FP'] if labels else "auto", + yticklabels=names + ['background FN'] if labels else "auto").set_facecolor((1, 1, 1)) + fig.axes[0].set_xlabel('True') + fig.axes[0].set_ylabel('Predicted') + fig.savefig(Path(save_dir) / 'confusion_matrix.png', dpi=250) + plt.close() + except Exception as e: + print(f'WARNING: ConfusionMatrix plot failure: {e}') + + def print(self): + for i in range(self.nc + 1): + print(' '.join(map(str, self.matrix[i]))) + + +def bbox_iou(box1, box2, x1y1x2y2=True, GIoU=False, DIoU=False, CIoU=False, eps=1e-7): + # Returns the IoU of box1 to box2. box1 is 4, box2 is nx4 + box2 = box2.T + + # Get the coordinates of bounding boxes + if x1y1x2y2: # x1, y1, x2, y2 = box1 + b1_x1, b1_y1, b1_x2, b1_y2 = box1[0], box1[1], box1[2], box1[3] + b2_x1, b2_y1, b2_x2, b2_y2 = box2[0], box2[1], box2[2], box2[3] + else: # transform from xywh to xyxy + b1_x1, b1_x2 = box1[0] - box1[2] / 2, box1[0] + box1[2] / 2 + b1_y1, b1_y2 = box1[1] - box1[3] / 2, box1[1] + box1[3] / 2 + b2_x1, b2_x2 = box2[0] - box2[2] / 2, box2[0] + box2[2] / 2 + b2_y1, b2_y2 = box2[1] - box2[3] / 2, box2[1] + box2[3] / 2 + + # Intersection area + inter = (torch.min(b1_x2, b2_x2) - torch.max(b1_x1, b2_x1)).clamp(0) * \ + (torch.min(b1_y2, b2_y2) - torch.max(b1_y1, b2_y1)).clamp(0) + + # Union Area + w1, h1 = b1_x2 - b1_x1, b1_y2 - b1_y1 + eps + w2, h2 = b2_x2 - b2_x1, b2_y2 - b2_y1 + eps + union = w1 * h1 + w2 * h2 - inter + eps + + iou = inter / union + if CIoU or DIoU or GIoU: + cw = torch.max(b1_x2, b2_x2) - torch.min(b1_x1, b2_x1) # convex (smallest enclosing box) width + ch = torch.max(b1_y2, b2_y2) - torch.min(b1_y1, b2_y1) # convex height + if CIoU or DIoU: # Distance or Complete IoU https://arxiv.org/abs/1911.08287v1 + c2 = cw ** 2 + ch ** 2 + eps # convex diagonal squared + rho2 = ((b2_x1 + b2_x2 - b1_x1 - b1_x2) ** 2 + + (b2_y1 + b2_y2 - b1_y1 - b1_y2) ** 2) / 4 # center distance squared + if CIoU: # https://github.com/Zzh-tju/DIoU-SSD-pytorch/blob/master/utils/box/box_utils.py#L47 + v = (4 / math.pi ** 2) * torch.pow(torch.atan(w2 / h2) - torch.atan(w1 / h1), 2) + with torch.no_grad(): + alpha = v / (v - iou + (1 + eps)) + return iou - (rho2 / c2 + v * alpha) # CIoU + return iou - rho2 / c2 # DIoU + c_area = cw * ch + eps # convex area + return iou - (c_area - union) / c_area # GIoU https://arxiv.org/pdf/1902.09630.pdf + return iou # IoU + + +def box_iou(box1, box2): + # https://github.com/pytorch/vision/blob/master/torchvision/ops/boxes.py + """ + Return intersection-over-union (Jaccard index) of boxes. + Both sets of boxes are expected to be in (x1, y1, x2, y2) format. + Arguments: + box1 (Tensor[N, 4]) + box2 (Tensor[M, 4]) + Returns: + iou (Tensor[N, M]): the NxM matrix containing the pairwise + IoU values for every element in boxes1 and boxes2 + """ + + def box_area(box): + # box = 4xn + return (box[2] - box[0]) * (box[3] - box[1]) + + area1 = box_area(box1.T) + area2 = box_area(box2.T) + + # inter(N,M) = (rb(N,M,2) - lt(N,M,2)).clamp(0).prod(2) + inter = (torch.min(box1[:, None, 2:], box2[:, 2:]) - torch.max(box1[:, None, :2], box2[:, :2])).clamp(0).prod(2) + return inter / (area1[:, None] + area2 - inter) # iou = inter / (area1 + area2 - inter) + + +def bbox_ioa(box1, box2, eps=1E-7): + """ Returns the intersection over box2 area given box1, box2. Boxes are x1y1x2y2 + box1: np.array of shape(4) + box2: np.array of shape(nx4) + returns: np.array of shape(n) + """ + + box2 = box2.transpose() + + # Get the coordinates of bounding boxes + b1_x1, b1_y1, b1_x2, b1_y2 = box1[0], box1[1], box1[2], box1[3] + b2_x1, b2_y1, b2_x2, b2_y2 = box2[0], box2[1], box2[2], box2[3] + + # Intersection area + inter_area = (np.minimum(b1_x2, b2_x2) - np.maximum(b1_x1, b2_x1)).clip(0) * \ + (np.minimum(b1_y2, b2_y2) - np.maximum(b1_y1, b2_y1)).clip(0) + + # box2 area + box2_area = (b2_x2 - b2_x1) * (b2_y2 - b2_y1) + eps + + # Intersection over box2 area + return inter_area / box2_area + + +def wh_iou(wh1, wh2): + # Returns the nxm IoU matrix. wh1 is nx2, wh2 is mx2 + wh1 = wh1[:, None] # [N,1,2] + wh2 = wh2[None] # [1,M,2] + inter = torch.min(wh1, wh2).prod(2) # [N,M] + return inter / (wh1.prod(2) + wh2.prod(2) - inter) # iou = inter / (area1 + area2 - inter) + + +# Plots ---------------------------------------------------------------------------------------------------------------- + +def plot_pr_curve(px, py, ap, save_dir='pr_curve.png', names=()): + # Precision-recall curve + fig, ax = plt.subplots(1, 1, figsize=(9, 6), tight_layout=True) + py = np.stack(py, axis=1) + + if 0 < len(names) < 21: # display per-class legend if < 21 classes + for i, y in enumerate(py.T): + ax.plot(px, y, linewidth=1, label=f'{names[i]} {ap[i, 0]:.3f}') # plot(recall, precision) + else: + ax.plot(px, py, linewidth=1, color='grey') # plot(recall, precision) + + ax.plot(px, py.mean(1), linewidth=3, color='blue', label='all classes %.3f mAP@0.5' % ap[:, 0].mean()) + ax.set_xlabel('Recall') + ax.set_ylabel('Precision') + ax.set_xlim(0, 1) + ax.set_ylim(0, 1) + plt.legend(bbox_to_anchor=(1.04, 1), loc="upper left") + fig.savefig(Path(save_dir), dpi=250) + plt.close() + + +def plot_mc_curve(px, py, save_dir='mc_curve.png', names=(), xlabel='Confidence', ylabel='Metric'): + # Metric-confidence curve + fig, ax = plt.subplots(1, 1, figsize=(9, 6), tight_layout=True) + + if 0 < len(names) < 21: # display per-class legend if < 21 classes + for i, y in enumerate(py): + ax.plot(px, y, linewidth=1, label=f'{names[i]}') # plot(confidence, metric) + else: + ax.plot(px, py.T, linewidth=1, color='grey') # plot(confidence, metric) + + y = py.mean(0) + ax.plot(px, y, linewidth=3, color='blue', label=f'all classes {y.max():.2f} at {px[y.argmax()]:.3f}') + ax.set_xlabel(xlabel) + ax.set_ylabel(ylabel) + ax.set_xlim(0, 1) + ax.set_ylim(0, 1) + plt.legend(bbox_to_anchor=(1.04, 1), loc="upper left") + fig.savefig(Path(save_dir), dpi=250) + plt.close() diff --git a/utils/plots.py b/utils/plots.py new file mode 100644 index 0000000..6c3f5bc --- /dev/null +++ b/utils/plots.py @@ -0,0 +1,471 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Plotting utils +""" + +import math +import os +from copy import copy +from pathlib import Path + +import cv2 +import matplotlib +import matplotlib.pyplot as plt +import numpy as np +import pandas as pd +import seaborn as sn +import torch +from PIL import Image, ImageDraw, ImageFont + +from utils.general import (CONFIG_DIR, FONT, LOGGER, Timeout, check_font, check_requirements, clip_coords, + increment_path, is_ascii, is_chinese, try_except, xywh2xyxy, xyxy2xywh) +from utils.metrics import fitness + +# Settings +RANK = int(os.getenv('RANK', -1)) +matplotlib.rc('font', **{'size': 11}) +matplotlib.use('Agg') # for writing to files only + + +class Colors: + # Ultralytics color palette https://ultralytics.com/ + def __init__(self): + # hex = matplotlib.colors.TABLEAU_COLORS.values() + hex = ('FF3838', 'FF9D97', 'FF701F', 'FFB21D', 'CFD231', '48F90A', '92CC17', '3DDB86', '1A9334', '00D4BB', + '2C99A8', '00C2FF', '344593', '6473FF', '0018EC', '8438FF', '520085', 'CB38FF', 'FF95C8', 'FF37C7') + self.palette = [self.hex2rgb('#' + c) for c in hex] + self.n = len(self.palette) + + def __call__(self, i, bgr=False): + c = self.palette[int(i) % self.n] + return (c[2], c[1], c[0]) if bgr else c + + @staticmethod + def hex2rgb(h): # rgb order (PIL) + return tuple(int(h[1 + i:1 + i + 2], 16) for i in (0, 2, 4)) + + +colors = Colors() # create instance for 'from utils.plots import colors' + + +def check_pil_font(font=FONT, size=10): + # Return a PIL TrueType Font, downloading to CONFIG_DIR if necessary + font = Path(font) + font = font if font.exists() else (CONFIG_DIR / font.name) + try: + return ImageFont.truetype(str(font) if font.exists() else font.name, size) + except Exception: # download if missing + check_font(font) + try: + return ImageFont.truetype(str(font), size) + except TypeError: + check_requirements('Pillow>=8.4.0') # known issue https://github.com/ultralytics/yolov5/issues/5374 + + +class Annotator: + if RANK in (-1, 0): + check_pil_font() # download TTF if necessary + + # YOLOv5 Annotator for train/val mosaics and jpgs and detect/hub inference annotations + def __init__(self, im, line_width=None, font_size=None, font='Arial.ttf', pil=False, example='abc'): + assert im.data.contiguous, 'Image not contiguous. Apply np.ascontiguousarray(im) to Annotator() input images.' + self.pil = pil or not is_ascii(example) or is_chinese(example) + if self.pil: # use PIL + self.im = im if isinstance(im, Image.Image) else Image.fromarray(im) + self.draw = ImageDraw.Draw(self.im) + self.font = check_pil_font(font='Arial.Unicode.ttf' if is_chinese(example) else font, + size=font_size or max(round(sum(self.im.size) / 2 * 0.035), 12)) + else: # use cv2 + self.im = im + self.lw = line_width or max(round(sum(im.shape) / 2 * 0.003), 2) # line width + + def box_label(self, box, label='', color=(128, 128, 128), txt_color=(255, 255, 255)): + # Add one xyxy box to image with label + if self.pil or not is_ascii(label): + self.draw.rectangle(box, width=self.lw, outline=color) # box + if label: + w, h = self.font.getsize(label) # text width, height + outside = box[1] - h >= 0 # label fits outside box + self.draw.rectangle((box[0], + box[1] - h if outside else box[1], + box[0] + w + 1, + box[1] + 1 if outside else box[1] + h + 1), fill=color) + # self.draw.text((box[0], box[1]), label, fill=txt_color, font=self.font, anchor='ls') # for PIL>8.0 + self.draw.text((box[0], box[1] - h if outside else box[1]), label, fill=txt_color, font=self.font) + else: # cv2 + p1, p2 = (int(box[0]), int(box[1])), (int(box[2]), int(box[3])) + cv2.rectangle(self.im, p1, p2, color, thickness=self.lw, lineType=cv2.LINE_AA) + if label: + tf = max(self.lw - 1, 1) # font thickness + w, h = cv2.getTextSize(label, 0, fontScale=self.lw / 3, thickness=tf)[0] # text width, height + outside = p1[1] - h - 3 >= 0 # label fits outside box + p2 = p1[0] + w, p1[1] - h - 3 if outside else p1[1] + h + 3 + cv2.rectangle(self.im, p1, p2, color, -1, cv2.LINE_AA) # filled + cv2.putText(self.im, label, (p1[0], p1[1] - 2 if outside else p1[1] + h + 2), 0, self.lw / 3, txt_color, + thickness=tf, lineType=cv2.LINE_AA) + + def rectangle(self, xy, fill=None, outline=None, width=1): + # Add rectangle to image (PIL-only) + self.draw.rectangle(xy, fill, outline, width) + + def text(self, xy, text, txt_color=(255, 255, 255)): + # Add text to image (PIL-only) + w, h = self.font.getsize(text) # text width, height + self.draw.text((xy[0], xy[1] - h + 1), text, fill=txt_color, font=self.font) + + def result(self): + # Return annotated image as array + return np.asarray(self.im) + + +def feature_visualization(x, module_type, stage, n=32, save_dir=Path('runs/detect/exp')): + """ + x: Features to be visualized + module_type: Module type + stage: Module stage within model + n: Maximum number of feature maps to plot + save_dir: Directory to save results + """ + if 'Detect' not in module_type: + batch, channels, height, width = x.shape # batch, channels, height, width + if height > 1 and width > 1: + f = save_dir / f"stage{stage}_{module_type.split('.')[-1]}_features.png" # filename + + blocks = torch.chunk(x[0].cpu(), channels, dim=0) # select batch index 0, block by channels + n = min(n, channels) # number of plots + fig, ax = plt.subplots(math.ceil(n / 8), 8, tight_layout=True) # 8 rows x n/8 cols + ax = ax.ravel() + plt.subplots_adjust(wspace=0.05, hspace=0.05) + for i in range(n): + ax[i].imshow(blocks[i].squeeze()) # cmap='gray' + ax[i].axis('off') + + LOGGER.info(f'Saving {f}... ({n}/{channels})') + plt.savefig(f, dpi=300, bbox_inches='tight') + plt.close() + np.save(str(f.with_suffix('.npy')), x[0].cpu().numpy()) # npy save + + +def hist2d(x, y, n=100): + # 2d histogram used in labels.png and evolve.png + xedges, yedges = np.linspace(x.min(), x.max(), n), np.linspace(y.min(), y.max(), n) + hist, xedges, yedges = np.histogram2d(x, y, (xedges, yedges)) + xidx = np.clip(np.digitize(x, xedges) - 1, 0, hist.shape[0] - 1) + yidx = np.clip(np.digitize(y, yedges) - 1, 0, hist.shape[1] - 1) + return np.log(hist[xidx, yidx]) + + +def butter_lowpass_filtfilt(data, cutoff=1500, fs=50000, order=5): + from scipy.signal import butter, filtfilt + + # https://stackoverflow.com/questions/28536191/how-to-filter-smooth-with-scipy-numpy + def butter_lowpass(cutoff, fs, order): + nyq = 0.5 * fs + normal_cutoff = cutoff / nyq + return butter(order, normal_cutoff, btype='low', analog=False) + + b, a = butter_lowpass(cutoff, fs, order=order) + return filtfilt(b, a, data) # forward-backward filter + + +def output_to_target(output): + # Convert model output to target format [batch_id, class_id, x, y, w, h, conf] + targets = [] + for i, o in enumerate(output): + for *box, conf, cls in o.cpu().numpy(): + targets.append([i, cls, *list(*xyxy2xywh(np.array(box)[None])), conf]) + return np.array(targets) + + +def plot_images(images, targets, paths=None, fname='images.jpg', names=None, max_size=1920, max_subplots=16): + # Plot image grid with labels + if isinstance(images, torch.Tensor): + images = images.cpu().float().numpy() + if isinstance(targets, torch.Tensor): + targets = targets.cpu().numpy() + if np.max(images[0]) <= 1: + images *= 255 # de-normalise (optional) + bs, _, h, w = images.shape # batch size, _, height, width + bs = min(bs, max_subplots) # limit plot images + ns = np.ceil(bs ** 0.5) # number of subplots (square) + + # Build Image + mosaic = np.full((int(ns * h), int(ns * w), 3), 255, dtype=np.uint8) # init + for i, im in enumerate(images): + if i == max_subplots: # if last batch has fewer images than we expect + break + x, y = int(w * (i // ns)), int(h * (i % ns)) # block origin + im = im.transpose(1, 2, 0) + mosaic[y:y + h, x:x + w, :] = im + + # Resize (optional) + scale = max_size / ns / max(h, w) + if scale < 1: + h = math.ceil(scale * h) + w = math.ceil(scale * w) + mosaic = cv2.resize(mosaic, tuple(int(x * ns) for x in (w, h))) + + # Annotate + fs = int((h + w) * ns * 0.01) # font size + annotator = Annotator(mosaic, line_width=round(fs / 10), font_size=fs, pil=True, example=names) + for i in range(i + 1): + x, y = int(w * (i // ns)), int(h * (i % ns)) # block origin + annotator.rectangle([x, y, x + w, y + h], None, (255, 255, 255), width=2) # borders + if paths: + annotator.text((x + 5, y + 5 + h), text=Path(paths[i]).name[:40], txt_color=(220, 220, 220)) # filenames + if len(targets) > 0: + ti = targets[targets[:, 0] == i] # image targets + boxes = xywh2xyxy(ti[:, 2:6]).T + classes = ti[:, 1].astype('int') + labels = ti.shape[1] == 6 # labels if no conf column + conf = None if labels else ti[:, 6] # check for confidence presence (label vs pred) + + if boxes.shape[1]: + if boxes.max() <= 1.01: # if normalized with tolerance 0.01 + boxes[[0, 2]] *= w # scale to pixels + boxes[[1, 3]] *= h + elif scale < 1: # absolute coords need scale if image scales + boxes *= scale + boxes[[0, 2]] += x + boxes[[1, 3]] += y + for j, box in enumerate(boxes.T.tolist()): + cls = classes[j] + color = colors(cls) + cls = names[cls] if names else cls + if labels or conf[j] > 0.25: # 0.25 conf thresh + label = f'{cls}' if labels else f'{cls} {conf[j]:.1f}' + annotator.box_label(box, label, color=color) + annotator.im.save(fname) # save + + +def plot_lr_scheduler(optimizer, scheduler, epochs=300, save_dir=''): + # Plot LR simulating training for full epochs + optimizer, scheduler = copy(optimizer), copy(scheduler) # do not modify originals + y = [] + for _ in range(epochs): + scheduler.step() + y.append(optimizer.param_groups[0]['lr']) + plt.plot(y, '.-', label='LR') + plt.xlabel('epoch') + plt.ylabel('LR') + plt.grid() + plt.xlim(0, epochs) + plt.ylim(0) + plt.savefig(Path(save_dir) / 'LR.png', dpi=200) + plt.close() + + +def plot_val_txt(): # from utils.plots import *; plot_val() + # Plot val.txt histograms + x = np.loadtxt('val.txt', dtype=np.float32) + box = xyxy2xywh(x[:, :4]) + cx, cy = box[:, 0], box[:, 1] + + fig, ax = plt.subplots(1, 1, figsize=(6, 6), tight_layout=True) + ax.hist2d(cx, cy, bins=600, cmax=10, cmin=0) + ax.set_aspect('equal') + plt.savefig('hist2d.png', dpi=300) + + fig, ax = plt.subplots(1, 2, figsize=(12, 6), tight_layout=True) + ax[0].hist(cx, bins=600) + ax[1].hist(cy, bins=600) + plt.savefig('hist1d.png', dpi=200) + + +def plot_targets_txt(): # from utils.plots import *; plot_targets_txt() + # Plot targets.txt histograms + x = np.loadtxt('targets.txt', dtype=np.float32).T + s = ['x targets', 'y targets', 'width targets', 'height targets'] + fig, ax = plt.subplots(2, 2, figsize=(8, 8), tight_layout=True) + ax = ax.ravel() + for i in range(4): + ax[i].hist(x[i], bins=100, label=f'{x[i].mean():.3g} +/- {x[i].std():.3g}') + ax[i].legend() + ax[i].set_title(s[i]) + plt.savefig('targets.jpg', dpi=200) + + +def plot_val_study(file='', dir='', x=None): # from utils.plots import *; plot_val_study() + # Plot file=study.txt generated by val.py (or plot all study*.txt in dir) + save_dir = Path(file).parent if file else Path(dir) + plot2 = False # plot additional results + if plot2: + ax = plt.subplots(2, 4, figsize=(10, 6), tight_layout=True)[1].ravel() + + fig2, ax2 = plt.subplots(1, 1, figsize=(8, 4), tight_layout=True) + # for f in [save_dir / f'study_coco_{x}.txt' for x in ['yolov5n6', 'yolov5s6', 'yolov5m6', 'yolov5l6', 'yolov5x6']]: + for f in sorted(save_dir.glob('study*.txt')): + y = np.loadtxt(f, dtype=np.float32, usecols=[0, 1, 2, 3, 7, 8, 9], ndmin=2).T + x = np.arange(y.shape[1]) if x is None else np.array(x) + if plot2: + s = ['P', 'R', 'mAP@.5', 'mAP@.5:.95', 't_preprocess (ms/img)', 't_inference (ms/img)', 't_NMS (ms/img)'] + for i in range(7): + ax[i].plot(x, y[i], '.-', linewidth=2, markersize=8) + ax[i].set_title(s[i]) + + j = y[3].argmax() + 1 + ax2.plot(y[5, 1:j], y[3, 1:j] * 1E2, '.-', linewidth=2, markersize=8, + label=f.stem.replace('study_coco_', '').replace('yolo', 'YOLO')) + + ax2.plot(1E3 / np.array([209, 140, 97, 58, 35, 18]), [34.6, 40.5, 43.0, 47.5, 49.7, 51.5], + 'k.-', linewidth=2, markersize=8, alpha=.25, label='EfficientDet') + + ax2.grid(alpha=0.2) + ax2.set_yticks(np.arange(20, 60, 5)) + ax2.set_xlim(0, 57) + ax2.set_ylim(25, 55) + ax2.set_xlabel('GPU Speed (ms/img)') + ax2.set_ylabel('COCO AP val') + ax2.legend(loc='lower right') + f = save_dir / 'study.png' + print(f'Saving {f}...') + plt.savefig(f, dpi=300) + + +@try_except # known issue https://github.com/ultralytics/yolov5/issues/5395 +@Timeout(30) # known issue https://github.com/ultralytics/yolov5/issues/5611 +def plot_labels(labels, names=(), save_dir=Path('')): + # plot dataset labels + LOGGER.info(f"Plotting labels to {save_dir / 'labels.jpg'}... ") + c, b = labels[:, 0], labels[:, 1:].transpose() # classes, boxes + nc = int(c.max() + 1) # number of classes + x = pd.DataFrame(b.transpose(), columns=['x', 'y', 'width', 'height']) + + # seaborn correlogram + sn.pairplot(x, corner=True, diag_kind='auto', kind='hist', diag_kws=dict(bins=50), plot_kws=dict(pmax=0.9)) + plt.savefig(save_dir / 'labels_correlogram.jpg', dpi=200) + plt.close() + + # matplotlib labels + matplotlib.use('svg') # faster + ax = plt.subplots(2, 2, figsize=(8, 8), tight_layout=True)[1].ravel() + y = ax[0].hist(c, bins=np.linspace(0, nc, nc + 1) - 0.5, rwidth=0.8) + try: # color histogram bars by class + [y[2].patches[i].set_color([x / 255 for x in colors(i)]) for i in range(nc)] # known issue #3195 + except Exception: + pass + ax[0].set_ylabel('instances') + if 0 < len(names) < 30: + ax[0].set_xticks(range(len(names))) + ax[0].set_xticklabels(names, rotation=90, fontsize=10) + else: + ax[0].set_xlabel('classes') + sn.histplot(x, x='x', y='y', ax=ax[2], bins=50, pmax=0.9) + sn.histplot(x, x='width', y='height', ax=ax[3], bins=50, pmax=0.9) + + # rectangles + labels[:, 1:3] = 0.5 # center + labels[:, 1:] = xywh2xyxy(labels[:, 1:]) * 2000 + img = Image.fromarray(np.ones((2000, 2000, 3), dtype=np.uint8) * 255) + for cls, *box in labels[:1000]: + ImageDraw.Draw(img).rectangle(box, width=1, outline=colors(cls)) # plot + ax[1].imshow(img) + ax[1].axis('off') + + for a in [0, 1, 2, 3]: + for s in ['top', 'right', 'left', 'bottom']: + ax[a].spines[s].set_visible(False) + + plt.savefig(save_dir / 'labels.jpg', dpi=200) + matplotlib.use('Agg') + plt.close() + + +def plot_evolve(evolve_csv='path/to/evolve.csv'): # from utils.plots import *; plot_evolve() + # Plot evolve.csv hyp evolution results + evolve_csv = Path(evolve_csv) + data = pd.read_csv(evolve_csv) + keys = [x.strip() for x in data.columns] + x = data.values + f = fitness(x) + j = np.argmax(f) # max fitness index + plt.figure(figsize=(10, 12), tight_layout=True) + matplotlib.rc('font', **{'size': 8}) + print(f'Best results from row {j} of {evolve_csv}:') + for i, k in enumerate(keys[7:]): + v = x[:, 7 + i] + mu = v[j] # best single result + plt.subplot(6, 5, i + 1) + plt.scatter(v, f, c=hist2d(v, f, 20), cmap='viridis', alpha=.8, edgecolors='none') + plt.plot(mu, f.max(), 'k+', markersize=15) + plt.title(f'{k} = {mu:.3g}', fontdict={'size': 9}) # limit to 40 characters + if i % 5 != 0: + plt.yticks([]) + print(f'{k:>15}: {mu:.3g}') + f = evolve_csv.with_suffix('.png') # filename + plt.savefig(f, dpi=200) + plt.close() + print(f'Saved {f}') + + +def plot_results(file='path/to/results.csv', dir=''): + # Plot training results.csv. Usage: from utils.plots import *; plot_results('path/to/results.csv') + save_dir = Path(file).parent if file else Path(dir) + fig, ax = plt.subplots(2, 5, figsize=(12, 6), tight_layout=True) + ax = ax.ravel() + files = list(save_dir.glob('results*.csv')) + assert len(files), f'No results.csv files found in {save_dir.resolve()}, nothing to plot.' + for fi, f in enumerate(files): + try: + data = pd.read_csv(f) + s = [x.strip() for x in data.columns] + x = data.values[:, 0] + for i, j in enumerate([1, 2, 3, 4, 5, 8, 9, 10, 6, 7]): + y = data.values[:, j] + # y[y == 0] = np.nan # don't show zero values + ax[i].plot(x, y, marker='.', label=f.stem, linewidth=2, markersize=8) + ax[i].set_title(s[j], fontsize=12) + # if j in [8, 9, 10]: # share train and val loss y axes + # ax[i].get_shared_y_axes().join(ax[i], ax[i - 5]) + except Exception as e: + LOGGER.info(f'Warning: Plotting error for {f}: {e}') + ax[1].legend() + fig.savefig(save_dir / 'results.png', dpi=200) + plt.close() + + +def profile_idetection(start=0, stop=0, labels=(), save_dir=''): + # Plot iDetection '*.txt' per-image logs. from utils.plots import *; profile_idetection() + ax = plt.subplots(2, 4, figsize=(12, 6), tight_layout=True)[1].ravel() + s = ['Images', 'Free Storage (GB)', 'RAM Usage (GB)', 'Battery', 'dt_raw (ms)', 'dt_smooth (ms)', 'real-world FPS'] + files = list(Path(save_dir).glob('frames*.txt')) + for fi, f in enumerate(files): + try: + results = np.loadtxt(f, ndmin=2).T[:, 90:-30] # clip first and last rows + n = results.shape[1] # number of rows + x = np.arange(start, min(stop, n) if stop else n) + results = results[:, x] + t = (results[0] - results[0].min()) # set t0=0s + results[0] = x + for i, a in enumerate(ax): + if i < len(results): + label = labels[fi] if len(labels) else f.stem.replace('frames_', '') + a.plot(t, results[i], marker='.', label=label, linewidth=1, markersize=5) + a.set_title(s[i]) + a.set_xlabel('time (s)') + # if fi == len(files) - 1: + # a.set_ylim(bottom=0) + for side in ['top', 'right']: + a.spines[side].set_visible(False) + else: + a.remove() + except Exception as e: + print(f'Warning: Plotting error for {f}; {e}') + ax[1].legend() + plt.savefig(Path(save_dir) / 'idetection_profile.png', dpi=200) + + +def save_one_box(xyxy, im, file='image.jpg', gain=1.02, pad=10, square=False, BGR=False, save=True): + # Save image crop as {file} with crop size multiple {gain} and {pad} pixels. Save and/or return crop + xyxy = torch.tensor(xyxy).view(-1, 4) + b = xyxy2xywh(xyxy) # boxes + if square: + b[:, 2:] = b[:, 2:].max(1)[0].unsqueeze(1) # attempt rectangle to square + b[:, 2:] = b[:, 2:] * gain + pad # box wh * gain + pad + xyxy = xywh2xyxy(b).long() + clip_coords(xyxy, im.shape) + crop = im[int(xyxy[0, 1]):int(xyxy[0, 3]), int(xyxy[0, 0]):int(xyxy[0, 2]), ::(1 if BGR else -1)] + if save: + file.parent.mkdir(parents=True, exist_ok=True) # make directory + cv2.imwrite(str(increment_path(file).with_suffix('.jpg')), crop) + return crop diff --git a/utils/torch_utils.py b/utils/torch_utils.py new file mode 100644 index 0000000..c5257c6 --- /dev/null +++ b/utils/torch_utils.py @@ -0,0 +1,329 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +PyTorch utils +""" + +import datetime +import math +import os +import platform +import subprocess +import time +import warnings +from contextlib import contextmanager +from copy import deepcopy +from pathlib import Path + +import torch +import torch.distributed as dist +import torch.nn as nn +import torch.nn.functional as F + +from utils.general import LOGGER + +try: + import thop # for FLOPs computation +except ImportError: + thop = None + +# Suppress PyTorch warnings +warnings.filterwarnings('ignore', message='User provided device_type of \'cuda\', but CUDA is not available. Disabling') + + +@contextmanager +def torch_distributed_zero_first(local_rank: int): + """ + Decorator to make all processes in distributed training wait for each local_master to do something. + """ + if local_rank not in [-1, 0]: + dist.barrier(device_ids=[local_rank]) + yield + if local_rank == 0: + dist.barrier(device_ids=[0]) + + +def date_modified(path=__file__): + # return human-readable file modification date, i.e. '2021-3-26' + t = datetime.datetime.fromtimestamp(Path(path).stat().st_mtime) + return f'{t.year}-{t.month}-{t.day}' + + +def git_describe(path=Path(__file__).parent): # path must be a directory + # return human-readable git description, i.e. v5.0-5-g3e25f1e https://git-scm.com/docs/git-describe + s = f'git -C {path} describe --tags --long --always' + try: + return subprocess.check_output(s, shell=True, stderr=subprocess.STDOUT).decode()[:-1] + except subprocess.CalledProcessError: + return '' # not a git repository + + +def device_count(): + # Returns number of CUDA devices available. Safe version of torch.cuda.device_count(). Only works on Linux. + assert platform.system() == 'Linux', 'device_count() function only works on Linux' + try: + cmd = 'nvidia-smi -L | wc -l' + return int(subprocess.run(cmd, shell=True, capture_output=True, check=True).stdout.decode().split()[-1]) + except Exception: + return 0 + + +def select_device(device='', batch_size=0, newline=True): + # device = 'cpu' or '0' or '0,1,2,3' + s = f'YOLOv5 🚀 {git_describe() or date_modified()} torch {torch.__version__} ' # string + device = str(device).strip().lower().replace('cuda:', '') # to string, 'cuda:0' to '0' + cpu = device == 'cpu' + if cpu: + os.environ['CUDA_VISIBLE_DEVICES'] = '-1' # force torch.cuda.is_available() = False + elif device: # non-cpu device requested + os.environ['CUDA_VISIBLE_DEVICES'] = device # set environment variable - must be before assert is_available() + assert torch.cuda.is_available() and torch.cuda.device_count() >= len(device.replace(',', '')), \ + f"Invalid CUDA '--device {device}' requested, use '--device cpu' or pass valid CUDA device(s)" + + cuda = not cpu and torch.cuda.is_available() + if cuda: + devices = device.split(',') if device else '0' # range(torch.cuda.device_count()) # i.e. 0,1,6,7 + n = len(devices) # device count + if n > 1 and batch_size > 0: # check batch_size is divisible by device_count + assert batch_size % n == 0, f'batch-size {batch_size} not multiple of GPU count {n}' + space = ' ' * (len(s) + 1) + for i, d in enumerate(devices): + p = torch.cuda.get_device_properties(i) + s += f"{'' if i == 0 else space}CUDA:{d} ({p.name}, {p.total_memory / 1024 ** 2:.0f}MiB)\n" # bytes to MB + else: + s += 'CPU\n' + + if not newline: + s = s.rstrip() + LOGGER.info(s.encode().decode('ascii', 'ignore') if platform.system() == 'Windows' else s) # emoji-safe + return torch.device('cuda:0' if cuda else 'cpu') + + +def time_sync(): + # pytorch-accurate time + if torch.cuda.is_available(): + torch.cuda.synchronize() + return time.time() + + +def profile(input, ops, n=10, device=None): + # YOLOv5 speed/memory/FLOPs profiler + # + # Usage: + # input = torch.randn(16, 3, 640, 640) + # m1 = lambda x: x * torch.sigmoid(x) + # m2 = nn.SiLU() + # profile(input, [m1, m2], n=100) # profile over 100 iterations + + results = [] + device = device or select_device() + print(f"{'Params':>12s}{'GFLOPs':>12s}{'GPU_mem (GB)':>14s}{'forward (ms)':>14s}{'backward (ms)':>14s}" + f"{'input':>24s}{'output':>24s}") + + for x in input if isinstance(input, list) else [input]: + x = x.to(device) + x.requires_grad = True + for m in ops if isinstance(ops, list) else [ops]: + m = m.to(device) if hasattr(m, 'to') else m # device + m = m.half() if hasattr(m, 'half') and isinstance(x, torch.Tensor) and x.dtype is torch.float16 else m + tf, tb, t = 0, 0, [0, 0, 0] # dt forward, backward + try: + flops = thop.profile(m, inputs=(x,), verbose=False)[0] / 1E9 * 2 # GFLOPs + except Exception: + flops = 0 + + try: + for _ in range(n): + t[0] = time_sync() + y = m(x) + t[1] = time_sync() + try: + _ = (sum(yi.sum() for yi in y) if isinstance(y, list) else y).sum().backward() + t[2] = time_sync() + except Exception: # no backward method + # print(e) # for debug + t[2] = float('nan') + tf += (t[1] - t[0]) * 1000 / n # ms per op forward + tb += (t[2] - t[1]) * 1000 / n # ms per op backward + mem = torch.cuda.memory_reserved() / 1E9 if torch.cuda.is_available() else 0 # (GB) + s_in = tuple(x.shape) if isinstance(x, torch.Tensor) else 'list' + s_out = tuple(y.shape) if isinstance(y, torch.Tensor) else 'list' + p = sum(list(x.numel() for x in m.parameters())) if isinstance(m, nn.Module) else 0 # parameters + print(f'{p:12}{flops:12.4g}{mem:>14.3f}{tf:14.4g}{tb:14.4g}{str(s_in):>24s}{str(s_out):>24s}') + results.append([p, flops, mem, tf, tb, s_in, s_out]) + except Exception as e: + print(e) + results.append(None) + torch.cuda.empty_cache() + return results + + +def is_parallel(model): + # Returns True if model is of type DP or DDP + return type(model) in (nn.parallel.DataParallel, nn.parallel.DistributedDataParallel) + + +def de_parallel(model): + # De-parallelize a model: returns single-GPU model if model is of type DP or DDP + return model.module if is_parallel(model) else model + + +def initialize_weights(model): + for m in model.modules(): + t = type(m) + if t is nn.Conv2d: + pass # nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu') + elif t is nn.BatchNorm2d: + m.eps = 1e-3 + m.momentum = 0.03 + elif t in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6, nn.SiLU]: + m.inplace = True + + +def find_modules(model, mclass=nn.Conv2d): + # Finds layer indices matching module class 'mclass' + return [i for i, m in enumerate(model.module_list) if isinstance(m, mclass)] + + +def sparsity(model): + # Return global model sparsity + a, b = 0, 0 + for p in model.parameters(): + a += p.numel() + b += (p == 0).sum() + return b / a + + +def prune(model, amount=0.3): + # Prune model to requested global sparsity + import torch.nn.utils.prune as prune + print('Pruning model... ', end='') + for name, m in model.named_modules(): + if isinstance(m, nn.Conv2d): + prune.l1_unstructured(m, name='weight', amount=amount) # prune + prune.remove(m, 'weight') # make permanent + print(' %.3g global sparsity' % sparsity(model)) + + +def fuse_conv_and_bn(conv, bn): + # Fuse convolution and batchnorm layers https://tehnokv.com/posts/fusing-batchnorm-and-conv/ + fusedconv = nn.Conv2d(conv.in_channels, + conv.out_channels, + kernel_size=conv.kernel_size, + stride=conv.stride, + padding=conv.padding, + groups=conv.groups, + bias=True).requires_grad_(False).to(conv.weight.device) + + # prepare filters + w_conv = conv.weight.clone().view(conv.out_channels, -1) + w_bn = torch.diag(bn.weight.div(torch.sqrt(bn.eps + bn.running_var))) + fusedconv.weight.copy_(torch.mm(w_bn, w_conv).view(fusedconv.weight.shape)) + + # prepare spatial bias + b_conv = torch.zeros(conv.weight.size(0), device=conv.weight.device) if conv.bias is None else conv.bias + b_bn = bn.bias - bn.weight.mul(bn.running_mean).div(torch.sqrt(bn.running_var + bn.eps)) + fusedconv.bias.copy_(torch.mm(w_bn, b_conv.reshape(-1, 1)).reshape(-1) + b_bn) + + return fusedconv + + +def model_info(model, verbose=False, img_size=640): + # Model information. img_size may be int or list, i.e. img_size=640 or img_size=[640, 320] + n_p = sum(x.numel() for x in model.parameters()) # number parameters + n_g = sum(x.numel() for x in model.parameters() if x.requires_grad) # number gradients + if verbose: + print(f"{'layer':>5} {'name':>40} {'gradient':>9} {'parameters':>12} {'shape':>20} {'mu':>10} {'sigma':>10}") + for i, (name, p) in enumerate(model.named_parameters()): + name = name.replace('module_list.', '') + print('%5g %40s %9s %12g %20s %10.3g %10.3g' % + (i, name, p.requires_grad, p.numel(), list(p.shape), p.mean(), p.std())) + + try: # FLOPs + from thop import profile + stride = max(int(model.stride.max()), 32) if hasattr(model, 'stride') else 32 + img = torch.zeros((1, model.yaml.get('ch', 3), stride, stride), device=next(model.parameters()).device) # input + flops = profile(deepcopy(model), inputs=(img,), verbose=False)[0] / 1E9 * 2 # stride GFLOPs + img_size = img_size if isinstance(img_size, list) else [img_size, img_size] # expand if int/float + fs = ', %.1f GFLOPs' % (flops * img_size[0] / stride * img_size[1] / stride) # 640x640 GFLOPs + except (ImportError, Exception): + fs = '' + + LOGGER.info(f"Model Summary: {len(list(model.modules()))} layers, {n_p} parameters, {n_g} gradients{fs}") + + +def scale_img(img, ratio=1.0, same_shape=False, gs=32): # img(16,3,256,416) + # scales img(bs,3,y,x) by ratio constrained to gs-multiple + if ratio == 1.0: + return img + else: + h, w = img.shape[2:] + s = (int(h * ratio), int(w * ratio)) # new size + img = F.interpolate(img, size=s, mode='bilinear', align_corners=False) # resize + if not same_shape: # pad/crop img + h, w = (math.ceil(x * ratio / gs) * gs for x in (h, w)) + return F.pad(img, [0, w - s[1], 0, h - s[0]], value=0.447) # value = imagenet mean + + +def copy_attr(a, b, include=(), exclude=()): + # Copy attributes from b to a, options to only include [...] and to exclude [...] + for k, v in b.__dict__.items(): + if (len(include) and k not in include) or k.startswith('_') or k in exclude: + continue + else: + setattr(a, k, v) + + +class EarlyStopping: + # YOLOv5 simple early stopper + def __init__(self, patience=30): + self.best_fitness = 0.0 # i.e. mAP + self.best_epoch = 0 + self.patience = patience or float('inf') # epochs to wait after fitness stops improving to stop + self.possible_stop = False # possible stop may occur next epoch + + def __call__(self, epoch, fitness): + if fitness >= self.best_fitness: # >= 0 to allow for early zero-fitness stage of training + self.best_epoch = epoch + self.best_fitness = fitness + delta = epoch - self.best_epoch # epochs without improvement + self.possible_stop = delta >= (self.patience - 1) # possible stop may occur next epoch + stop = delta >= self.patience # stop training if patience exceeded + if stop: + LOGGER.info(f'Stopping training early as no improvement observed in last {self.patience} epochs. ' + f'Best results observed at epoch {self.best_epoch}, best model saved as best.pt.\n' + f'To update EarlyStopping(patience={self.patience}) pass a new patience value, ' + f'i.e. `python train.py --patience 300` or use `--patience 0` to disable EarlyStopping.') + return stop + + +class ModelEMA: + """ Updated Exponential Moving Average (EMA) from https://github.com/rwightman/pytorch-image-models + Keeps a moving average of everything in the model state_dict (parameters and buffers) + For EMA details see https://www.tensorflow.org/api_docs/python/tf/train/ExponentialMovingAverage + """ + + def __init__(self, model, decay=0.9999, updates=0): + # Create EMA + self.ema = deepcopy(de_parallel(model)).eval() # FP32 EMA + # if next(model.parameters()).device.type != 'cpu': + # self.ema.half() # FP16 EMA + self.updates = updates # number of EMA updates + self.decay = lambda x: decay * (1 - math.exp(-x / 2000)) # decay exponential ramp (to help early epochs) + for p in self.ema.parameters(): + p.requires_grad_(False) + + def update(self, model): + # Update EMA parameters + with torch.no_grad(): + self.updates += 1 + d = self.decay(self.updates) + + msd = de_parallel(model).state_dict() # model state_dict + for k, v in self.ema.state_dict().items(): + if v.dtype.is_floating_point: + v *= d + v += (1 - d) * msd[k].detach() + + def update_attr(self, model, include=(), exclude=('process_group', 'reducer')): + # Update EMA attributes + copy_attr(self.ema, model, include, exclude) diff --git a/val.py b/val.py new file mode 100644 index 0000000..78abbda --- /dev/null +++ b/val.py @@ -0,0 +1,383 @@ +# YOLOv5 🚀 by Ultralytics, GPL-3.0 license +""" +Validate a trained YOLOv5 model accuracy on a custom dataset + +Usage: + $ python path/to/val.py --weights yolov5s.pt --data coco128.yaml --img 640 + +Usage - formats: + $ python path/to/val.py --weights yolov5s.pt # PyTorch + yolov5s.torchscript # TorchScript + yolov5s.onnx # ONNX Runtime or OpenCV DNN with --dnn + yolov5s.xml # OpenVINO + yolov5s.engine # TensorRT + yolov5s.mlmodel # CoreML (MacOS-only) + yolov5s_saved_model # TensorFlow SavedModel + yolov5s.pb # TensorFlow GraphDef + yolov5s.tflite # TensorFlow Lite + yolov5s_edgetpu.tflite # TensorFlow Edge TPU +""" + +import argparse +import json +import os +import sys +from pathlib import Path +from threading import Thread + +import numpy as np +import torch +from tqdm import tqdm + +FILE = Path(__file__).resolve() +ROOT = FILE.parents[0] # YOLOv5 root directory +if str(ROOT) not in sys.path: + sys.path.append(str(ROOT)) # add ROOT to PATH +ROOT = Path(os.path.relpath(ROOT, Path.cwd())) # relative + +from models.common import DetectMultiBackend +from utils.callbacks import Callbacks +from utils.datasets import create_dataloader +from utils.general import (LOGGER, box_iou, check_dataset, check_img_size, check_requirements, check_yaml, + coco80_to_coco91_class, colorstr, increment_path, non_max_suppression, print_args, + scale_coords, xywh2xyxy, xyxy2xywh) +from utils.metrics import ConfusionMatrix, ap_per_class +from utils.plots import output_to_target, plot_images, plot_val_study +from utils.torch_utils import select_device, time_sync + + +def save_one_txt(predn, save_conf, shape, file): + # Save one txt result + gn = torch.tensor(shape)[[1, 0, 1, 0]] # normalization gain whwh + for *xyxy, conf, cls in predn.tolist(): + xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist() # normalized xywh + line = (cls, *xywh, conf) if save_conf else (cls, *xywh) # label format + with open(file, 'a') as f: + f.write(('%g ' * len(line)).rstrip() % line + '\n') + + +def save_one_json(predn, jdict, path, class_map): + # Save one JSON result {"image_id": 42, "category_id": 18, "bbox": [258.15, 41.29, 348.26, 243.78], "score": 0.236} + image_id = int(path.stem) if path.stem.isnumeric() else path.stem + box = xyxy2xywh(predn[:, :4]) # xywh + box[:, :2] -= box[:, 2:] / 2 # xy center to top-left corner + for p, b in zip(predn.tolist(), box.tolist()): + jdict.append({'image_id': image_id, + 'category_id': class_map[int(p[5])], + 'bbox': [round(x, 3) for x in b], + 'score': round(p[4], 5)}) + + +def process_batch(detections, labels, iouv): + """ + Return correct predictions matrix. Both sets of boxes are in (x1, y1, x2, y2) format. + Arguments: + detections (Array[N, 6]), x1, y1, x2, y2, conf, class + labels (Array[M, 5]), class, x1, y1, x2, y2 + Returns: + correct (Array[N, 10]), for 10 IoU levels + """ + correct = torch.zeros(detections.shape[0], iouv.shape[0], dtype=torch.bool, device=iouv.device) + iou = box_iou(labels[:, 1:], detections[:, :4]) + x = torch.where((iou >= iouv[0]) & (labels[:, 0:1] == detections[:, 5])) # IoU above threshold and classes match + if x[0].shape[0]: + matches = torch.cat((torch.stack(x, 1), iou[x[0], x[1]][:, None]), 1).cpu().numpy() # [label, detection, iou] + if x[0].shape[0] > 1: + matches = matches[matches[:, 2].argsort()[::-1]] + matches = matches[np.unique(matches[:, 1], return_index=True)[1]] + # matches = matches[matches[:, 2].argsort()[::-1]] + matches = matches[np.unique(matches[:, 0], return_index=True)[1]] + matches = torch.Tensor(matches).to(iouv.device) + correct[matches[:, 1].long()] = matches[:, 2:3] >= iouv + return correct + + +@torch.no_grad() +def run(data, + weights=None, # model.pt path(s) + batch_size=32, # batch size + imgsz=640, # inference size (pixels) + conf_thres=0.001, # confidence threshold + iou_thres=0.6, # NMS IoU threshold + task='val', # train, val, test, speed or study + device='', # cuda device, i.e. 0 or 0,1,2,3 or cpu + workers=8, # max dataloader workers (per RANK in DDP mode) + single_cls=False, # treat as single-class dataset + augment=False, # augmented inference + verbose=False, # verbose output + save_txt=False, # save results to *.txt + save_hybrid=False, # save label+prediction hybrid results to *.txt + save_conf=False, # save confidences in --save-txt labels + save_json=False, # save a COCO-JSON results file + project=ROOT / 'runs/val', # save to project/name + name='exp', # save to project/name + exist_ok=False, # existing project/name ok, do not increment + half=True, # use FP16 half-precision inference + dnn=False, # use OpenCV DNN for ONNX inference + model=None, + dataloader=None, + save_dir=Path(''), + plots=True, + callbacks=Callbacks(), + compute_loss=None, + ): + # Initialize/load model and set device + training = model is not None + if training: # called by train.py + device, pt, jit, engine = next(model.parameters()).device, True, False, False # get model device, PyTorch model + + half &= device.type != 'cpu' # half precision only supported on CUDA + model.half() if half else model.float() + else: # called directly + device = select_device(device, batch_size=batch_size) + + # Directories + save_dir = increment_path(Path(project) / name, exist_ok=exist_ok) # increment run + (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True) # make dir + + # Load model + model = DetectMultiBackend(weights, device=device, dnn=dnn, data=data) + stride, pt, jit, onnx, engine = model.stride, model.pt, model.jit, model.onnx, model.engine + imgsz = check_img_size(imgsz, s=stride) # check image size + half &= (pt or jit or onnx or engine) and device.type != 'cpu' # FP16 supported on limited backends with CUDA + if pt or jit: + model.model.half() if half else model.model.float() + elif engine: + batch_size = model.batch_size + else: + half = False + batch_size = 1 # export.py models default to batch-size 1 + device = torch.device('cpu') + LOGGER.info(f'Forcing --batch-size 1 square inference shape(1,3,{imgsz},{imgsz}) for non-PyTorch backends') + + # Data + data = check_dataset(data) # check + + # Configure + model.eval() + is_coco = isinstance(data.get('val'), str) and data['val'].endswith('coco/val2017.txt') # COCO dataset + nc = 1 if single_cls else int(data['nc']) # number of classes + iouv = torch.linspace(0.5, 0.95, 10).to(device) # iou vector for mAP@0.5:0.95 + niou = iouv.numel() + + # Dataloader + if not training: + model.warmup(imgsz=(1 if pt else batch_size, 3, imgsz, imgsz), half=half) # warmup + pad = 0.0 if task in ('speed', 'benchmark') else 0.5 + rect = False if task == 'benchmark' else pt # square inference for benchmarks + task = task if task in ('train', 'val', 'test') else 'val' # path to train/val/test images + dataloader = create_dataloader(data[task], imgsz, batch_size, stride, single_cls, pad=pad, rect=rect, + workers=workers, prefix=colorstr(f'{task}: '))[0] + + seen = 0 + confusion_matrix = ConfusionMatrix(nc=nc) + names = {k: v for k, v in enumerate(model.names if hasattr(model, 'names') else model.module.names)} + class_map = coco80_to_coco91_class() if is_coco else list(range(1000)) + s = ('%20s' + '%11s' * 6) % ('Class', 'Images', 'Labels', 'P', 'R', 'mAP@.5', 'mAP@.5:.95') + dt, p, r, f1, mp, mr, map50, map = [0.0, 0.0, 0.0], 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0 + loss = torch.zeros(3, device=device) + jdict, stats, ap, ap_class = [], [], [], [] + pbar = tqdm(dataloader, desc=s, bar_format='{l_bar}{bar:10}{r_bar}{bar:-10b}') # progress bar + for batch_i, (im, targets, paths, shapes) in enumerate(pbar): + t1 = time_sync() + if pt or jit or engine: + im = im.to(device, non_blocking=True) + targets = targets.to(device) + im = im.half() if half else im.float() # uint8 to fp16/32 + im /= 255 # 0 - 255 to 0.0 - 1.0 + nb, _, height, width = im.shape # batch size, channels, height, width + t2 = time_sync() + dt[0] += t2 - t1 + + # Inference + out, train_out = model(im) if training else model(im, augment=augment, val=True) # inference, loss outputs + dt[1] += time_sync() - t2 + + # Loss + if compute_loss: + loss += compute_loss([x.float() for x in train_out], targets)[1] # box, obj, cls + + # NMS + targets[:, 2:] *= torch.Tensor([width, height, width, height]).to(device) # to pixels + lb = [targets[targets[:, 0] == i, 1:] for i in range(nb)] if save_hybrid else [] # for autolabelling + t3 = time_sync() + out = non_max_suppression(out, conf_thres, iou_thres, labels=lb, multi_label=True, agnostic=single_cls) + dt[2] += time_sync() - t3 + + # Metrics + for si, pred in enumerate(out): + labels = targets[targets[:, 0] == si, 1:] + nl = len(labels) + tcls = labels[:, 0].tolist() if nl else [] # target class + path, shape = Path(paths[si]), shapes[si][0] + seen += 1 + + if len(pred) == 0: + if nl: + stats.append((torch.zeros(0, niou, dtype=torch.bool), torch.Tensor(), torch.Tensor(), tcls)) + continue + + # Predictions + if single_cls: + pred[:, 5] = 0 + predn = pred.clone() + scale_coords(im[si].shape[1:], predn[:, :4], shape, shapes[si][1]) # native-space pred + + # Evaluate + if nl: + tbox = xywh2xyxy(labels[:, 1:5]) # target boxes + scale_coords(im[si].shape[1:], tbox, shape, shapes[si][1]) # native-space labels + labelsn = torch.cat((labels[:, 0:1], tbox), 1) # native-space labels + correct = process_batch(predn, labelsn, iouv) + if plots: + confusion_matrix.process_batch(predn, labelsn) + else: + correct = torch.zeros(pred.shape[0], niou, dtype=torch.bool) + stats.append((correct.cpu(), pred[:, 4].cpu(), pred[:, 5].cpu(), tcls)) # (correct, conf, pcls, tcls) + + # Save/log + if save_txt: + save_one_txt(predn, save_conf, shape, file=save_dir / 'labels' / (path.stem + '.txt')) + if save_json: + save_one_json(predn, jdict, path, class_map) # append to COCO-JSON dictionary + callbacks.run('on_val_image_end', pred, predn, path, names, im[si]) + + # Plot images + if plots and batch_i < 3: + f = save_dir / f'val_batch{batch_i}_labels.jpg' # labels + Thread(target=plot_images, args=(im, targets, paths, f, names), daemon=True).start() + f = save_dir / f'val_batch{batch_i}_pred.jpg' # predictions + Thread(target=plot_images, args=(im, output_to_target(out), paths, f, names), daemon=True).start() + + # Compute metrics + stats = [np.concatenate(x, 0) for x in zip(*stats)] # to numpy + if len(stats) and stats[0].any(): + tp, fp, p, r, f1, ap, ap_class = ap_per_class(*stats, plot=plots, save_dir=save_dir, names=names) + ap50, ap = ap[:, 0], ap.mean(1) # AP@0.5, AP@0.5:0.95 + mp, mr, map50, map = p.mean(), r.mean(), ap50.mean(), ap.mean() + nt = np.bincount(stats[3].astype(np.int64), minlength=nc) # number of targets per class + else: + nt = torch.zeros(1) + + # Print results + pf = '%20s' + '%11i' * 2 + '%11.3g' * 4 # print format + LOGGER.info(pf % ('all', seen, nt.sum(), mp, mr, map50, map)) + + # Print results per class + if (verbose or (nc < 50 and not training)) and nc > 1 and len(stats): + for i, c in enumerate(ap_class): + LOGGER.info(pf % (names[c], seen, nt[c], p[i], r[i], ap50[i], ap[i])) + + # Print speeds + t = tuple(x / seen * 1E3 for x in dt) # speeds per image + if not training: + shape = (batch_size, 3, imgsz, imgsz) + LOGGER.info(f'Speed: %.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape {shape}' % t) + + # Plots + if plots: + confusion_matrix.plot(save_dir=save_dir, names=list(names.values())) + callbacks.run('on_val_end') + + # Save JSON + if save_json and len(jdict): + w = Path(weights[0] if isinstance(weights, list) else weights).stem if weights is not None else '' # weights + anno_json = str(Path(data.get('path', '../coco')) / 'annotations/instances_val2017.json') # annotations json + pred_json = str(save_dir / f"{w}_predictions.json") # predictions json + LOGGER.info(f'\nEvaluating pycocotools mAP... saving {pred_json}...') + with open(pred_json, 'w') as f: + json.dump(jdict, f) + + try: # https://github.com/cocodataset/cocoapi/blob/master/PythonAPI/pycocoEvalDemo.ipynb + check_requirements(['pycocotools']) + from pycocotools.coco import COCO + from pycocotools.cocoeval import COCOeval + + anno = COCO(anno_json) # init annotations api + pred = anno.loadRes(pred_json) # init predictions api + eval = COCOeval(anno, pred, 'bbox') + if is_coco: + eval.params.imgIds = [int(Path(x).stem) for x in dataloader.dataset.img_files] # image IDs to evaluate + eval.evaluate() + eval.accumulate() + eval.summarize() + map, map50 = eval.stats[:2] # update results (mAP@0.5:0.95, mAP@0.5) + except Exception as e: + LOGGER.info(f'pycocotools unable to run: {e}') + + # Return results + model.float() # for training + if not training: + s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else '' + LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}{s}") + maps = np.zeros(nc) + map + for i, c in enumerate(ap_class): + maps[c] = ap[i] + return (mp, mr, map50, map, *(loss.cpu() / len(dataloader)).tolist()), maps, t + + +def parse_opt(): + parser = argparse.ArgumentParser() + parser.add_argument('--data', type=str, default=ROOT / 'data/coco128.yaml', help='dataset.yaml path') + parser.add_argument('--weights', nargs='+', type=str, default=ROOT / 'yolov5s.pt', help='model.pt path(s)') + parser.add_argument('--batch-size', type=int, default=32, help='batch size') + parser.add_argument('--imgsz', '--img', '--img-size', type=int, default=640, help='inference size (pixels)') + parser.add_argument('--conf-thres', type=float, default=0.001, help='confidence threshold') + parser.add_argument('--iou-thres', type=float, default=0.6, help='NMS IoU threshold') + parser.add_argument('--task', default='val', help='train, val, test, speed or study') + parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu') + parser.add_argument('--workers', type=int, default=8, help='max dataloader workers (per RANK in DDP mode)') + parser.add_argument('--single-cls', action='store_true', help='treat as single-class dataset') + parser.add_argument('--augment', action='store_true', help='augmented inference') + parser.add_argument('--verbose', action='store_true', help='report mAP by class') + parser.add_argument('--save-txt', action='store_true', help='save results to *.txt') + parser.add_argument('--save-hybrid', action='store_true', help='save label+prediction hybrid results to *.txt') + parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels') + parser.add_argument('--save-json', action='store_true', help='save a COCO-JSON results file') + parser.add_argument('--project', default=ROOT / 'runs/val', help='save to project/name') + parser.add_argument('--name', default='exp', help='save to project/name') + parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment') + parser.add_argument('--half', action='store_true', help='use FP16 half-precision inference') + parser.add_argument('--dnn', action='store_true', help='use OpenCV DNN for ONNX inference') + opt = parser.parse_args() + opt.data = check_yaml(opt.data) # check YAML + opt.save_json |= opt.data.endswith('coco.yaml') + opt.save_txt |= opt.save_hybrid + print_args(FILE.stem, opt) + return opt + + +def main(opt): + check_requirements(requirements=ROOT / 'requirements.txt', exclude=('tensorboard', 'thop')) + + if opt.task in ('train', 'val', 'test'): # run normally + if opt.conf_thres > 0.001: # https://github.com/ultralytics/yolov5/issues/1466 + LOGGER.info(f'WARNING: confidence threshold {opt.conf_thres} >> 0.001 will produce invalid mAP values.') + run(**vars(opt)) + + else: + weights = opt.weights if isinstance(opt.weights, list) else [opt.weights] + opt.half = True # FP16 for fastest results + if opt.task == 'speed': # speed benchmarks + # python val.py --task speed --data coco.yaml --batch 1 --weights yolov5n.pt yolov5s.pt... + opt.conf_thres, opt.iou_thres, opt.save_json = 0.25, 0.45, False + for opt.weights in weights: + run(**vars(opt), plots=False) + + elif opt.task == 'study': # speed vs mAP benchmarks + # python val.py --task study --data coco.yaml --iou 0.7 --weights yolov5n.pt yolov5s.pt... + for opt.weights in weights: + f = f'study_{Path(opt.data).stem}_{Path(opt.weights).stem}.txt' # filename to save to + x, y = list(range(256, 1536 + 128, 128)), [] # x axis (image sizes), y axis + for opt.imgsz in x: # img-size + LOGGER.info(f'\nRunning {f} --imgsz {opt.imgsz}...') + r, _, t = run(**vars(opt), plots=False) + y.append(r + t) # results and times + np.savetxt(f, y, fmt='%10.4g') # save + os.system('zip -r study.zip study_*.txt') + plot_val_study(x=x) # plot + + +if __name__ == "__main__": + opt = parse_opt() + main(opt)