From dae0700d5a0c567ec94920b3c36328025bfd067b Mon Sep 17 00:00:00 2001 From: Sahan Paliskara Date: Wed, 26 Oct 2022 18:23:26 -0700 Subject: [PATCH] Shorten ReadMe ghstack-source-id: f41d888c40f2a2505c75bc63f5f531a16c483e80 Pull Request resolved: https://github.com/pytorch/multipy/pull/237 --- README.md | 305 ++---------------------------------------- docs/source/index.rst | 5 +- 2 files changed, 11 insertions(+), 299 deletions(-) diff --git a/README.md b/README.md index a7128a39..90cca945 100644 --- a/README.md +++ b/README.md @@ -17,7 +17,7 @@ Requirements: > ℹ️ This is project is in Beta. `torch::deploy` is ready for use in production environments but may have some rough edges that we're continuously working on improving. We're always interested in hearing feedback and usecases that you might have. Feel free to reach out! -## Installation +## The Easy Path to Installation ### Building via Docker @@ -40,92 +40,12 @@ docker run --rm multipy multipy/runtime/build/test_deploy ### Installing via `pip install` -We support installing both python modules and the runtime libs using `pip -install`, with the caveat of having to manually install the C++ dependencies -first. This serves as a single-command source build, essentially being a wrapper -around `python setup.py develop`, once all the dependencies have been installed. +The second easiest way of using `torch::deploy` is through our single command `pip install`. +However, the C++ dependencies have to manually be installed before hand. Specifically a `-fpic` +enabled version of python. For full instructions for getting the C++ dependencies up and +running and more detailed guide on `torch::deploy` installation can be found [here](https://pytorch.org/multipy/latest/setup.html#installing-via-pip-install). - -To start with, the multipy repo should be cloned first: - -```shell -git clone --recurse-submodules https://github.com/pytorch/multipy.git -cd multipy - -# (optional) if using existing checkout -git submodule sync && git submodule update --init --recursive -``` - -#### Installing System Dependencies - -The runtime system dependencies are specified in `build-requirements.txt`. To install them on Debian-based systems, one could run: - -```shell -sudo apt update -xargs sudo apt install -y -qq --no-install-recommends **NOTE** We support Python versions 3.7 through 3.10 for `multipy`; note that for `conda` environments the `libpython-static` libraries are available for `3.8` onwards. With `virtualenv`/`pyenv` any version from 3.7 through 3.10 can be used, as the PIC library is built explicitly. - -
-Click to expand - -Example commands for installing conda: -```shell -curl -fsSL -v -o ~/miniconda.sh -O https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh && \ -chmod +x ~/miniconda.sh && \ -~/miniconda.sh -b -p /opt/conda && \ -rm ~/miniconda.sh -``` -Virtualenv / pyenv can be installed as follows: -```shell -pip3 install virtualenv -git clone https://github.com/pyenv/pyenv.git ~/.pyenv -``` -
- - -#### Installing python, pytorch and related dependencies - -Multipy requires the latest version of pytorch to run models successfully, and we recommend fetching the latest _nightlies_ for pytorch and also cuda, if required. - -##### In a `conda` environment, we would do the following: -```shell -conda create -n newenv -conda activate newenv -conda install python=3.8 -conda install -c conda-forge libpython-static=3.8 - -# cuda -conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch-nightly - -# cpu only -conda install pytorch torchvision torchaudio cpuonly -c pytorch-nightly -``` - -##### For a `pyenv` / `virtualenv` setup, one could do: -```shell -export CFLAGS="-fPIC -g" -~/.pyenv/bin/pyenv install --force 3.8.6 -virtualenv -p ~/.pyenv/versions/3.8.6/bin/python3 ~/venvs/multipy -source ~/venvs/multipy/bin/activate -pip install -r dev-requirements.txt - -# cuda -pip3 install --pre torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/nightly/cu113 - -# cpu only -pip3 install --pre torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/nightly/cpu - -``` - -#### Running `pip install` - -Once all the dependencies are successfully installed, most importantly including a PIC-library of python and the latest nightly of pytorch, we can run the following, in either `conda` or `virtualenv`, to install both the python modules and the runtime/interpreter libraries: +Once all the dependencies are successfully installed, you can run the following, in either `conda` or `virtualenv`, to install both the python modules and the runtime/interpreter libraries: ```shell # from base multipy directory pip install -e . @@ -137,216 +57,9 @@ Alternatively, one can install only the python modules without invoking `cmake` pip install -e . --install-option="--cmakeoff" ``` -> **NOTE** As of 10/11/2022 the linking of prebuilt static fPIC versions of python downloaded from `conda-forge` can be problematic on certain systems (for example Centos 8), with linker errors like `libpython_multipy.a: error adding symbols: File format not recognized`. This seems to be an issue with `binutils`, and the steps in https://wiki.gentoo.org/wiki/Project:Toolchain/Binutils_2.32_upgrade_notes/elfutils_0.175:_unable_to_initialize_decompress_status_for_section_.debug_info can help. Alternatively, the user can go with the `virtualenv`/`pyenv` flow above. - -## Development - -### Manually building `multipy::runtime` from source - -Both `docker` and `pip install` options above are wrappers around the `cmake` -build of multipy's runtime. For development purposes it's often helpful to -invoke `cmake` separately. - -See the install section for how to correctly setup the Python environment. - -```bash -# checkout repo -git clone --recurse-submodules https://github.com/pytorch/multipy.git -cd multipy - -# (optional) if using existing checkout -git submodule sync && git submodule update --init --recursive - -cd multipy -# install python parts of `torch::deploy` in multipy/multipy/utils -pip install -e . --install-option="--cmakeoff" - -cd multipy/runtime - -# configure runtime to build/ -cmake -S . -B build -# if you need to override the ABI setting you can pass -cmake -S . -B build -D_GLIBCXX_USE_CXX11_ABI=<0/1> - -# compile the files in build/ -cmake --build build --config Release -j -``` - -### Running unit tests for `multipy::runtime` - -We first need to generate the neccessary examples. First make sure your python environment has [torch](https://pytorch.org). Afterwards, once `multipy::runtime` is built, run the following (executed automatically for `docker` and `pip` above): - -``` -cd multipy/multipy/runtime -python example/generate_examples.py -cd build -./test_deploy -``` - -## Examples - -See the [examples directory](./examples) for complete examples. - -### Packaging a model `for multipy::runtime` - -``multipy::runtime`` can load and run Python models that are packaged with -``torch.package``. You can learn more about ``torch.package`` in the ``torch.package`` [documentation](https://pytorch.org/docs/stable/package.html#tutorials). - -For now, let's create a simple model that we can load and run in ``multipy::runtime``. - -```python -from torch.package import PackageExporter -import torchvision - -# Instantiate some model -model = torchvision.models.resnet.resnet18() - -# Package and export it. -with PackageExporter("my_package.pt") as e: - e.intern("torchvision.**") - e.extern("numpy.**") - e.extern("sys") - e.extern("PIL.*") - e.extern("typing_extensions") - e.save_pickle("model", "model.pkl", model) -``` - -Note that since "numpy", "sys", "PIL" were marked as "extern", `torch.package` will -look for these dependencies on the system that loads this package. They will not be packaged -with the model. - -Now, there should be a file named ``my_package.pt`` in your working directory. - -
- -### Load the model in C++ -```cpp -#include -#include -#include -#include - -#include -#include - -int main(int argc, const char* argv[]) { - if (argc != 2) { - std::cerr << "usage: example-app \n"; - return -1; - } - - // Start an interpreter manager governing 4 embedded interpreters. - std::shared_ptr env = - std::make_shared( - std::getenv("PATH_TO_EXTERN_PYTHON_PACKAGES") // Ensure to set this environment variable (e.g. /home/user/anaconda3/envs/multipy-example/lib/python3.8/site-packages) - ); - multipy::runtime::InterpreterManager manager(4, env); - - try { - // Load the model from the multipy.package. - multipy::runtime::Package package = manager.loadPackage(argv[1]); - multipy::runtime::ReplicatedObj model = package.loadPickle("model", "model.pkl"); - } catch (const c10::Error& e) { - std::cerr << "error loading the model\n"; - std::cerr << e.msg(); - return -1; - } - - std::cout << "ok\n"; -} - -``` - -This small program introduces many of the core concepts of ``multipy::runtime``. - -An ``InterpreterManager`` abstracts over a collection of independent Python -interpreters, allowing you to load balance across them when running your code. - -``PathEnvironment`` enables you to specify the location of Python -packages on your system which are external, but necessary, for your model. - -Using the ``InterpreterManager::loadPackage`` method, you can load a -``multipy.package`` from disk and make it available to all interpreters. - -``Package::loadPickle`` allows you to retrieve specific Python objects -from the package, like the ResNet model we saved earlier. - -Finally, the model itself is a ``ReplicatedObj``. This is an abstract handle to -an object that is replicated across multiple interpreters. When you interact -with a ``ReplicatedObj`` (for example, by calling ``forward``), it will select -an free interpreter to execute that interaction. - -
- -### Build and execute the C++ example - -Assuming the above C++ program was stored in a file called, `example-app.cpp`, a -minimal `CMakeLists.txt` file would look like: - -```cmake -cmake_minimum_required(VERSION 3.12 FATAL_ERROR) -project(multipy_tutorial) - -set(MULTIPY_PATH ".." CACHE PATH "The repo where multipy is built or the PYTHONPATH") - -# include the multipy utils to help link against -include(${MULTIPY_PATH}/multipy/runtime/utils.cmake) - -# add headers from multipy -include_directories(${MULTIPY_PATH}) - -# link the multipy prebuilt binary -add_library(multipy_internal STATIC IMPORTED) -set_target_properties(multipy_internal - PROPERTIES - IMPORTED_LOCATION - ${MULTIPY_PATH}/multipy/runtime/build/libtorch_deploy.a) -caffe2_interface_library(multipy_internal multipy) - -add_executable(example-app example-app.cpp) -target_link_libraries(example-app PUBLIC "-Wl,--no-as-needed -rdynamic" dl pthread util multipy c10 torch_cpu) -``` - -Currently, it is necessary to build ``multipy::runtime`` as a static library. -In order to correctly link to a static library, the utility ``caffe2_interface_library`` -is used to appropriately set and unset ``--whole-archive`` flag. - -Furthermore, the ``-rdynamic`` flag is needed when linking to the executable -to ensure that symbols are exported to the dynamic table, making them accessible -to the deploy interpreters (which are dynamically loaded). - -**Updating LIBRARY_PATH and LD_LIBRARY_PATH** - -In order to locate dependencies provided by PyTorch (e.g. `libshm`), we need to update the `LIBRARY_PATH` and `LD_LIBRARY_PATH` environment variables to include the path to PyTorch's C++ libraries. If you installed PyTorch using pip or conda, this path is usually in the site-packages. An example of this is provided below. - -```bash -export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/home/user/anaconda3/envs/multipy-example/lib/python3.8/site-packages/torch/lib" -export LIBRARY_PATH="$LIBRARY_PATH:/home/user/anaconda3/envs/multipy-example/lib/python3.8/site-packages/torch/lib" -``` - -The last step is configuring and building the project. Assuming that our code -directory is laid out like this: - -``` -example-app/ - CMakeLists.txt - example-app.cpp -``` - - -We can now run the following commands to build the application from within the -``example-app/`` folder: - -```bash -cmake -S . -B build -DMULTIPY_PATH="/home/user/repos/multipy" # the parent directory of multipy (i.e. the git repo) -cmake --build build --config Release -j -``` - -Now we can run our app: - -```bash -./example-app /path/to/my_package.pt -``` +## Getting Started with `torch::deploy` +Once you have `torch::deploy` built, check out our [tutorials](https://pytorch.org/multipy/latest/tutorials/tutorial_root.html) and +[API documentation](https://pytorch.org/multipy/latest/api/library_root.html). ## Contributing diff --git a/docs/source/index.rst b/docs/source/index.rst index 045b4263..70babad6 100644 --- a/docs/source/index.rst +++ b/docs/source/index.rst @@ -3,9 +3,8 @@ ``torch::deploy`` [Beta] ===================== -``torch::deploy`` is a system that allows you to load multiple python interpreters which execute PyTorch models, and run them in a single C++ process. Effectively, it allows people to multithread their pytorch models. -For more information on how torch::deploy works please see the related `arXiv paper `_. We plan to further generalize ``torch::deploy`` into a more generic system, ``multipy::runtime``, -which is more suitable for arbitrary python programs rather than just pytorch applications. +``torch::deploy`` (MultiPy for non-PyTorch use cases) is a C++ library that enables you to run eager mode PyTorch models in production without any modifications to your model to support tracing. ``torch::deploy`` provides a way to run using multiple independent Python interpreters in a single process without a shared global interpreter lock (GIL). +For more information on how ``torch::deploy`` works please see the related `arXiv paper `_. Documentation