From e6e3103e33c06836b3b436ecf6cbed0bb7fae46f Mon Sep 17 00:00:00 2001 From: OHASHI Takeshi Date: Sun, 11 Sep 2022 17:11:55 +0900 Subject: [PATCH 1/2] Support PyTorch1.11 and OpenCV-4.5.5 --- Dockerfile | 16 ++++++++++------ docs/install.md | 16 ++++++++-------- requirements.txt | 8 +++++--- utils/nms_rotated/src/poly_nms_cuda.cu | 21 ++++++++++----------- 4 files changed, 33 insertions(+), 28 deletions(-) diff --git a/Dockerfile b/Dockerfile index 9a55005a..f605ade6 100644 --- a/Dockerfile +++ b/Dockerfile @@ -1,18 +1,22 @@ # 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 +FROM nvcr.io/nvidia/pytorch:22.04-py3 +#FROM nvcr.io/nvidia/pytorch:21.04-py3 # Install linux packages -RUN apt update && apt install -y zip htop screen libgl1-mesa-glx +RUN apt update && apt install -y zip htop tmux libgl1-mesa-glx # Install python dependencies COPY requirements.txt . RUN python -m pip install --upgrade pip -RUN pip uninstall -y nvidia-tensorboard nvidia-tensorboard-plugin-dlprof -RUN pip install --no-cache -r requirements.txt coremltools onnx gsutil notebook wandb>=0.12.2 -RUN pip install --no-cache -U torch torchvision numpy Pillow -# RUN pip install --no-cache torch==1.10.0+cu113 torchvision==0.11.1+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html +# RUN pip uninstall -y nvidia-tensorboard nvidia-tensorboard-plugin-dlprof +RUN pip install --no-cache -r requirements.txt +RUN pip install --no-cache coremltools onnx gsutil notebook wandb>=0.12.2 +RUN pip install --no-cache -U numpy Pillow +RUN pip install --no-cache torch==1.12.1+cu116 torchvision==0.13.1+cu116 torchaudio==0.12.1+cu116 --extra-index-url https://download.pytorch.org/whl/cu116 +#RUN pip install --no-cache torch==1.12.1+cu113 torchvision==0.13.1+cu113 torchaudio==0.12.1+cu113 --extra-index-url https://download.pytorch.org/whl/cu113 +RUN pip install --no-cache roboflow shapely # Create working directory RUN mkdir -p /usr/src/app diff --git a/docs/install.md b/docs/install.md index 0da4a94a..8a75eaa7 100644 --- a/docs/install.md +++ b/docs/install.md @@ -2,20 +2,20 @@ ## Requirements * Linux **(Recommend)**, Windows **(not Recommend, Please refer to this [issue](https://github.com/hukaixuan19970627/yolov5_obb/issues/224) if you have difficulty in generating utils/nms_rotated_ext.cpython-XX-XX-XX-XX.so)** * Python 3.7+ -* PyTorch ≥ 1.7 -* CUDA 9.0 or higher +* PyTorch ≥ 1.11 +* CUDA 11.3 or higher I have tested the following versions of OS and softwares： -* OS：Ubuntu 16.04/18.04 -* CUDA: 10.0/10.1/10.2/11.3 +* OS：Ubuntu 20.04 +* CUDA: 11.3/11.6 ## Install **CUDA Driver Version ≥ CUDA Toolkit Version(runtime version) = torch.version.cuda** a. Create a conda virtual environment and activate it, e.g., ``` -conda create -n Py39_Torch1.10_cu11.3 python=3.9 -y -source activate Py39_Torch1.10_cu11.3 +conda create -n Py39_Torch1.11_cu11.3 python=3.9 -y +source activate Py39_Torch1.11_cu11.3 ``` b. Make sure your CUDA runtime api version ≤ CUDA driver version. (for example 11.3 ≤ 11.4) ``` @@ -24,7 +24,7 @@ nvidia-smi ``` c. Install PyTorch and torchvision following the [official instructions](https://pytorch.org/), Make sure cudatoolkit version same as CUDA runtime api version, e.g., ``` -pip3 install torch==1.10.1+cu113 torchvision==0.11.2+cu113 torchaudio==0.10.1+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html +pip3 install --no-cache torch==1.12.1+cu116 torchvision==0.13.1+cu116 torchaudio==0.12.1+cu116 --extra-index-url https://download.pytorch.org/whl/cu116 nvcc -V python >>> import torch @@ -33,7 +33,7 @@ python ``` d. Clone the yolov5-obb repository. ``` -git clone https://github.com/hukaixuan19970627/yolov5_obb.git +git clone https://github.com/ohashi/yolov5_obb.git cd yolov5_obb ``` e. Install yolov5-obb. diff --git a/requirements.txt b/requirements.txt index 8ff89d1f..e3a7a76c 100644 --- a/requirements.txt +++ b/requirements.txt @@ -3,13 +3,15 @@ # Base ---------------------------------------- matplotlib>=3.2.2 numpy>=1.18.5 -opencv-python>=4.5.4 +opencv-python==4.5.5.64 +opencv-contrib-python==4.5.5.64 +opencv-python-headless==4.5.5.64 Pillow>=7.1.2 PyYAML>=5.3.1 requests>=2.23.0 scipy>=1.4.1 -torch>=1.7.0 -torchvision>=0.8.1 +#torch>=1.7.0 +#torchvision>=0.8.1 tqdm>=4.41.0 # Logging ------------------------------------- diff --git a/utils/nms_rotated/src/poly_nms_cuda.cu b/utils/nms_rotated/src/poly_nms_cuda.cu index efa3286b..ef58f609 100644 --- a/utils/nms_rotated/src/poly_nms_cuda.cu +++ b/utils/nms_rotated/src/poly_nms_cuda.cu @@ -1,8 +1,7 @@ #include +#include #include - -#include -#include +#include #include #include @@ -188,7 +187,7 @@ __global__ void poly_nms_kernel(const int n_polys, const float nms_overlap_thres t |= 1ULL << i; } } - const int col_blocks = THCCeilDiv(n_polys, threadsPerBlock); + const int col_blocks = at::ceil_div(n_polys, threadsPerBlock); dev_mask[cur_box_idx * col_blocks + col_start] = t; } } @@ -206,18 +205,18 @@ at::Tensor poly_nms_cuda(const at::Tensor boxes, float nms_overlap_thresh) { int boxes_num = boxes.size(0); - const int col_blocks = THCCeilDiv(boxes_num, threadsPerBlock); + const int col_blocks = at::ceil_div(boxes_num, threadsPerBlock); scalar_t* boxes_dev = boxes_sorted.data_ptr(); - THCState *state = at::globalContext().lazyInitCUDA(); + at::globalContext().lazyInitCUDA(); unsigned long long* mask_dev = NULL; - mask_dev = (unsigned long long*) THCudaMalloc(state, boxes_num * col_blocks * sizeof(unsigned long long)); + mask_dev = (unsigned long long*) c10::cuda::CUDACachingAllocator::raw_alloc(boxes_num * col_blocks * sizeof(unsigned long long)); - dim3 blocks(THCCeilDiv(boxes_num, threadsPerBlock), - THCCeilDiv(boxes_num, threadsPerBlock)); + dim3 blocks(at::ceil_div(boxes_num, threadsPerBlock), + at::ceil_div(boxes_num, threadsPerBlock)); dim3 threads(threadsPerBlock); poly_nms_kernel<<>>(boxes_num, nms_overlap_thresh, @@ -225,7 +224,7 @@ at::Tensor poly_nms_cuda(const at::Tensor boxes, float nms_overlap_thresh) { mask_dev); std::vector mask_host(boxes_num * col_blocks); - THCudaCheck(cudaMemcpyAsync( + C10_CUDA_CHECK(cudaMemcpyAsync( &mask_host[0], mask_dev, sizeof(unsigned long long) * boxes_num * col_blocks, @@ -253,7 +252,7 @@ at::Tensor poly_nms_cuda(const at::Tensor boxes, float nms_overlap_thresh) { } } - THCudaFree(state, mask_dev); + c10::cuda::CUDACachingAllocator::raw_delete(mask_dev); return order_t.index({ keep.narrow(/*dim=*/0, /*start=*/0, /*length=*/num_to_keep).to( From 48b59f470ba837f26c924939238625de6ab1dcc6 Mon Sep 17 00:00:00 2001 From: OHASHI Takeshi Date: Tue, 13 Sep 2022 19:49:12 +0900 Subject: [PATCH 2/2] Bug fixed utils/loss.py --- utils/loss.py | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/utils/loss.py b/utils/loss.py index eee5727e..39ca7cb8 100644 --- a/utils/loss.py +++ b/utils/loss.py @@ -211,7 +211,7 @@ def build_targets(self, p, targets): # ttheta, tgaussian_theta = [], [] tgaussian_theta = [] # gain = torch.ones(7, device=targets.device) # normalized to gridspace gain - feature_wh = torch.ones(2, device=targets.device) # feature_wh + feature_wh = torch.ones(2, device=targets.device).long() # feature_wh ai = torch.arange(na, device=targets.device).float().view(na, 1).repeat(1, nt) # same as .repeat_interleave(nt) # targets (tensor): (n_gt_all_batch, c) -> (na, n_gt_all_batch, c) -> (na, n_gt_all_batch, c+1) # targets (tensor): (na, n_gt_all_batch, [img_index, clsid, cx, cy, l, s, theta, gaussian_θ_labels, anchor_index]]) @@ -226,7 +226,7 @@ def build_targets(self, p, targets): for i in range(self.nl): anchors = self.anchors[i] # gain[2:6] = torch.tensor(p[i].shape)[[3, 2, 3, 2]] # xyxy gain=[1, 1, w, h, w, h, 1, 1] - feature_wh[0:2] = torch.tensor(p[i].shape)[[3, 2]] # xyxy gain=[w_f, h_f] + feature_wh[0:2] = torch.tensor(p[i].shape)[[3, 2]].long() # xyxy gain=[w_f, h_f] # Match targets to anchors # t = targets * gain # xywh featuremap pixel @@ -240,9 +240,9 @@ def build_targets(self, p, targets): t = t[j] # filter; Tensor.size(n_filter1, c+1) # Offsets - gxy = t[:, 2:4] # grid xy; (n_filter1, 2) + gxy = t[:, 2:4].long() # grid xy; (n_filter1, 2) # gxi = gain[[2, 3]] - gxy # inverse - gxi = feature_wh[[0, 1]] - gxy # inverse + gxi = (feature_wh[[0, 1]] - gxy).long() # 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)) # (5, n_filter1) @@ -254,17 +254,17 @@ def build_targets(self, p, targets): # Define, t (tensor): (n_filter2, [img_index, clsid, cx, cy, l, s, theta, gaussian_θ_labels, anchor_index]) b, c = t[:, :2].long().T # image, class; (n_filter2) - gxy = t[:, 2:4] # grid xy - gwh = t[:, 4:6] # grid wh + gxy = t[:, 2:4].long() # grid xy + gwh = t[:, 4:6].long() # grid wh # theta = t[:, 6] - gaussian_theta_labels = t[:, 7:-1] + gaussian_theta_labels = t[:, 7:-1].long() gij = (gxy - offsets).long() gi, gj = gij.T # grid xy indices # Append a = t[:, -1].long() # anchor indices 取整 # 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, feature_wh[1] - 1), gi.clamp_(0, feature_wh[0] - 1))) # image, anchor, grid indices + indices.append((b, a, gj.clamp_(0, feature_wh[1] - 1).long(), gi.clamp_(0, feature_wh[0] - 1).long())) # image, anchor, grid indices tbox.append(torch.cat((gxy - gij, gwh), 1)) # box anch.append(anchors[a]) # anchors tcls.append(c) # class