在腾讯云IDE上部署 YOLOv5-v6.0 环境

0 前言

本章节主要记录在远古设备上使用v6.0版本的yolov5的搭建，主要原因还是经常性有人来问，太长了。

这次的平台在 腾讯云IDE 上搭建，2024-11-18仍可免费。

• 硬件与软件列表：

  • NVIDIA Tesla T4 (8h32g 8TFlops SP)
  • Linux VM-0-203-ubuntu 5.4.0-166-generic #183-Ubuntu SMP Mon Oct 2 11:28:33 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux
  • CUDA12.0 （用nvidia-smi查看）

1 环境

先把yolov5-v6.0的pull到本地，https://github.com/ultralytics/yolov5/releases/tag/v6.0 。

1.1 Conda 环境

conda create -n yolov5-v6.0 python=3.8
conda activate yolov5-v6.0

1.2 pip 环境

一般都是 pytorch 版本不好找，去 Pytroch官方 找比较方便，直接搜对应的CUDA版本， 没有就往下降，如CUDA12.0，选择CUDA11.8的也行

# CUDA 11.8
conda install pytorch==2.0.0 torchvision==0.15.0 torchaudio==2.0.0 pytorch-cuda=11.8 -c pytorch -c nvidia

不建议使用 pip install -r requirements.txt ，单独安装比较好：

# Base ----------------------------------------
pip install matplotlib==3.2.2
pip install numpy==1.18.5
pip install opencv-python==4.1.2.30
pip install Pillow==7.1.2
pip install PyYAML==5.3.1
pip install requests>=2.23.0
pip install scipy==1.4.1
pip install tqdm==4.41.0
pip install protobuf==3.20.3
# Logging -------------------------------------
pip install tensorboard==2.4.1
# Plotting ------------------------------------
pip install pandas==1.1.4
pip install seaborn==0.11.0
# Export --------------------------------------
pip install onnx==1.9.0  # ONNX export
pip install onnx-simplifier==0.3.6  # ONNX simplifier
pip install onnxruntime==1.6.0
# pip install numpy==1.18.5
# Extras --------------------------------------
pip install thop  # FLOPs computation

1.2.1 依赖导致报错问题

另外在腾讯云的平台上缺了mkl依赖，导致报错：

from torch._C import * # noqa: F403......site-packages/torch/lib/libtorch_cpu.so: undefined symbol: iJIT_NotifyEvent

补上

pip install mkl==2024.0.0

另外torch版本太高了，会报错：AttributeError: 'Upsample' object has no attribute 'recompute_scale_factor'

定位到.../site-packages/torch/nn/modules/upsampling.py” 修改成：

def forward(self, input: Tensor) -> Tensor:
    # return F.interpolate(input, self.size, self.scale_factor, self.mode, self.align_corners,
    #                      recompute_scale_factor=self.recompute_scale_factor)
    return F.interpolate(input, self.size, self.scale_factor, self.mode, self.align_corners)

报错：RuntimeError: result type Float can't be cast to the desired output type long int

修改loss.py

# anchors = self.anchors[i]
anchors, shape = self.anchors[i], p[i].shape

# 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

py38_yolov5-6.0_requirements.txt :

Package                Version
---------------------- --------------------
absl-py                2.1.0
cachetools             4.2.4
certifi                2024.8.30
charset-normalizer     3.4.0
coloredlogs            15.0.1
cycler                 0.12.1
flatbuffers            24.3.25
google-auth            1.35.0
google-auth-oauthlib   0.4.6
grpcio                 1.68.0
humanfriendly          10.0
idna                   3.10
importlib_metadata     8.5.0
intel-cmplr-lib-ur     2024.2.1
intel-openmp           2024.2.1
kiwisolver             1.4.7
Markdown               3.7
MarkupSafe             2.1.5
matplotlib             3.2.2
mkl                    2024.0.0
mpmath                 1.3.0
numpy                  1.18.5
oauthlib               3.2.2
onnx                   1.9.0
onnx-simplifier        0.3.6
onnxoptimizer          0.3.13
onnxruntime            1.6.0
opencv-python          4.1.2.30
packaging              24.2
pandas                 1.1.4
Pillow                 7.1.2
pip                    24.3.1
protobuf               5.28.3
pyasn1                 0.6.1
pyasn1_modules         0.4.1
pyparsing              3.1.4
python-dateutil        2.9.0.post0
pytz                   2024.2
PyYAML                 5.3.1
requests               2.32.3
requests-oauthlib      2.0.0
rsa                    4.9
scipy                  1.4.1
seaborn                0.11.0
setuptools             75.3.0
six                    1.16.0
sympy                  1.13.3
tbb                    2021.13.1
tensorboard            2.4.1
tensorboard-plugin-wit 1.8.1
thop                   0.1.1.post2209072238
torch                  1.10.0
torchaudio             0.10.0
torchvision            0.11.0
tqdm                   4.41.0
typing_extensions      4.12.2
urllib3                2.2.3
Werkzeug               3.0.6
wheel                  0.45.0
zipp                   3.20.2

1.3 测试

python -c "import torch; print(torch.cuda.is_available())"

python detect.py
python export.py --weights yolov5s.pt --include onnx --dynamic
python train.py