为 pytorch 安装预训练模型时出错

Question

我正在使用 Windows 10 机器（是的，我知道，别笑，）。 并使用 python 3,7: 我正在尝试在此处安装预训练模型：

https://github.com/meliketoy/fine-tuning.pytorch

该网站建议的命令是：

$ git clone https://github.com/Cadene/pretrained-models.pytorch.git
$ pretrained-models.pytorch
$ python setup.py install

虽然网站上说这是针对 Python 3.5，而我有 3.7，但我认为 3.7 版本应该是向后兼容的，对吧？

我成功运行了git clone ，而pretrained-models.pytorch实际上是一个cd命令（这让我陷入了一个循环。）。 然后我遇到了python setup.py install的麻烦

我得到的错误是：

[Errno 2] No such file or directory: 'build\\bdist.win-amd64\\egg\\pretrainedmodels\\models\\resnext_features\\__pycache__\\resnext101_32x4d_features.cpython-37.pyc.1702181039952'

我该如何解决这个错误？

编辑（回应评论）：有人要求完整的追溯。 这里是！

(base) G:\>python setup.py install
running install
running bdist_egg
running egg_info
creating pretrainedmodels.egg-info
writing pretrainedmodels.egg-info\PKG-INFO
writing dependency_links to pretrainedmodels.egg-info\dependency_links.txt
writing requirements to pretrainedmodels.egg-info\requires.txt
writing top-level names to pretrainedmodels.egg-info\top_level.txt
writing manifest file 'pretrainedmodels.egg-info\SOURCES.txt'
reading manifest file 'pretrainedmodels.egg-info\SOURCES.txt'
writing manifest file 'pretrainedmodels.egg-info\SOURCES.txt'
installing library code to build\bdist.win-amd64\egg
running install_lib
running build_py
creating build
creating build\lib
creating build\lib\pretrainedmodels
copying pretrainedmodels\utils.py -> build\lib\pretrainedmodels
copying pretrainedmodels\version.py -> build\lib\pretrainedmodels
copying pretrainedmodels\__init__.py -> build\lib\pretrainedmodels
creating build\lib\pretrainedmodels\datasets
copying pretrainedmodels\datasets\utils.py -> build\lib\pretrainedmodels\datasets
copying pretrainedmodels\datasets\voc.py -> build\lib\pretrainedmodels\datasets
copying pretrainedmodels\datasets\__init__.py -> build\lib\pretrainedmodels\datasets
creating build\lib\pretrainedmodels\models
copying pretrainedmodels\models\bninception.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\cafferesnet.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\dpn.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\fbresnet.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\inceptionresnetv2.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\inceptionv4.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\nasnet.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\nasnet_mobile.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\pnasnet.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\polynet.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\resnext.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\senet.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\torchvision_models.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\utils.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\vggm.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\wideresnet.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\xception.py -> build\lib\pretrainedmodels\models
copying pretrainedmodels\models\__init__.py -> build\lib\pretrainedmodels\models
creating build\lib\pretrainedmodels\models\resnext_features
copying pretrainedmodels\models\resnext_features\resnext101_32x4d_features.py -> build\lib\pretrainedmodels\models\resnext_features
copying pretrainedmodels\models\resnext_features\resnext101_64x4d_features.py -> build\lib\pretrainedmodels\models\resnext_features
copying pretrainedmodels\models\resnext_features\__init__.py -> build\lib\pretrainedmodels\models\resnext_features
creating build\bdist.win-amd64
creating build\bdist.win-amd64\egg
creating build\bdist.win-amd64\egg\pretrainedmodels
creating build\bdist.win-amd64\egg\pretrainedmodels\datasets
copying build\lib\pretrainedmodels\datasets\utils.py -> build\bdist.win-amd64\egg\pretrainedmodels\datasets
copying build\lib\pretrainedmodels\datasets\voc.py -> build\bdist.win-amd64\egg\pretrainedmodels\datasets
copying build\lib\pretrainedmodels\datasets\__init__.py -> build\bdist.win-amd64\egg\pretrainedmodels\datasets
creating build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\bninception.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\cafferesnet.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\dpn.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\fbresnet.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\inceptionresnetv2.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\inceptionv4.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\nasnet.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\nasnet_mobile.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\pnasnet.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\polynet.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\resnext.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
creating build\bdist.win-amd64\egg\pretrainedmodels\models\resnext_features
copying build\lib\pretrainedmodels\models\resnext_features\resnext101_32x4d_features.py -> build\bdist.win-amd64\egg\pretrainedmodels\models\resnext_features
copying build\lib\pretrainedmodels\models\resnext_features\resnext101_64x4d_features.py -> build\bdist.win-amd64\egg\pretrainedmodels\models\resnext_features
copying build\lib\pretrainedmodels\models\resnext_features\__init__.py -> build\bdist.win-amd64\egg\pretrainedmodels\models\resnext_features
copying build\lib\pretrainedmodels\models\senet.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\torchvision_models.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\utils.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\vggm.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\wideresnet.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\xception.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\models\__init__.py -> build\bdist.win-amd64\egg\pretrainedmodels\models
copying build\lib\pretrainedmodels\utils.py -> build\bdist.win-amd64\egg\pretrainedmodels
copying build\lib\pretrainedmodels\version.py -> build\bdist.win-amd64\egg\pretrainedmodels
copying build\lib\pretrainedmodels\__init__.py -> build\bdist.win-amd64\egg\pretrainedmodels
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\datasets\utils.py to utils.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\datasets\voc.py to voc.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\datasets\__init__.py to __init__.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\models\bninception.py to bninception.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\models\cafferesnet.py to cafferesnet.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\models\dpn.py to dpn.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\models\fbresnet.py to fbresnet.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\models\inceptionresnetv2.py to inceptionresnetv2.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\models\inceptionv4.py to inceptionv4.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\models\nasnet.py to nasnet.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\models\nasnet_mobile.py to nasnet_mobile.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\models\pnasnet.py to pnasnet.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\models\polynet.py to polynet.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\models\resnext.py to resnext.cpython-37.pyc
byte-compiling build\bdist.win-amd64\egg\pretrainedmodels\models\resnext_features\resnext101_32x4d_features.py to resnext101_32x4d_features.cpython-37.pyc
error: [Errno 2] No such file or directory: 'build\\bdist.win-amd64\\egg\\pretrainedmodels\\models\\resnext_features\\__pycache__\\resnext101_32x4d_features.cpython-37.pyc.1702181039952'

Answer 1

一种选择是使用 docker 图像，我经常使用的是来自datascience-notebook图像。

为了这：

• 1. 为 Windows 安装 docker 桌面，参考此链接。
• 2. 在 Docker Destop Settings 中启用文件共享

正如您从C:users\amtre中看到的那样，我可以将任何子目录挂载到容器中，例如Documents文件夹中的所有子目录。

• 3. 一旦 docker 可以访问容器的挂载目录，我们将使用 jupyter datascience-notebook ，因为它已经默认附带了一些软件包。 在终端上输入

docker run -it -e GRANT_SUDO=yes --user root --rm -p 8888:8888 -p 4040:4040 -v C:/users/amtre/Documents:/home/jovyan/work jupyter/datascience-notebook

拉动 docker 映像需要一段时间，但最后您将获得 URL 来访问笔记本，如上图所示。

4. 在 Jupyter 中，打开一个终端并输入

git clone https://github.com/Cadene/pretrained-models.pytorch.git
cd pretrained-models.pytorch
python setup.py install

这也将安装'torch', 'torchvision', 'munch', 'tqdm' ，因为它在setup.py的install_requires中。 安装完成后，您应该可以开始使用预训练模型了