Yolov5: can't use cuda

[liguoyu666]

I create a conda env as your requirements:
cuda: 10.1
python 3.8.0
pytorch 1.6.0
torchvision: 0.7.0
but I can't use the cuda.
import torch print(torch.cuda.is_available()) # False

[github-actions[bot]]

Hello @liguoyu666, thank you for your interest in our work! Please visit our Custom Training Tutorial to get started, and see our Jupyter Notebook , Docker Image, and Google Cloud Quickstart Guide for example environments.

If this is a bug report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom model or data training question, please note Ultralytics does not provide free personal support. As a leader in vision ML and AI, we do offer professional consulting, from simple expert advice up to delivery of fully customized, end-to-end production solutions for our clients, such as:

• Cloud-based AI systems operating on hundreds of HD video streams in realtime.
• Edge AI integrated into custom iOS and Android apps for realtime 30 FPS video inference.
• Custom data training, hyperparameter evolution, and model exportation to any destination.

For more information please visit https://www.ultralytics.com.

[liguoyu666]

My previous cuda version was 10.0，today I install cuda 10.1, conda environments are OK such as yolov4, except yolov5.

docker pull ultralytics/yolov5, In the docker environment，the cuda can't be used either.

[learncrazy]

You can check whether the configuration path of cuda is correct? I remember that you need to add the path of the new version of cuda10.1 to .bashrc (in fact, you only need to modify the original cuda10.0 to 10.1).

[liguoyu666]

You can check whether the configuration path of cuda is correct? I remember that you need to add the path of the new version of cuda10.1 to .bashrc (in fact, you only need to modify the original cuda10.0 to 10.1).

In fact, I installed cuda 10.0 and cuda10.1, and I modify the original cuda 10.0 to 10.1.
yolov3 and yolov4 can use cuda in their environment.
This seems to be caused by a mismatch between the cuda version and the pytorch version.

[liguoyu666]

In the yolov5 environment, I change pytorch 1.6.0 to 1.4.0, torch.cuda.is_available() is True.
I uninstall cuda 10.0 and 10.1, then reinstall cuda 10.1, the same problem remains: in pytorch 1.6.0, torch.cuda.is_available() is False; in pytorch 1.2.0 and 1.4.0, torch.cuda.is_available() is True.
My driver version is 430.40.

[learncrazy]

You can try it and download torch again. Use the command pip install torch==1.6.0+cu101, which may be useful.

[liguoyu666]

You can try it and download torch again. Use the command pip install torch==1.6.0+cu101, which may be useful.

Because I changed the source of pip, using pip install torch==1.6.0+cu101 can't install, so I installed it according to the official Python installation:
pip install torch==1.6.0+cu101 torchvision==0.7.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html,
the cuda can be cused. Thank you!