[Errno 95] Operation not supported

[madiltalay]

Thanks for your error report and we appreciate it a lot.

Checklist

1. I have searched related issues but cannot get the expected help.
2. The bug has not been fixed in the latest version.

Describe the bug
While running train.py, the epoch runs, but when the time to save checkpoint is reached, I get:

OSError Traceback (most recent call last)
in ()
----> 1 train_detector(model, datasets, cfg, distributed=distributed, validate=validate, timestamp=timestamp, meta=meta)
7 frames
/content/packages/mmcv/utils/path.py in symlink(src, dst, overwrite, **kwargs)
36 if os.path.lexists(dst) and overwrite:
37 os.remove(dst)
---> 38 os.symlink(src, dst, **kwargs)
39
40
OSError: [Errno 95] Operation not supported: 'epoch_1.pth' -> '/content/gdrive/My Drive/Developer/Project_1/checkpoints/resnet_retina/epoch_1.pth'

Checking the issues, I find that I have to set create_symlink=False in runner.py in mmcv.runner.
I set create_symlink=False and rerun the code. Then the epoch does not run and I get the error:

RuntimeError: cuda runtime error (209) : no kernel image is available for execution on the device at mmdet/ops/sigmoid_focal_loss/src/cuda/sigmoid_focal_loss_cuda.cu:128

I don't understand what am I doing wrong.
Reproduction

2. Did you make any modifications on the code or config? Did you understand what you have modified?
3. What dataset did you use?

Environment

1. Please run python mmdet/utils/collect_env.py to collect necessary environment infomation and paste it here.
2. You may add addition that may be helpful for locating the problem, such as
   • How you installed PyTorch [e.g., pip, conda, source]
   • Other environment variables that may be related (such as $PATH, $LD_LIBRARY_PATH, $PYTHONPATH, etc.)

Error traceback
If applicable, paste the error trackback here.

A placeholder for trackback.

Bug fix
If you have already identified the reason, you can provide the information here. If you are willing to create a PR to fix it, please also leave a comment here and that would be much appreciated!

[madiltalay]

Well it seems that I cannot use symlink twice, and as I was using it earlier in the script, the second time usage gave an error. I found a workaround to avoid the usage in the first instance which led to error-free running.
However, it is still strange that why does setting create_symlink = False leads to cuda error.
If that's a bug, it might be fixed, else the issue maybe closed.
Thanks.

[hellock]

Thanks for reporting and we will check it.

[ProBuro]

how did you solve the problem? I have the same in colab

[ProBuro]

I found a solution to the problem. there it says: replace with create_symlink = False. strange that no one has yet answered this question. / usr / ... /mmcv/runner/epoch_based_runner.py
works in colab