RuntimeError: max_pool3d_with_indices_backward_cuda does not have a deterministic implementation, but you set 'torch.use_deterministic_algorithms(True)'.

[WtElio]

I found this problem. Can you help me look at it?

[mkang315]

Did you configure your GPU Device 0?

[Xiaoliang-Liu666]

I also have the same problem, and how to solve this problem?

[mkang315]

I suppose your problem is related to hyperparameters that are set in ./data/hyps/hyp.scratch-low.yaml because the error message can be tracked in ./train.py
line 69: def train(hyp, opt, device, callbacks): # hyp is path/to/hyp.yaml or hyp dictionary
line 479: def main(opt, callbacks=Callbacks()):

To test your hardware and software environment, you may run YOLOv5. Thank you!

[mkang315]

Please look at YOLOv5 Multi-GPU Tutorial and follow it.