Flux controlnet can't be trained, do this script really work?

[liuyu19970607]

Describe the bug

run with one num processes, the code broke down and returns:
RuntimeError: Expected to have finished reduction in the prior iteration before starting a new one. This error indicates that your module has parameters that were not used in producing loss. You can enable unused parameter detection by passing the keyword argument find_unused_parameters=True to torch.nn.parallel.DistributedDataParallel, and by

run with more than one processes, the code broke down and returns:
Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data.

Reproduction

just follow the instructions and it will be reproduced

Logs

No response

System Info

diffusers v0.32

Who can help?

No response

[sayakpaul]

Does it run suitably within a single GPU? Cc: @PromeAIpro

[PromeAIpro]

yesterday just tested on release 0.31.0, single a100 80G, works fine. see #9857
just tested good a few minute ago in 0.32.0dev0, could you provide your python-installations and commad line script for me to further debug @liuyu19970607

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[sayakpaul]

Closing due to inactivities.