Closed ghd98522 closed 11 months ago
The reason for this warning is that we added the parameter "find_unused_parameters=True" to Pytorch DDP. As mentioned in the warning,we have flow control in model forwarding, so in some iterations, there will be no unused parameters. I think this warning don't have any impact.
I use 6 sheets of 3090 distributed training, does this warning have any impact?![1692080679106](https://github.com/InternLandMark/LandMark/assets/141723481/ed9dffe4-a741-451c-9451-c98081ac2910)