Partially freeze SWIN backbone

[levan92]

Hi, can I check how we can partially freeze a SWIN backbone? I've tried adding frozen_stages=3 to the backbone in the config, but met with the following error:

RuntimeError: Expected to have finished reduction in the prior iteration before starting a new one. This error indicate
s 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
making sure all `forward` function outputs participate in calculating loss.
If you already have done the above, then the distributed data parallel module wasn't able to locate the output tensors
in the return value of your module's `forward` function. Please include the loss function and the structure of the return value of `forward` of your module when reporting this issue (e.g. list, dict, iterable).
Parameter indices which did not receive grad for rank 0: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
24 25 26 27 28 29 30 31 32 325 326

Is there something else that needs to be set? Thank you!

[levan92]

@BIGWangYuDong any updates on this?

[levan92]

any updates? @BIGWangYuDong

[austinmw]

I get the same error when using frozen_stages with CSPNeXt .

I believe this can be fixed by adding self._freeze_stages() at the end of the backbone __init__ methods, similar to the resnet code (here for swin and here for CSPNeXt).

[gachiemchiep]

@austinmw I got the same error. Your code fix it. Thank you very much