maybe_compute_generative_loss error

[azuredsky]

superclip/training/train.py", line 432, in maybe_compute_generative_loss
return F.cross_entropy(token_logits.permute(0, 2, 1), token_labels)
RuntimeError: permute(sparse_coo): number of dimensions in the tensor input does not match the length of the desired ordering of dimensions i.e. input.dim() = 2 is not equal to len(dims) = 3
run===== torch.Size([32, 49408]) torch.Size([32, 77]) torch.Size([32, 512])run=====

[Bernice123]

same error...

[SleepingWhz]

Thanks for your interest and for reporting the issue.

Using our provided training config, we don’t seem to encounter this problem.
Could you share more details about your setup (e.g., the config and launch command) so we can try to reproduce it?