MartinPernus

Results 4 comments of MartinPernus

RuntimeError: Function 'SqrtBackward' returned nan values in its 0th output.

Having the same problem using alexnet. Solved by adding epsilon under the sqrt.

re-trained model cannot get the IS score as the thesis provides

Same problem here, following the instructions, the final inception score on CUB is only 4.00 EDIT: By evaluating many trained models, one of them achieved IS of 4.26. I would...

Command '['ninja', '-v']' returned non-zero exit status 1.

Hi, it seems that cuda is not properly installed. Can you try reinstalling? For other compilation issues I have also found [this dockerfile](https://github.com/rosinality/alias-free-gan-pytorch/blob/main/Dockerfile) to be working well.

[Bug] Attention to qkv should use QKVParallelLinear

@philippe-eecs the paper says they do not use TP (too much comm overhead), just FSDP + unified sequence parallel