Pedro Cuenca
Pedro Cuenca
Closing as this appears to be some sort of problem with GCP images.
> Hey @BenjaminIrwin, > > This is actually quite easily doable. You just need to pass a config parameter that will change the size of your input channels to the...
> Thanks a lot @SkyTNT for being so responsive here and replying so quickly, that's super nice! > For the future we should probably change community pipelines to automatically detect...
So far I've not been able to observe this, will keep watching for a few days.
In the second part of [this comment](https://github.com/pytorch/pytorch/issues/84039#issuecomment-1241973904) I documented a similar behaviour using `einsum` vs `matmul`. I used @patil-suraj's branch to run the tests without the warmup pass, but they...
> Is this still relevant Yes, unfortunately, still an issue with the RC of PyTorch 1.13 from the `test` wheel. @Birch-san In my experience this is only required the first...
Thanks @Birch-san! Great work posting all those details in PyTorch's github, it looks like they are really helping! I'll do some more testing about the non-determinism to try to isolate...
I'll ask in the PyTorch forums again.
This doesn't fully work for me yet. The way I'm testing is by commenting these lines: https://github.com/huggingface/diffusers/blob/97958bcdc808649d13b0fa3b0f7fad686caf3866/tests/test_modeling_common.py#L53-L55 and then running the tests for `test_from_save_pretrained` on `UNet2DModelTests`. If the lines are...
Update: the bot was right in this case. This was never resolved in PyTorch 1.13, but works in PyTorch 2.