Pedro Cuenca
Pedro Cuenca
Thanks a lot @yiyixuxu, checking it out!
Hi @yiyixuxu! ~~What hardware are you using to test? It works fine for me on a single A100, and OOMs in a TPUv3-8. Are you using multiple GPUs?~~ I was...
Yeah, I realized later that the default `sample_batch_size` for class images generation is `4`.
Let's please try to keep the discussion civil. This is clearly a sensitive topic and we could all benefit from different points of view as long as arguments are presented...
Works fine for me on 3090 (`cuda:0`) and 2080 Ti (`cuda:1`). I don't currently have any 4090s to test.
@patrickvonplaten I agree, I think it's a bit too specific. The training scripts are in my opinion meant to be examples, so people are free to extend and modify them...
Hi @zihao12! I've been trying to reproduce this issue but the settings you posted work fine for me. Would you mind providing a little more information so we can try...
Thanks for your response! Your setup is very similar to mine, except that I have slightly newer versions of PyTorch (1.13 instead of 1.12.1) and accelerate (0.15.0 instead of 0.13.2)....
Thanks @tzvc! I'll test on an A100 tomorrow.
Sorry for the delay in getting back to you! I could finally replicate this issue using A100 (40 GB) instances from Google Cloud, but not elsewhere. Specifically, I've found the...