Devansh Bisla

Only setting seed does not guarantee deterministic behaviour on GPU. Have you tried using ```torch.use_deterministic_algorithms``` or ```torch.backends.cudnn.deterministic```. Here is an article about it just in case you haven't seen it...

Hi @noahgolmant, I would like to work on this issue. Do you have some starting points? Thanks

You could replace the function alltogether def apply_ckpt(self, ckpt: Union[None, str, dict]): if ckpt is None: return if ckpt.endswith("ckpt"): sd = torch.load(ckpt, map_location="cpu")["state_dict"] elif ckpt.endswith("safetensors"): sd = load_safetensors(ckpt) else: raise...