wandb: Currently logged in as: anony-moose-529863. Use `wandb login --relogin` to force relogin I have similar problems. I am trying to resume a wandb run on a SLURM job, by running: And I get the errors: Sign up for freeto join this conversation on GitHub. Already have an account...
The notebook was working fine till a day before and I was storing checkpoints but now when I try to run either from the checkpoint or by loading t5-small, I get asked for the wandb API key on running the trainer. I don't even have a profile. ...
I like to work with SDXL or Juggernaut XL, as the results are most natural in my opinion. SD1.5 usually lacks details because of the small size when its about depicting whole (posing) bodies. Although SDXL and Juggernaut are much better with fingers and hand poses, e.g. male genitals ...
Azure Open AI Embedding API, text-embedding-ada-002, supports 1536 dimensions. Elastic search, Lucene based engine, supports 1024 dimensions as a max. Open search can insert 16,000 dimensions as a vector storage. Open search is available to use as a vector database with Azure Open AI ...
🐛 Describe the bug I tried to implement the causal_lower_right masking in flex attention. This requires the masking function to know the difference in lengths of keys and queries: QL = query.size(2) KL = key.size(2) def causal_mask(b, h,...
= -1: # -1 rank indicates serial code dist.destroy_process_group() def get_batch(batch: Tuple[Tensor, Tensor], rank) -> Tuple[Tensor, Tensor]: x, y = batch if torch.cuda.is_available(): x, y = x.to(rank), y.to(rank) else: # I don't think this is needed... # x, ...