Xiang Jiang

it works after a reboot.

it stopped working again. seems to be quite random.

Hello, Our implicit alignment approach focuses on organizing data in the input space, especially with class imbalance, and can be combined with any domain adaptation algorithm. If domain adaptation can...

it is still available. you should use `lfs` to download the code.

I'm wondering what dataset are you referring to? I didn't notice this issue with `num_workers` but I reduced the pseudo-label update frequency to speed up the process. If the pseudolabels...

Thanks for the explanation. I'm wondering do you experience slow loading when the source has a single domain, and are you observing 50% slower loading due to one additional domain...

Office31 is a small dataset and should not take more than a couple hours to train. Maybe there are some issues with the training configurations.

@suchenzang @stephenroller Could you help elaborate how `convert_to_singleton.py` helps with this case? It appears to be merging `reshard*.pt` files, but the OPT-IML checkpoint does not have `reshard*.pt`. We are encountering...

> > @Charles-ux-bit Assuming 30B-IML checkpoint is same as 30B checkpoint in architecture, a hack that works is to copy "shard_metadata" from the later and then use reshard_fsdp script The...

@sriniiyer , could you fix the issue in the checkpoint or upload resharded version to unblock us from running inference on OPT-IML?