squall/picotron

Go to file

ferdinand.mom f1053e3cbe revert to use huggingface cli + hf_transfers (this will not create snapshots/blob folder etc through CLI use)		2024-12-18 15:51:04 +00:00
picotron	Revert to @zzhhjjj class naming as it is more expressive	2024-12-17 15:55:18 +00:00
template	breaking: add new version of initi meta device but memory leaks	2024-12-17 15:46:16 +00:00
tests	Revert to @zzhhjjj class naming as it is more expressive	2024-12-17 15:55:18 +00:00
.gitignore	small changes	2024-12-17 05:01:35 +00:00
create_config.py	revert to use huggingface cli + hf_transfers (this will not create snapshots/blob folder etc through CLI use)	2024-12-18 15:51:04 +00:00
extract_metrics.py	add mfu parsing	2024-12-04 13:08:28 +00:00
README.md	Initial commit	2024-09-18 14:01:22 +02:00
requirements.txt	use hf_transfer which improve download time by 3	2024-12-18 14:51:14 +00:00
setup.py	tesnsor parallel, will clean later	2024-10-18 05:13:44 +00:00
submit_slurm_jobs.py	can now load big model through safetensors (sharded and single file)	2024-12-01 19:39:16 +00:00
train.py	download safetensors when creating config time. If we do it in training, barrier() may tiemout while waiting for download	2024-12-17 15:46:16 +00:00

README.md

picotron