4 Commits

Author	SHA1	Message	Date
Wing Lian	3ebf22464b	qlora-fsdp ram efficient loading with hf trainer (#1791) ... * fix 405b with lower cpu ram requirements * make sure to use doouble quant and only skip output embeddings * set model attributes * more fixes for sharded fsdp loading * update the base model in example to use pre-quantized nf4-bf16 weights * upstream fixes for qlora+fsdp	2024-07-30 19:21:38 -04:00
Wing Lian	891ae8aa13	fix ray install (#1630)	2024-05-16 01:25:42 -04:00
Wing Lian	0c49ecc429	more fixes to work with runpod + skypilot (#1629)	2024-05-16 00:05:56 -04:00
Wing Lian	60113437e4	cloud image w/o tmux (#1628)	2024-05-15 22:27:40 -04:00