diff --git a/train_v4.py b/train_v4.py index 9a52496..c94cb0c 100644 --- a/train_v4.py +++ b/train_v4.py @@ -151,7 +151,7 @@ trainer = SFTTrainer( save_steps=100, save_total_limit=2, warmup_ratio=WARMUP_RATIO, - optim="adamw_8bit", + optim="adamw_torch", seed=42, report_to="none", max_seq_length=MAX_SEQ,