4 Commits (79e36e17e898e35c00a0c128a413b51d4ccbd536)

Author	SHA1	Message	Date
wanghao	79e36e17e8	Fix: Remove attn_mask when using is_causal=True in Flash Attention ... PyTorch's scaled_dot_product_attention cannot use attn_mask and is_causal=True simultaneously. This commit sets attn_mask=None when is_causal=True to avoid parameter conflicts.	2025-10-18 21:02:46 +08:00
jingyaogong	4e35fb9da8	[fix] update model	2025-10-17 00:09:32 +08:00
jingyaogong	29454c31af	fix bugs	2025-04-27 09:56:49 +08:00
jingyaogong	a62faf34bd	250426	2025-04-26 10:05:47 +08:00