Pass quant_config to attention gate projection (#28649)

2026-06-18 20:04:25 +08:00
parent bb9d31f22d
commit 97e3b8998d
1 changed files with 1 additions and 1 deletions
@@ -304,7 +304,7 @@ class LagunaAttention(nn.Module):
                g_proj_dim,
                bias=False,
                gather_output=False,
-                quant_config=None,
+                quant_config=quant_config,
                tp_rank=attn_tp_rank,
                tp_size=attn_tp_size,
                prefix=add_prefix("g_proj", prefix),