Fix llama model o_proj lora_ids passing for finite lorax

quic-jouachen · quic-jouachen · commit 506fc72a93e4 · 2025-09-24T18:02:10.000-07:00
diff --git a/QEfficient/transformers/models/llama/modeling_llama.py b/QEfficient/transformers/models/llama/modeling_llama.py
@@ -174,7 +174,7 @@ def forward(
         )
 
         attn_output = attn_output.reshape(*input_shape, -1).contiguous()
-        attn_output = self.o_proj(attn_output)
+        attn_output = self.o_proj(attn_output, **kwargs)
         return attn_output, attn_weights, past_key_value
 
 

Original file line number	Diff line number	Diff line change
`@@ -174,7 +174,7 @@ def forward(`
`174`	`174`	`)`
`175`	`175`
`176`	`176`	`attn_output = attn_output.reshape(*input_shape, -1).contiguous()`
`177`		`- attn_output = self.o_proj(attn_output)`
	`177`	`+ attn_output = self.o_proj(attn_output, **kwargs)`
`178`	`178`	`return attn_output, attn_weights, past_key_value`
`179`	`179`
`180`	`180`