use two method to add bias

4caf1539 · zhuwenwen · 767bb050 · 4caf1539
Commit 4caf1539 authored Jul 20, 2024 by zhuwenwen
Hide whitespace changes
Inline Side-by-side

Showing with 4 additions and 1 deletion

vllm/model_executor/layers/linear.py vllm/model_executor/layers/linear.py +4 -1

No files found.
--- a/vllm/model_executor/layers/linear.py
+++ b/vllm/model_executor/layers/linear.py
@@ -115,7 +115,10 @@ class UnquantizedLinearMethod(LinearMethodBase):
        if self.use_llama_nn:
            if bias is not None:
-                return torch.addmm(bias, x, weight)
+                if len(x.shape) == 2: 
+                    return torch.addmm(bias, x, weight)
+                else:
+                    return torch.matmul(x, weight) + bias
            else:
                return torch.matmul(x, weight)
        else: