No print

unsloth/kernels/utils.py

@@ -180,7 +180,6 @@ pass
 def fast_linear_forward(proj, X, temp_lora = None, out = None):
     W, W_quant, lora_A, lora_B, lora_S = get_lora_parameters(proj)
     out = fast_gemv(X, W, W_quant, out = out)
-    print(out.shape, X.shape)
     if lora_A is not None:
         dtype = X.dtype
         temp_lora = torch.matmul(X, lora_A.to(dtype).t(), out = temp_lora)

unsloth/models/llama.py

@@ -153,7 +153,7 @@ def LlamaAttention_fast_forward_inference(
 
     # A = self.o_proj(A)
     A = fast_linear_forward(self.o_proj, A)
-    A = On.reshape(1, 1, self.hidden_size)
+    A = A.reshape(1, 1, self.hidden_size)
 
     return A, (Kn, Vn)
 pass