Fix incorrect use of bias in awq

server/text_generation_server/utils/awq/quantize/qmodule.py

@@ -35,10 +35,7 @@ class WQLinear(nn.Module):
         self.qweight = qweight
         self.qzeros = qzeros
         self.scales = scales
-        if bias:
-            self.bias = bias
-        else:
-            self.bias = None
+        self.bias = bias
 
     @torch.no_grad()
     def forward(self, x):

server/text_generation_server/utils/layers.py

@@ -335,7 +335,7 @@ def get_linear(weight, bias, quantize):
             qweight=qweight,
             qzeros=qzeros,
             scales=scales,
-            bias=bias is not None,
+            bias=bias,
         )
     else:
         raise NotImplementedError(f"Quantization `{quantize}` is not implemented yet.")