some minor fix (#3048)

Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

server/text_generation_server/layers/gptq/__init__.py

@@ -171,6 +171,7 @@ class GPTQWeightsLoader(WeightsLoader):
             g_idx=g_idx,
             bits=self.bits,
             groupsize=self.groupsize,
+            use_awq_kernel=self.quantize == "awq",
             use_exllama=use_exllama,
         )
 

server/text_generation_server/layers/moe/unquantized.py

@@ -85,6 +85,8 @@ class UnquantizedSparseMoELayer(nn.Module):
                 use_grouped_topk=self.n_expert_group is not None,
                 num_expert_group=self.n_expert_group,
                 topk_group=self.topk_group,
+                scoring_func=self.scoring_func,
+                e_score_correction_bias=self.e_score_correction_bias,
             )
         return fused_moe(
             x,