awq fix

qa_engine/qa_engine.py

@@ -66,7 +66,7 @@ class TransformersPipelineModel(LLM):
         tokenizer = AutoTokenizer.from_pretrained(model_id)
         if "AWQ" in model_id:
             model = AutoAWQForCausalLM.from_quantized(model_id, fuse_layers=True,
-                                          trust_remote_code=False, safetensors=True, torch_dtype=torch.bfloat16).model
+                                          trust_remote_code=False, safetensors=True).model
         else:
             model = AutoModelForCausalLM.from_pretrained(
                 model_id,