Update model.py

model.py

@@ -12,21 +12,7 @@ bnb_config = BitsAndBytesConfig(
     llm_int8_skip_modules=None
 )
 
-def load_model(model_path="meta-llama/Meta-Llama-3-8B-Instruct"):
-    # 不要使用 torch.device 來傳入 pipeline，改用 device_map="auto"
-    pipe = pipeline(
-        "text-generation",
-        model=model_path,
-        model_kwargs={
-            "quantization_config": bnb_config,
-            "device_map": "auto",
-            "torch_dtype": torch.float16
-        },
-        token=hf_token
-    )
-    return pipe
 
-'''
 def load_model(model_path="meta-llama/Meta-Llama-3-8B-Instruct"):
     device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
@@ -34,10 +20,10 @@ def load_model(model_path="meta-llama/Meta-Llama-3-8B-Instruct"):
         "text-generation",
         model=model_path,
         model_kwargs={"torch_dtype": torch.float16} if torch.cuda.is_available() else {},
-        quantization_config=bnb_config,
+#        quantization_config=bnb_config,
         device=device,
         token=hf_token
     )
     return pipe
-'''
+