Update app.py

app.py

@@ -22,8 +22,8 @@ base_model = AutoModelForCausalLM.from_pretrained(
     'meta-llama/Llama-2-7b-chat-hf',
     token=access_token,
     trust_remote_code=True, 
-    device_map="cpu",
-    torch_dtype=torch.float16,
+    device_map=None,
+    torch_dtype="auto",
     offload_folder="offload/"
 )
 model = PeftModel.from_pretrained(