Update app.py

app.py

@@ -19,7 +19,7 @@ def main():
     # Load the full, merged model (quantization_config has been removed)
     model = AutoModelForCausalLM.from_pretrained(
         MODEL_ID,
-        torch_dtype=torch.float32, # 在CPU上使用 torch.float32 更稳定
+        torch_dtype=torch.float16,
         device_map="auto",
         trust_remote_code=True
     )