Update app.py

app.py

@@ -20,7 +20,7 @@ async def load_models():
         tokenizers[model_key] = AutoTokenizer.from_pretrained(model_name)
         models[model_key] = AutoModelForCausalLM.from_pretrained(
             model_name,
-            torch_dtype=torch.float16,
+            torch_dtype="auto",
             device_map="auto"
         )
     print("All models loaded!")