use gpu

app.py

@@ -43,8 +43,9 @@ def load_model():
     access_token = os.getenv("HF_TOKEN")
     try:
         device = 0 if torch.cuda.is_available() else -1
-        tokenizer = AutoTokenizer.from_pretrained(model_name, use_auth_token=access_token, clean_up_tokenization_spaces=False)
-        model = AutoModelForCausalLM.from_pretrained(model_name, use_auth_token=access_token)
+        tokenizer = AutoTokenizer.from_pretrained(model_name, token=access_token, clean_up_tokenization_spaces=False)
+        model = AutoModelForCausalLM.from_pretrained(model_name, token=access_token)
+        device = 0 if torch.cuda.is_available() else -1
         return pipeline("text-generation", model=model, tokenizer=tokenizer, max_new_tokens=150, temperature=0.1)
     except Exception as e:
         print(f"Error loading model: {e}")