Add GPU decorator to generate_response function and improve login feedback

app.py

@@ -58,7 +58,6 @@ llm_model = AutoModelForCausalLM.from_pretrained(
     quantization_config=model_config,
     device_map="auto")
 
-@spaces.GPU
 def initialize_llm():
     """
     Initialize the LLM with careful memory management.
@@ -79,6 +78,7 @@ def initialize_llm():
     
     return model, tokenizer
 
+@spaces.GPU
 def generate_response(prompt:str, history: Optional[list], llm: Optional[AutoModelForCausalLM], tokenizer, max_length: int = 100) -> str:
     """
     Generate a response from the LLM model given a prompt.
@@ -111,8 +111,8 @@ demo =  gr.ChatInterface(
 
 if __name__ == "__main__":
     auth = HuggingFaceLogin()
-    auth.login()
+    if auth.login():
+        print("Login successful!")
     
-    # Initialize the model and tokenizer
     llm_model, llm_tokenizer = initialize_llm()
     demo.launch()