Update app.py

app.py

@@ -9,7 +9,7 @@ MODEL_PATH = "Ozaii/zephyr-bae"
 max_seq_length = 2048
 
 print("Attempting to load Zephyr... Cross your fingers! 🤞")
-
+@spaces.GPU
 try:
     peft_config = PeftConfig.from_pretrained(MODEL_PATH)
     base_model_name = peft_config.base_model_name_or_path
@@ -35,6 +35,7 @@ except Exception as e:
     print(f"Oops! Zephyr seems to be playing hide and seek. Error: {str(e)}")
     raise
 
+@spaces.GPU
 def generate_response(prompt, max_new_tokens=128):
     inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=2048).to(model.device)
     streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)