Set higher ZeroGPU limit for visitors.

README.md

@@ -29,5 +29,6 @@ I wanted my prototype to have 0 setup for the user if at all possible, and HF Sp
 2. Don't try to run an LLM locally on unknown hardware specs.
 3. Don't use my key for a publicly-facing app.
 
-I already have up to 25 minutes of GPU inference per day with my Pro account, so there's no extra expense for me, and the wait for a GPU is minimal.  There's a small risk of a malicious stranger using up all 25 minutes before you guys get to it. 
+I've set the limit for visitors to 300s.  That does mean that the wait for a GPU could be longer, but you'll be able to run more than 1 or 2 trials.
+
 

app.py

@@ -120,4 +120,4 @@ with gr.Blocks() as demo:
         outputs=result_out,
     )
 
-demo.launch(share=True)
+demo.launch()

quiz_generator.py

@@ -43,11 +43,12 @@ model = AutoModelForCausalLM.from_pretrained(
 
 pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
 
-@spaces.GPU
+@spaces.GPU(duration=300)
 def run_inference(prompt_message: str):
     """
     @spaces.GPU is a Hugging Face decorator for GPU inference.
     Required for the ZeroGPU setting in HF Spaces.
+    duration=300 allows visitors to use up to 300s of inference.
     See https://huggingface.co/docs/hub/en/spaces-zerogpu
 
     :param prompt_message: The user message submitted to the LLM