Update app.py

app.py

@@ -71,9 +71,9 @@ def generate_text(
         temp += out["choices"][0]["text"]
         yield temp
 
-# Define the Gradio interface without timeout
+# Define the Gradio interface without retry_btn
 demo = gr.Interface(
-   fn=generate_text,
+    fn=generate_text,
     inputs="text",
     outputs="text",
     title="llama-cpp-python on GPU with ChromaDB",
@@ -91,7 +91,6 @@ demo = gr.Interface(
         gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
         gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
     ],
-    processing_timeout=None,  # No timeout limit
 )
 
 if __name__ == "__main__":