local inference page with fixes to gpu with zerogpu + add accelerate for device mapping - removed previous and fixed overall

app.py

@@ -341,4 +341,4 @@ with gr.Blocks() as demo:
     )
 
 logger.info("Starting Gradio interface...")
-demo.queue().launch(ssr_mode=False) 
+demo.queue().launch() 

requirements.txt

@@ -1,5 +1,5 @@
 transformers
-gradio
+gradio>=4.44.1
 torch
 accelerate
 SentencePiece