zerogpu test

app.py

@@ -6,6 +6,7 @@ import gradio as gr
 import numpy as np
 import torch
 import torchaudio
+import spaces
 from transformers import AutoProcessor, SeamlessM4TModel
 
 from lang_list import (
@@ -39,10 +40,13 @@ DEFAULT_TARGET_LANGUAGE = "French"
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
-processor = AutoProcessor.from_pretrained("ylacombe/hf-seamless-m4t-large")
-model = SeamlessM4TModel.from_pretrained("ylacombe/hf-seamless-m4t-large").to(device)
+# processor = AutoProcessor.from_pretrained("ylacombe/hf-seamless-m4t-large")
+# model = SeamlessM4TModel.from_pretrained("ylacombe/hf-seamless-m4t-large").to(device)
+processor = AutoProcessor.from_pretrained("facebook/hf-seamless-m4t-medium")
+model = SeamlessM4TModel.from_pretrained("facebook/hf-seamless-m4t-medium").to(device)
 
 
+@spaces.GPU
 def predict(
     task_name: str,
     audio_source: str,
@@ -430,7 +434,8 @@ with gr.Blocks(css="style.css") as demo:
         outputs=[output_audio, output_text],
         api_name="run",
     )
-demo.queue(max_size=50).launch()
+demo.launch()
+# demo.queue(max_size=50).launch()
 
 # Linking models to the space
 # 'facebook/seamless-m4t-large'