Update app.py

app.py

@@ -3,8 +3,7 @@ os.system("pip install gradio==3.3")
 import gradio as gr
 import numpy as np
 import streamlit as st
-
-from fairseq.checkpoint_utils import load_model_ensemble_and_task_from_hf_hub
+from audio_pipe import SpeechToSpeechPipeline
 
 title = "SpeechMatrix Speech-to-speech Translation"
 
@@ -21,18 +20,14 @@ for src in SRC_LIST:
             MODEL_LIST.append(f"textless_sm_{src}_{tgt}")
         
 examples = []
-MODEL_LIST = ['textless_sm_es_en', 'textless_sm_es_fr', 'textless_sm_et_en']
-print(MODEL_LIST)
 
 # io_dict = {model: gr.Interface.load(f"huggingface/facebook/{model}", api_key=st.secrets["api_key"]) for model in MODEL_LIST}
-# io_dict = {model: gr.Interface.load(f"huggingface/facebook/{model}") for model in MODEL_LIST}
-io_dict = {model: load_model_ensemble_and_task_from_hf_hub(f"facebook/{model}")[0] for model in MODEL_LIST}
+pipe_dict = {model: SpeechToSpeechPipeline(f"facebook/{model}") for model in MODEL_LIST}
    
 def inference(audio, model):
-    print(io_dict)
-    print(io_dict[model])
-    out_audio = io_dict[model](audio)   
-    return out_audio 
+    out_audio = pipe_dict[model](audio).get_config()["value"]["name"]
+    return out_audio
+    
 gr.Interface(
     inference,
     [gr.inputs.Audio(source="microphone", type="filepath", label="Input"),gr.inputs.Dropdown(choices=MODEL_LIST, default="xm_transformer_sm_all-en",type="value", label="Model")