Update app.py

app.py

@@ -122,6 +122,7 @@ def generate_audio(text, speaker_id=None):
     inputs = tokenizer(text, return_tensors="pt")#.input_ids
 
     speaker_embeddings = None
+    model=get_model(name_model)
     #torch.cuda.empty_cache()
     with torch.no_grad():
         for chunk in _inference_forward_stream(model,input_ids=inputs.input_ids,attention_mask=inputs.attention_mask,speaker_embeddings= speaker_embeddings,chunk_size=64):