Update app.py

app.py

@@ -192,7 +192,6 @@ def inference_batch_transformers(
     ).to(device)
 
     speech_end_id = tok.convert_tokens_to_ids("<|SPEECH_GENERATION_END|>")
-    pad_id = tok.pad_token_id if tok.pad_token_id is not None else (tok.eos_token_id or speech_end_id)
 
     outputs = lm.generate(
         input_ids=input_ids_1,
@@ -203,7 +202,6 @@ def inference_batch_transformers(
         repetition_penalty=1.05,
         max_new_tokens=max_new_tokens,
         eos_token_id=speech_end_id,
-        pad_token_id=pad_id,
         num_return_sequences=batch_size, 
         use_cache=True,
     )