up

run_demo_multi_models.py

@@ -60,7 +60,7 @@ def maybe_load_cached_pipeline(model_name):
         # set forced_decoder_ids
         pipe.model.config.forced_decoder_ids = pipe.tokenizer.get_decoder_prompt_ids(language="fr", task="transcribe")
         # limit genneration max length
-        pipe.model.config.max_length = MAX_NEW_TOKENS + 1
+        # pipe.model.config.max_length = MAX_NEW_TOKENS + 1
 
         logger.info(f"`{model_name}` pipeline has been initialized")
         print_memory_info()