Update web_demo.py

web_demo.py

@@ -167,7 +167,7 @@ def _launch_demo(model, tokenizer, audio_tokenizer):
                 model.generation_config.do_sample = False
 
 
-
+        
         add_generation_prompt =True
         input_ids = tokenizer.apply_chat_template(
             messages,
@@ -184,7 +184,7 @@ def _launch_demo(model, tokenizer, audio_tokenizer):
 
         input_ids = torch.tensor([input_ids], dtype=torch.long).to("cuda")
 
-        # print("input", tokenizer.decode(input_ids[0], skip_special_tokens=False), flush=True)
+        print("input", tokenizer.decode(input_ids[0], skip_special_tokens=False), flush=True)
 
 
         if audio_path_list == []:
@@ -328,8 +328,7 @@ def _launch_demo(model, tokenizer, audio_tokenizer):
 
 def main():
 
-    # model_name_or_path = "VITA-MLLM/VITA-Audio-Plus-Vanilla"
-    model_name_or_path = "VITA-MLLM/VITA-Audio-Plus-Boost"
+    model_name_or_path = "VITA-MLLM/VITA-Audio-Plus-Vanilla"
 
     device_map = "cuda:0"
 
@@ -385,7 +384,7 @@ def main():
     model.generation_config.top_p = 1.0
     model.generation_config.num_beams = 1
     model.generation_config.pad_token_id = tokenizer.pad_token_id
-    #model.generation_config.mtp_inference_mode = [8192,10]
+    model.generation_config.mtp_inference_mode = [8192,10]
 
 
     _launch_demo(model, tokenizer, audio_tokenizer)