diff --git a/applications/ColossalChat/coati/distributed/consumer.py b/applications/ColossalChat/coati/distributed/consumer.py
index fa0e331e8..45840e7e4 100644
--- a/applications/ColossalChat/coati/distributed/consumer.py
+++ b/applications/ColossalChat/coati/distributed/consumer.py
@@ -165,7 +165,6 @@ class BaseConsumer:
                     state_dict, src=self.num_producers, device=self.device, group_name="sync_model"
                 )
         del state_dict
-        print(f"[C{self.rank}]: Sync model before training done")
         torch.cuda.empty_cache()
         self.profiler.exit("sync_model")
 
diff --git a/applications/ColossalChat/coati/distributed/inference_backend.py b/applications/ColossalChat/coati/distributed/inference_backend.py
index e6eab1669..9f9d8e36b 100644
--- a/applications/ColossalChat/coati/distributed/inference_backend.py
+++ b/applications/ColossalChat/coati/distributed/inference_backend.py
@@ -381,7 +381,6 @@ class AsyncVLLMInferenceBackend(AsyncInferenceBackend):
         log_probs = []
         response_idx = []
         while len(self.running_requests) >= self.microbatch_size:
-            # print(f"Current running {len(self.running_requests)}/{self.microbatch_size} requests, waiting...")
             await asyncio.sleep(0.1)
         self.running_requests.append(request_id)  # enqueue
         # pop the first input_ids and attention_mask