diff --git a/README.md b/README.md
index 7a1bc8e8b..332aaa0af 100644
--- a/README.md
+++ b/README.md
@@ -9,9 +9,9 @@
 SGLang is a structured generation language designed for large language models (LLMs).
 It makes your interaction with LLMs faster and more controllable by co-designing the frontend language and the runtime system.
 
-The core features of SGLang include:
+The core features include:
 - **A Flexible Front-End Language**: This allows for easy programming of LLM applications with multiple chained generation calls, advanced prompting techniques, control flow, multiple modalities, parallelism, and external interaction.
-- **A High-Performance Runtime with RadixAttention**: This feature significantly accelerates the execution of complex LLM programs by automatic KV cache reuse across multiple calls. It also supports other common techniques like continuous batching and tensor parallelism.
+- **A High-Performance Runtime with RadixAttention**: This feature significantly accelerates the execution of complex LLM programs by automatically reusing the KV cache across multiple calls. It can also be used as a standalone serving engine with all common techniques implemented, such as continuous batching and tensor parallelism.
 
 ## News
 - [2024/02] 🔥 SGLang enables **3x faster JSON decoding** with compressed finite state machine ([blog](https://lmsys.org/blog/2024-02-05-compressed-fsm/)).
diff --git a/docs/hyperparameter_tuning.md b/docs/hyperparameter_tuning.md
index 0263d557a..dec516bc9 100644
--- a/docs/hyperparameter_tuning.md
+++ b/docs/hyperparameter_tuning.md
@@ -6,7 +6,7 @@ Achieving a large batch size is the most important thing for attaining high thro
 
 When the server is running at full load, look for the following in the log:
 
-```[gpu_id=0] #running-req: 233, #token: 370959, token usage: 0.82, gen throughput (token/s): 4594.01, #queue-req: 417```
+```[gpu_id=0] Decode batch. #running-req: 233, #token: 370959, token usage: 0.82, gen throughput (token/s): 4594.01, #queue-req: 417```
 
 ### Tune Your Request Submission Speed
 `#queue-req` indicates the number of requests in the queue. If you frequently see `#queue-req == 0`, it suggests you are bottlenecked by the request submission speed.
diff --git a/python/sglang/backend/litellm.py b/python/sglang/backend/litellm.py
index 9a0060f33..dc89dc16d 100644
--- a/python/sglang/backend/litellm.py
+++ b/python/sglang/backend/litellm.py
@@ -9,6 +9,7 @@ try:
     import litellm
 except ImportError as e:
     litellm = e
+    litellm.num_retries = 1
 
 
 class LiteLLM(BaseBackend):
diff --git a/python/sglang/srt/managers/controller/tp_worker.py b/python/sglang/srt/managers/controller/tp_worker.py
index 30e209b22..7fb5e1b3b 100644
--- a/python/sglang/srt/managers/controller/tp_worker.py
+++ b/python/sglang/srt/managers/controller/tp_worker.py
@@ -111,7 +111,10 @@ class ModelTpServer:
             f"context_len={self.model_config.context_len}, "
         )
         if self.tp_rank == 0:
-            logger.info(f"server_args: {server_args.print_mode_args()}")
+            logger.info(
+                f"[gpu_id={self.gpu_id}] "
+                f"server_args: {server_args.print_mode_args()}"
+            )
 
         # Init cache
         self.tree_cache = RadixCache(
@@ -226,7 +229,7 @@ class ModelTpServer:
                             self.num_generated_tokens = 0
                             self.last_stats_tic = time.time()
                             logger.info(
-                                f"[gpu_id={self.gpu_id}] "
+                                f"[gpu_id={self.gpu_id}] Decode batch. "
                                 f"#running-req: {len(self.running_batch.reqs)}, "
                                 f"#token: {num_used}, "
                                 f"token usage: {num_used / self.max_total_num_tokens:.2f}, "
@@ -397,12 +400,13 @@ class ModelTpServer:
                 self.tree_cache_metrics["hit"] / self.tree_cache_metrics["total"]
             )
             logger.info(
-                f"new fill batch. #seq: {len(can_run_list)}. "
-                f"#cached_token: {hit_tokens}. "
-                f"#new_token: {new_batch_input_tokens}. "
-                f"#remaining_req: {len(self.forward_queue) - len(can_run_list)}. "
-                f"#running_req: {running_req}. "
-                f"tree_cache_hit_rate: {100.0 * tree_cache_hit_rate:.2f}%. "
+                f"[gpu_id={self.gpu_id}] Prefil batch. "
+                f"#new-seq: {len(can_run_list)}, "
+                f"#new-token: {new_batch_input_tokens}, "
+                f"#cached-token: {hit_tokens}, "
+                f"cache hit rate: {100.0 * tree_cache_hit_rate:.2f}%, "
+                f"#running-req: {running_req}, "
+                f"#queue-req: {len(self.forward_queue) - len(can_run_list)}"
             )
             # logger.debug(
             #    f"fsm_cache_hit_rate: {100.0 * self.regex_fsm_cache.get_cache_hit_rate():.2f}%. "