From 69193f71225edf37047597b03099011c26146c5e Mon Sep 17 00:00:00 2001
From: Ke Bao <ispobaoke@gmail.com>
Date: Sat, 1 Nov 2025 16:27:31 +0800
Subject: [PATCH] Filter tokenizer warning for kimi models (#12485)

---
 python/sglang/srt/utils/hf_transformers_utils.py | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/python/sglang/srt/utils/hf_transformers_utils.py b/python/sglang/srt/utils/hf_transformers_utils.py
index 637390ba5..30d0dc16d 100644
--- a/python/sglang/srt/utils/hf_transformers_utils.py
+++ b/python/sglang/srt/utils/hf_transformers_utils.py
@@ -15,6 +15,7 @@
 
 import contextlib
 import json
+import logging
 import os
 import tempfile
 import warnings
@@ -348,6 +349,12 @@ def get_context_length(config):
 _FAST_LLAMA_TOKENIZER = "hf-internal-testing/llama-tokenizer"
 
 
+# Filter warnings like: https://github.com/sgl-project/sglang/issues/8082
+class TokenizerWarningsFilter(logging.Filter):
+    def filter(self, record: logging.LogRecord) -> bool:
+        return "Calling super().encode with" not in record.getMessage()
+
+
 def get_tokenizer(
     tokenizer_name: str,
     *args,
@@ -393,6 +400,10 @@ def get_tokenizer(
             clean_up_tokenization_spaces=False,
             **kwargs,
         )
+        # Filter tokenizer warnings
+        logging.getLogger(tokenizer.__class__.__module__).addFilter(
+            TokenizerWarningsFilter()
+        )
     except TypeError as e:
         # The LLaMA tokenizer causes a protobuf error in some environments.
         err_msg = (