From c224a4c6cc466c70694ce2ba29f1b7ff777392cc Mon Sep 17 00:00:00 2001
From: Baizhou Zhang <sobereddiezhang@gmail.com>
Date: Tue, 14 Oct 2025 11:49:33 -0700
Subject: [PATCH] Fix log for chunked prefix cache (#11624)

---
 python/sglang/srt/model_executor/model_runner.py | 15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)

diff --git a/python/sglang/srt/model_executor/model_runner.py b/python/sglang/srt/model_executor/model_runner.py
index 10bc0f866..aceb572c9 100644
--- a/python/sglang/srt/model_executor/model_runner.py
+++ b/python/sglang/srt/model_executor/model_runner.py
@@ -174,6 +174,15 @@ MLA_ATTENTION_BACKENDS = [
     "nsa",
 ]
 
+CHUNKED_PREFIX_CACHE_SUPPORTED_ATTENTION_BACKENDS = [
+    "flashinfer",
+    "fa3",
+    "fa4",
+    "flashmla",
+    "cutlass_mla",
+    "trtllm_mla",
+]
+
 
 def add_mla_attention_backend(backend_name):
     if backend_name not in MLA_ATTENTION_BACKENDS:
@@ -604,7 +613,11 @@ class ModelRunner:
                     f"{self.model_config.hf_config.model_type}"
                 )
 
-        if not self.use_mla_backend:
+        if (
+            not self.use_mla_backend
+            or server_args.attention_backend
+            not in CHUNKED_PREFIX_CACHE_SUPPORTED_ATTENTION_BACKENDS
+        ):
             server_args.disable_chunked_prefix_cache = True
 
         if not server_args.disable_chunked_prefix_cache: