From f55d608c897965202335053bab955ac20010f95b Mon Sep 17 00:00:00 2001
From: fzyzcjy <5236035+fzyzcjy@users.noreply.github.com>
Date: Sun, 28 Dec 2025 11:22:20 +0800
Subject: [PATCH] Tiny fix cannot launch nvfp4 checkpoint with bf16 kv cache
 (#15986)

---
 python/sglang/srt/server_args.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/python/sglang/srt/server_args.py b/python/sglang/srt/server_args.py
index 84fcfd7cb..abb3fde8d 100644
--- a/python/sglang/srt/server_args.py
+++ b/python/sglang/srt/server_args.py
@@ -1594,9 +1594,9 @@ class ServerArgs:
                 )
                 self.page_size = 64
 
-            if self.kv_cache_dtype not in ["fp8_e4m3", "fp4_e2m1", "auto"]:
+            if self.kv_cache_dtype not in ["fp8_e4m3", "fp4_e2m1", "bf16", "auto"]:
                 raise ValueError(
-                    "TensorRT-LLM MLA backend only supports kv-cache-dtype of fp8_e4m3, fp4_e2m1, or auto."
+                    "TensorRT-LLM MLA backend only supports kv-cache-dtype of fp8_e4m3, fp4_e2m1, bf16, or auto."
                 )
 
         if (