From 1cba2e05d383720d4ac3861219cda9e41f0146c5 Mon Sep 17 00:00:00 2001
From: GoldPancake <56388518+Deleter-D@users.noreply.github.com>
Date: Tue, 18 Nov 2025 11:22:00 +0800
Subject: [PATCH] [ForRLRelease] temporary change mtp msg size (#5103)

---
 custom_ops/gpu_ops/speculate_decoding/speculate_msg.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/custom_ops/gpu_ops/speculate_decoding/speculate_msg.h b/custom_ops/gpu_ops/speculate_decoding/speculate_msg.h
index dbc177b4c..6279027c8 100644
--- a/custom_ops/gpu_ops/speculate_decoding/speculate_msg.h
+++ b/custom_ops/gpu_ops/speculate_decoding/speculate_msg.h
@@ -22,7 +22,7 @@
 #include "paddle/extension.h"
 
 #define MAX_BSZ 256
-#define MAX_DRAFT_TOKENS 6
+#define MAX_DRAFT_TOKENS 2
 
 // TODO: replace all msgdata in speculate-decoding
 struct speculate_msgdata {