From 533896fd63e496f6f60d97ee13bbd48c579fcd76 Mon Sep 17 00:00:00 2001
From: AIbin <37361953+chang-wenbin@users.noreply.github.com>
Date: Fri, 10 Oct 2025 21:31:38 +0800
Subject: [PATCH] fix paddle_peak_increase size (#4355)

---
 fastdeploy/worker/gpu_worker.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fastdeploy/worker/gpu_worker.py b/fastdeploy/worker/gpu_worker.py
index bf1b88dab..601efd16b 100644
--- a/fastdeploy/worker/gpu_worker.py
+++ b/fastdeploy/worker/gpu_worker.py
@@ -141,7 +141,7 @@ class GpuWorker(WorkerBase):
         paddle_allocated_mem_after_run = paddle.device.cuda.max_memory_allocated(local_rank)
 
         model_block_memory_used = self.cal_theortical_kvcache()
-        paddle_peak_increase = paddle_reserved_mem_after_run - paddle_allocated_mem_before_run
+        paddle_peak_increase = paddle_allocated_mem_after_run - paddle_allocated_mem_before_run
 
         paddle.device.cuda.empty_cache()