From 43c8349d51ae67494866be8927be869a0a32d41e Mon Sep 17 00:00:00 2001
From: songsenand <songsenand@163.com>
Date: Thu, 26 Feb 2026 14:36:25 +0800
Subject: [PATCH] =?UTF-8?q?fix(trainer):=20=E7=A7=BB=E9=99=A4=E5=9B=BA?=
 =?UTF-8?q?=E5=AE=9A=E6=80=BB=E6=AD=A5=E6=95=B0=EF=BC=8C=E4=BD=BF=E7=94=A8?=
 =?UTF-8?q?=E5=AE=9E=E9=99=85=E5=81=9C=E6=AD=A2=E6=89=B9=E6=AC=A1=E8=AE=A1?=
 =?UTF-8?q?=E7=AE=97=20warmup=20=E6=AD=A5=E6=95=B0?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 src/trainer/model.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/trainer/model.py b/src/trainer/model.py
index 933d45c..0132f02 100644
--- a/src/trainer/model.py
+++ b/src/trainer/model.py
@@ -455,7 +455,7 @@ class MoEModel(nn.Module):
 
         scaler = amp.GradScaler(enabled=mixed_precision)
 
-        total_steps = max(stop_batch, 2e5)
+        total_steps = stop_batch
         warmup_steps = int(total_steps * warmup_ratio)
         logger.info(f"Training Start: Steps={total_steps}, Warmup={warmup_steps}")
         processed_batches = 0