From b62633447522fbf84ca9bcc5e8ee063a4f268a1f Mon Sep 17 00:00:00 2001
From: Yineng Zhang <me@zhyncs.com>
Date: Mon, 8 Dec 2025 23:11:18 -0800
Subject: [PATCH] fix: make override DeepseekV2Model work (#14707)

---
 python/sglang/srt/models/deepseek_v2.py | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/python/sglang/srt/models/deepseek_v2.py b/python/sglang/srt/models/deepseek_v2.py
index ab67e03a6..2b1c3c04d 100644
--- a/python/sglang/srt/models/deepseek_v2.py
+++ b/python/sglang/srt/models/deepseek_v2.py
@@ -3202,7 +3202,6 @@ class DeepseekV2Model(nn.Module):
 class DeepseekV2ForCausalLM(nn.Module):
     # for quark model load
     packed_modules_mapping = {}
-    model_cls = DeepseekV2Model
 
     def __init__(
         self,
@@ -3228,7 +3227,7 @@ class DeepseekV2ForCausalLM(nn.Module):
         self.quant_config = quant_config
         self.determine_num_fused_shared_experts()
         self.use_nsa = is_deepseek_nsa(config)
-        self.model = self.model_cls(
+        self.model = DeepseekV2Model(
             config, quant_config, prefix=add_prefix("model", prefix)
         )
         if self.pp_group.is_last_rank: