From 15ff6982b94fad2fdf44607a040a146a5ae5d62c Mon Sep 17 00:00:00 2001
From: mrhaoxx <mr.haoxx@gmail.com>
Date: Thu, 27 Nov 2025 06:59:15 +0800
Subject: [PATCH] Support  KTransformers for Qwen3-VL moe (#13983)

---
 python/sglang/srt/models/qwen3_vl_moe.py | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/python/sglang/srt/models/qwen3_vl_moe.py b/python/sglang/srt/models/qwen3_vl_moe.py
index e810dae7e..e3e9e07d1 100644
--- a/python/sglang/srt/models/qwen3_vl_moe.py
+++ b/python/sglang/srt/models/qwen3_vl_moe.py
@@ -22,6 +22,7 @@ import torch
 import torch.nn as nn
 
 from sglang.srt.configs.qwen3_vl import Qwen3VLMoeConfig, Qwen3VLMoeTextConfig
+from sglang.srt.eplb.expert_location import ModelConfigForExpertLocation
 from sglang.srt.layers.moe.fused_moe_triton.layer import FusedMoE
 from sglang.srt.layers.quantization.base_config import QuantizationConfig
 from sglang.srt.model_executor.forward_batch_info import ForwardBatch, PPProxyTensors
@@ -326,5 +327,13 @@ class Qwen3VLMoeForConditionalGeneration(Qwen3VLForConditionalGeneration):
         #         if isinstance(self.model.layers[layer_id].mlp, Qwen3MoeSparseMoeBlock)
         #     }
 
+    @classmethod
+    def get_model_config_for_expert_location(cls, config):
+        return ModelConfigForExpertLocation(
+            num_layers=config.text_config.num_hidden_layers,
+            num_logical_experts=config.text_config.num_experts,
+            num_groups=None,
+        )
+
 
 EntryClass = Qwen3VLMoeForConditionalGeneration