From 385599cb04ff41c7bd3c0518c189ee5efee4ad11 Mon Sep 17 00:00:00 2001
From: fzyzcjy <5236035+fzyzcjy@users.noreply.github.com>
Date: Tue, 4 Nov 2025 02:17:43 +0800
Subject: [PATCH] Fix error when calling quantization (#12548)

---
 python/sglang/srt/layers/quantization/modelopt_quant.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/python/sglang/srt/layers/quantization/modelopt_quant.py b/python/sglang/srt/layers/quantization/modelopt_quant.py
index 0017f740d..06d87c048 100755
--- a/python/sglang/srt/layers/quantization/modelopt_quant.py
+++ b/python/sglang/srt/layers/quantization/modelopt_quant.py
@@ -56,6 +56,8 @@ try:
         from flashinfer import fp4_quantize
     else:
         from sgl_kernel import scaled_fp4_quant as fp4_quantize
+
+    from flashinfer import fp4_quantize as fp4_quantize_flashinfer
 except ImportError:
     fp4_quantize = None
 
@@ -1571,7 +1573,7 @@ class ModelOptNvFp4FusedMoEMethod(FusedMoEMethodBase):
 
                 # Quantize before comm, swizzle after.
                 if x.shape[0] > 0:
-                    x, x_sf = fp4_quantize(
+                    x, x_sf = fp4_quantize_flashinfer(
                         x, layer.w13_input_scale_quant, is_sf_swizzled_layout=False
                     )
                 else: