From 7ae078b4174279aed8ef8ab4d1cd433eb71e4e53 Mon Sep 17 00:00:00 2001
From: Chenjie Luo <108829653+cjluo-nv@users.noreply.github.com>
Date: Mon, 6 Oct 2025 11:21:38 -0700
Subject: [PATCH] Explicitly register real quant gemms

Signed-off-by: Chenjie Luo <108829653+cjluo-nv@users.noreply.github.com>
---
 .../torch/quantization/backends/__init__.py     | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

diff --git a/modelopt/torch/quantization/backends/__init__.py b/modelopt/torch/quantization/backends/__init__.py
index 92317d92b..c4c2fadd5 100644
--- a/modelopt/torch/quantization/backends/__init__.py
+++ b/modelopt/torch/quantization/backends/__init__.py
@@ -15,5 +15,18 @@
 
 """Quantization backends."""
 
-from .gemm_registry import *
-from .nvfp4_gemm import *
+from .fp8_per_tensor_gemm import Fp8PerTensorLinear, _fp8_availability_check
+from .gemm_registry import gemm_registry
+from .nvfp4_gemm import Nvfp4Linear, _nvfp4_availability_check
+
+# Register default implementations
+gemm_registry.register(
+    gemm_func=Fp8PerTensorLinear.apply,
+    availability_check=_fp8_availability_check,
+)
+
+# Register default implementations
+gemm_registry.register(
+    gemm_func=Nvfp4Linear.apply,
+    availability_check=_nvfp4_availability_check,
+)