From f566b9b74897b73034f0bad7a691e2558ec9c2dd Mon Sep 17 00:00:00 2001
From: biondizzle <biondizzle@gmail.com>
Date: Tue, 2 Jun 2026 10:02:01 +0000
Subject: [PATCH] Fix FP8 quantize return type (2-tuple not 3)

---
 dsv4/kernels/cuda/kv_quantize.cu | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/dsv4/kernels/cuda/kv_quantize.cu b/dsv4/kernels/cuda/kv_quantize.cu
index d29246ee..689c5d04 100644
--- a/dsv4/kernels/cuda/kv_quantize.cu
+++ b/dsv4/kernels/cuda/kv_quantize.cu
@@ -293,7 +293,7 @@ std::tuple<torch::Tensor, torch::Tensor> quantize_nvfp4_from_fp32_cuda(
     return {out_fp4.view(torch::kFloat4_e2m1fn_x2), out_sf.view(torch::kFloat8_e4m3fn)};
 }
 
-std::tuple<torch::Tensor, torch::Tensor, torch::Tensor> quantize_fp8_e4m3_from_fp32_cuda(
+std::tuple<torch::Tensor, torch::Tensor> quantize_fp8_e4m3_from_fp32_cuda(
     torch::Tensor input
 ) {
     int M = input.size(0);