From: Michael Wand <redacted>
Date: Sat, 10 Jan 2026 09:16:07 +0000 (-0800)
Subject: Corrected: changed s13 = src1->nb[3] instead of nb[2] (llama/18724)
X-Git-Tag: v0.9.6~88
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=4a4cd3735f6469107f2b372d3db506c9ba2132ec;p=pkg%2Fggml%2Fsources%2Fggml

Corrected: changed s13 = src1->nb[3] instead of nb[2] (llama/18724)
---

diff --git a/src/ggml-cuda/mmq.cu b/src/ggml-cuda/mmq.cu
index ceb95758..c9aa7024 100644
--- a/src/ggml-cuda/mmq.cu
+++ b/src/ggml-cuda/mmq.cu
@@ -190,7 +190,7 @@ void ggml_cuda_mul_mat_q(
     {
         const int64_t s11 = src1->nb[1] / ts_src1;
         const int64_t s12 = src1->nb[2] / ts_src1;
-        const int64_t s13 = src1->nb[2] / ts_src1;
+        const int64_t s13 = src1->nb[3] / ts_src1;
 
         if (use_native_mxfp4) {
             quantize_mmq_mxfp4_cuda(src1_d, ids_src1.get(), src1_q8_1.get(), src0->type, ne10, s11, s12, s13,