Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

vulkan: Support GGML_TYPE_NVFP4 (#21455)

Browse Source 
This adds nvfp4 support for get_rows, dequant, and mul_mat(_id). For
mul_mat, it does not add support for the dp4/q8_1 path, it's all via
fp16/fp32.
This commit is contained in:
Jeff Bolz
2026-04-14 11:34:23 +02:00
committed by GitHub
parent e489a5ca0e
commit 6a6780a232
8 changed files with 171 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-vulkan/vulkan-shaders/copy_from_quant.comp
+1 -1
View File
@@ -4,7 +4,7 @@
#include "generic_unary_head.glsl"
#include "dequant_funcs.glsl"
#if defined(DATA_A_IQ4_NL) || defined(DATA_A_MXFP4)
#if defined(DATA_A_IQ4_NL) || defined(DATA_A_MXFP4) || defined(DATA_A_NVFP4)
// 16 invocations needed for init_iq_shmem
layout(local_size_x = 16, local_size_y = 1, local_size_z = 1) in;
#else