]> git.djapps.eu Git - pkg/ggml/sources/ggml/commit
vulkan: matmul dequantization improvements (llama/12015)
authorEve <redacted>
Fri, 28 Feb 2025 07:20:08 +0000 (07:20 +0000)
committerGeorgi Gerganov <redacted>
Tue, 4 Mar 2025 19:24:42 +0000 (21:24 +0200)
commita55b828f8cd3a66f4865e07cdb18e8ba8846ae0c
tree01b266264fd54a6d017922efe79da91a858c2896
parente0c456ae09dceee483771ac42ba7b0c25d164c5a
vulkan: matmul dequantization improvements (llama/12015)

* faster dequant for old quants

* dont use unpack for iq4_nl

* vec2 unpack for q8
src/ggml-vulkan/vulkan-shaders/dequant_funcs.comp
src/ggml-vulkan/vulkan-shaders/dequant_funcs_cm2.comp
src/ggml-vulkan/vulkan-shaders/mul_mm.comp
src/ggml-vulkan/vulkan-shaders/types.comp
src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp