vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

vulkan: initial support for IQ1_S and IQ1_M quantizations

Browse source
This commit is contained in:
Rémy O 2025-01-30 04:28:49 +01:00
parent 98f6b0fd1e
commit fa92caae18
14 changed files with 537 additions and 28 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec.comp
View file

@ -133,7 +133,7 @@ void compute_outputs(const uint32_t first_row, const uint32_t num_rows) {
void main() {
const uint first_row = NUM_ROWS * (gl_WorkGroupID.x + gl_NumWorkGroups.x * gl_WorkGroupID.z);
#if defined(DATA_A_IQ2_XXS) || defined(DATA_A_IQ2_XS) || defined(DATA_A_IQ2_S) || defined(DATA_A_IQ3_XXS) || defined(DATA_A_IQ3_S) || defined(DATA_A_IQ4_XS) || defined(DATA_A_IQ4_NL)
#if defined(DATA_A_IQ1_S) || defined(DATA_A_IQ1_M) || defined(DATA_A_IQ2_XXS) || defined(DATA_A_IQ2_XS) || defined(DATA_A_IQ2_S) || defined(DATA_A_IQ3_XXS) || defined(DATA_A_IQ3_S) || defined(DATA_A_IQ4_XS) || defined(DATA_A_IQ4_NL)
init_iq_shmem(gl_WorkGroupSize);
#endif