ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-05-01 03:41:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

IQ2_XS_R4 (#155)

Browse Source 
* iq2_xs_r4: Zen4

* iq2_xs_r4: AVX2

* iq2_xs_r4: slightly better matrix x vector on AVX2

* iq2_xs_r4: NEON - not much better than iq2_xs

* iq2_xs_r4: slightly better NEON

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow
2024-12-21 08:32:39 +01:00
committed by GitHub
parent 9dfd69bd93
commit 6892554e43
10 changed files with 363 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/include/ggml.h
View File

@@ -419,6 +419,7 @@ extern "C" {
GGML_TYPE_Q5_K_R4 = 213,
GGML_TYPE_Q6_K_R4 = 214,
GGML_TYPE_IQ2_XXS_R4= 216,
GGML_TYPE_IQ2_XS_R4 = 217,
GGML_TYPE_IQ3_XXS_R4= 218,
GGML_TYPE_IQ4_NL_R4 = 220,
GGML_TYPE_IQ4_XS_R4 = 223,
@@ -499,6 +500,7 @@ extern "C" {
GGML_FTYPE_MOSTLY_Q5_K_R4 = 213, // except 1d tensors
GGML_FTYPE_MOSTLY_Q6_K_R4 = 214, // except 1d tensors
GGML_FTYPE_MOSTLY_IQ2_XXS_R4= 215, // except 1d tensors
GGML_FTYPE_MOSTLY_IQ2_XS_R4 = 216, // except 1d tensors
GGML_FTYPE_MOSTLY_IQ3_XXS_R4= 217, // except 1d tensors
GGML_FTYPE_MOSTLY_IQ4_NL_R4 = 219, // except 1d tensors
GGML_FTYPE_MOSTLY_IQ4_XS_R4 = 222, // except 1d tensors