ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-04-28 10:21:48 +00:00
Code Issues Packages Projects Releases Wiki Activity

IQ3_XXS_R4 (#153)

Browse Source 
* iq3_xxs_r4: 1st shot on Zen4

PP-512: 107 t/s -> 137 t/s
TG-128(1 thread): 2.64 t/s -> 3.44 t/s

* iq4_xxs_r4: WIP

* iq4_xxs_r4: 1st shot at AVX2

Note: there is a bug in the AVX2 implementation for nrc_y = 1
for IQ quants with blocks of 32. I have fixed it for now by
using the nrc_y > 1 implementation (which works) also for nrc_y = 1.

* iq3_xxs_r4: NEON

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow
2024-12-20 09:12:48 +01:00
committed by GitHub
parent dfa12b7f91
commit f4170f78bd
10 changed files with 351 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
include/llama.h
View File

@@ -188,6 +188,7 @@ extern "C" {
LLAMA_FTYPE_MOSTLY_Q4_K_R4 = 214, // except 1d tensors
LLAMA_FTYPE_MOSTLY_Q5_K_R4 = 216, // except 1d tensors
LLAMA_FTYPE_MOSTLY_Q6_K_R4 = 218, // except 1d tensors
LLAMA_FTYPE_MOSTLY_IQ3_XXS_R4 = 223, // except 1d tensors
LLAMA_FTYPE_MOSTLY_IQ4_NL_R4 = 225, // except 1d tensors
LLAMA_FTYPE_MOSTLY_IQ4_XS_R4 = 230, // except 1d tensors
LLAMA_FTYPE_MOSTLY_Q6_0_R4 = 335, // except 1d tensors