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

IQ4_XS: a 4.25 bpw quantization (#5747)

Browse source 
* Try IQ4_NL with blocks of 64 - does not look good

* iq4_xs: go to super-blocks of 256 and 6-bit scales for blocks of 32

* iq4_xs: CUDA works - 133.2 t/s

* iq4_xs: AVX2 dot product

* iq4_xs: ARM_NEON dot product

* iq4_nl: Metal implementation

As usual, Metal / Apple Silicon don't like my quants.

* iq3_xs: minor fix

* iq4_xs: shrink by using IQ3_S for attn_k and attn_q

* iq4_xs: revert using IQ3_S for attn_k and attn_v

PPL vs size is good, but CPU performance suffers: on M2 Max
TG-128 drops to 21.7 t/s from 28.8, and on a Ryzen-7950X
to 14.5 t/s from 15.8 t/s. On CUDA we have 135 t/s when
using IQ3_S vs 133 t/s with pure IQ4_XS.

* Fix CI

* iq4_xs: Added forgotten check for 256 divisibility

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow 2024-02-27 16:34:24 +02:00 • committed by GitHub
parent c24a2a6e60
commit 0becb22ac0
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
11 changed files with 668 additions and 38 deletions
Download patch file Download diff file Expand all files Collapse all files

1
llama.h
View file

@ -115,6 +115,7 @@ extern "C" {
LLAMA_FTYPE_MOSTLY_IQ3_M = 27, // except 1d tensors
LLAMA_FTYPE_MOSTLY_IQ2_S = 28, // except 1d tensors
LLAMA_FTYPE_MOSTLY_IQ2_M = 29, // except 1d tensors
LLAMA_FTYPE_MOSTLY_IQ4_XS = 30, // except 1d tensors
LLAMA_FTYPE_GUESSED = 1024, // not specified in the model file
};