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

1.5 bit quantization (#5453)

Browse source 
* iq1_s: WIP basics

* iq1_s: CUDA is working

* iq1_s: scalar CPU dot product

* iq1_s: WIP AVX2 dot product - something is not right

* Fix tests

* Fix shadow warnings

* Fix after merge with latest master

* iq1_s: AVX2 finally works

* iq1_s: ARM_NEON dot product. Works, but not very fast

* iq1_s: better grid

* iq1_s: use IQ2_XXS for attn_output

At a cost of 0.04 extra bpw this gives a big improvement in PPL.

* iq1_s: Metal basics

Dequantize works, but not dot product

* iq1_s: Metal works, but quite slow

As usual, Apple Silicon does not like the code I write.

* iq1_s: Tests

* iq1_s: slightly faster dot product

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow 2024-02-18 18:16:55 +02:00 • committed by GitHub
parent c8e0d7efeb
commit bd2d4e393b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
12 changed files with 1286 additions and 48 deletions
Download patch file Download diff file Expand all files Collapse all files

1
llama.h
View file

@ -100,6 +100,7 @@ extern "C" {
LLAMA_FTYPE_MOSTLY_Q2_K_S = 21, // except 1d tensors
LLAMA_FTYPE_MOSTLY_Q3_K_XS = 22, // except 1d tensors
LLAMA_FTYPE_MOSTLY_IQ3_XXS = 23, // except 1d tensors
LLAMA_FTYPE_MOSTLY_IQ1_S = 24, // except 1d tensors
LLAMA_FTYPE_GUESSED = 1024, // not specified in the model file
};