ggml-org/llama.cpp

Pull Requests Commits

iq1_s: fix dequantize on the CPU

Iwan Kawrakow committed 1 year ago

5440a127

iq1_s: very slightly faster dequantize on Metal

Iwan Kawrakow committed 1 year ago

436c65e1

iq1_s: make Metal work with new version

Iwan Kawrakow committed 1 year ago

da4528bc

iq1_s: make Neon work with new version.

Iwan Kawrakow committed 1 year ago

4fba3e00

iq1_s: make scalar and AVX2 work with the new version

Iwan Kawrakow committed 1 year ago

c09f7349

iq1_s: we can do even better

Iwan Kawrakow committed 1 year ago

82380acf

Better 1.5 bit quantization (#5971)

ikawrakow committed 1 year ago

Verified be858f62

[SYCL] Add q3_s and q1_s (#5886)

abhilash1910 committed 1 year ago

Verified ef3ced26

[SYCL] Add support for SYCL Nvidia target (#5738)

AidanBeltonS committed 1 year ago

Verified 3814a073

metal : move mm_id indices to shared mem (#5982)

ggerganov committed 1 year ago

Verified bb6d00bb

android : fix utf8 decoding error (#5935)

zhangfuwen committed 1 year ago

Verified 7ab7b733

readme : update hot topics

ggerganov committed 1 year ago

Verified d9f65c97

ggerganov committed 1 year ago

Verified b838b53a

ggml : try fix 32-bit arm compat (whisper/1938)

ggerganov committed 1 year ago

Verified df4dc3e7

ggml : remove __constant__ specifier for CUDA tables (#5940)

ggerganov committed 1 year ago

Verified bf47a5ee

server: ci: windows build and tests (#5968)

phymbert committed 1 year ago

Verified fa8a809a

llama : add support for GritLM (#5959)

dranger003 committed 1 year ago

Verified bcebd7db

grammar : verify parsed state (#5950)

HanClinto committed 1 year ago

Verified 2960eae8

nix: update flake.lock (#5969)

ggerganov committed 1 year ago

Verified c7854147

server: benchmark: chat/completions scenario and other llm servers comparison (#5941)

phymbert committed 1 year ago

Verified 621e86b3

server : print chat template info

ggerganov committed 1 year ago

Verified 77d1ac7e

perplexity : support using multiple sequences to allow larger batch sizes (#5946)

slaren committed 1 year ago

Verified d894f352

readme : update hot topics

ggerganov committed 1 year ago

Verified 098dbaab

ggml : fix unnecessary f32 -> f16 -> f32 casts (mmla) (#5951)

ggerganov committed 1 year ago

Verified 8380ecfb

server : fix metrics init (#5964)

ggerganov committed 1 year ago

Verified 58308a0e

ggml : remove old quantization functions (#5942)

ggerganov committed 1 year ago

Verified 5b097973

server : clarify some items in the readme (#5957)

ggerganov committed 1 year ago

Verified 97c09585

server : normalize embeddings (#5956)

redlion0929 committed 1 year ago

Verified fb215c38

tests : gitignore ggml-common.h

ggerganov committed 1 year ago

Verified 2c4f566c

server : fix passing prompt as tokens (#5955)

z80maniac committed 1 year ago

Verified 0db32bea

Older