ggml-org/llama.cpp

Pull Requests Commits

llama : remove experimental stuff

ggerganov committed 2 years ago

Verified e7e7b114

metal : revert the concurrnecy change because it was wrong

ggerganov committed 2 years ago

Verified e343b8b4

metal : relax conditions on fast matrix multiplication kernel

ggerganov committed 2 years ago

Verified 336afbcb

whisper : tokenizer fix + re-enable tokenizer test for LLaMa (#3096)

goerch committed 2 years ago

Verified 71ca2fad

cmake : add a compiler flag check for FP16 format (#3086)

RossComputerGuy committed 2 years ago

Verified 1b6c650d

CUDA: mul_mat_q RDNA2 tunings (#2910)

JohannesGaessler committed 2 years ago

Verified 0a5eebb4

speculative: add --n-gpu-layers-draft option (#3063)

sozforex committed 2 years ago

Verified 84e72365

arm64 support for windows (#3007)

es0m committed 2 years ago

Verified b52b29ab

CUDA: fix LoRAs (#3130)

JohannesGaessler committed 2 years ago

Verified 4f7cd6ba

CUDA: fix mul_mat_q not used for output tensor (#3127)

JohannesGaessler committed 2 years ago

Verified 89e89599

CUDA: lower GPU latency + fix Windows performance (#3110)

JohannesGaessler committed 2 years ago

Verified d54a4027

cmake : support build for iOS/tvOS (#3116)

jhen0409 committed 2 years ago

Verified 1b0d0925

CUDA: add device number to error messages (#3112)

JohannesGaessler committed 2 years ago

Verified 8a4ca9af

metal : PP speedup (#3084)

ikawrakow committed 2 years ago

Verified f31b6f4e

convert: remove most of the n_mult usage in convert.py (#3098)

Green-Sky committed 2 years ago

Verified 6eeb4d90

metal : support for Swift (#3078)

kchro3 committed 2 years ago

Verified 21ac3a15

metal : support build for iOS/tvOS (#3089)

jhen0409 committed 2 years ago

Verified 4fd54779

flake : add train-text-from-scratch to flake.nix (#3042)

takov751 committed 2 years ago

Verified ec2a24fe

readme : fix typo (#3043)

eltociear committed 2 years ago

Verified 7d99aca7

metal : Q3_K speedup (#2995)

ikawrakow committed 2 years ago

Verified ba7ffbb2

examples : make n_ctx warning work again (#3066)

cebtenzzre committed 2 years ago

Verified e64f5b55

readme : update hot tpoics

ggerganov committed 2 years ago

Verified 94f10b91

sync : ggml (CUDA GLM RoPE + POSIX) (#3082)

ggerganov committed 2 years ago

Verified b3e9852e

build : do not use _GNU_SOURCE gratuitously (#2035)

przemoc committed 2 years ago

Verified cb6c44c5

docker : add git to full-cuda.Dockerfile main-cuda.Dockerfile (#3044)

polym committed 2 years ago

Verified a21baeb1

Update deprecated GGML TheBloke links to GGUF (#3079)

SleepyYui committed 2 years ago

Verified 6ff712a6

ggml-alloc : correctly check mmap return value for errors (#3075)

slaren committed 2 years ago

Verified ebc96086

enable CPU HBM (#2603)

jikunshang committed 2 years ago

Verified 7f412dab

convert : fix F32 ftype not being saved (#3048)

cebtenzzre committed 2 years ago

Verified 6336d834

fix some warnings from gcc and clang-tidy (#3038)

cebtenzzre committed 2 years ago

Verified 00d62adb

Older