ggerganov/llama.cpp

Pull Requests Commits

ggml : speed-up Q5_0 + Q5_1 at 4 threads

ggerganov committed 2 years ago

Verified e116eb63

llama : fix model magic/version write

ggerganov committed 2 years ago

Verified ffd76e18

llama : produce error upon loading old model files

ggerganov committed 2 years ago

Verified 4201fa5c

ggml : uniform 5th bit extraction

ggerganov committed 2 years ago

Verified 0e48eb6f

AVX implementations (#1370)

sw committed 2 years ago

Verified 948d1248

scripts : add script for measuring the time per token

ggerganov committed 2 years ago

Verified d155f0f8

ggml : fix Q5_0 quantization

ggerganov committed 2 years ago

Verified 8fbf7777

ggml : minor formatting

ggerganov committed 2 years ago

Verified 60f62bbc

ggml : remove Q4_2 mode

ggerganov committed 2 years ago

Verified 7cdc08a5

ggml : update cuBLAS + normalize variable names

ggerganov committed 2 years ago

Verified b47bd287

ggml : fix Q4_1 quantization

ggerganov committed 2 years ago

Verified c2166569

ggml : remove WASM SIMD bit shuffling + remove vzip for ARM 32-bit

ggerganov committed 2 years ago

Verified 4991499a

ggml : simplify scalar dot

ggerganov committed 2 years ago

Verified ba953d6e

ggml : remove Q5_1 bit shuffling (ARM NEON + scalar)

ggerganov committed 2 years ago

Verified c7af9042

ggml : 2x faster scalar implementations

ggerganov committed 2 years ago

Verified 39bb8e7d

ggml : remove Q5_0 bit shuffling (ARM NEON)

ggerganov committed 2 years ago

Verified 796f8ae2

ggml : remove Q4_2 bit shuffling (WIP, BROKEN)

ggerganov committed 2 years ago

Verified a6a1d96c

ggml : nibbles_from_floats() + bytes_from_nibbles() (ARM NEON)

ggerganov committed 2 years ago

Verified 086cfea1

ggml : remove Q4_1 bit shuffling (ARM NEON + reference)

ggerganov committed 2 years ago

Verified edb6c8bb

ggml : remove Q4_0 bit shufling (ARM NEON)

ggerganov committed 2 years ago

Verified a546dc6d

readme : add TOC and Pygmalion instructions (#1359)

AlpinDale committed 2 years ago

Verified fe60904e

llama : fix hparams shadow (#1367)

prusnak committed 2 years ago

Verified 003ba2fb

llama : require first token to be BOS (#1303)

ggerganov committed 2 years ago

Verified f9a63649

convert: add ability to convert safetensors files (#1276)

ubik2 committed 2 years ago

Verified 95078cc5

Documented CUDA reproducibility, added warning (#1346)

JohannesGaessler committed 2 years ago

Verified 1f48b0ab

CI: add Windows CLBlast and OpenBLAS builds (#1277)

SlyEcho committed 2 years ago

Verified e1295513

ggml : Allow usage of CLBlast alongside Accelerate.framework (#1336)

swittk committed 2 years ago

Verified 1b0fd454

Remove default arguments from sampling functions (#1343)

j-f1 committed 2 years ago

Verified 39240885

makefile: automatic Arch Linux detection (#1332)

daniandtheweb committed 2 years ago

Verified 173d0e64

ci : add cublas to windows release (#1271)

Green-Sky committed 2 years ago

Verified a3b85b28

Older