ggml-org/llama.cpp

Pull Requests Commits

`ggml_cuda_cpy` support for 4d tensors and float16->float32 upcasting (ggml/686)

balisujohn committed 1 year ago

Verified 625a699b

gguf : add input validation, prevent integer overflows (ggml/709)

ggerganov committed 1 year ago

Verified a4b07c05

ci : fix yolo URLs + fix metal capture (ggml/712)

ggerganov committed 1 year ago

Verified 549a1e6c

metal : add debug capture backend function (ggml/694)

jmousseau committed 1 year ago

Verified 5f14ee0b

Faster AVX2 dot product for IQ2_XS (#5187)

ikawrakow committed 1 year ago

Verified 8e14e3dd

SOTA 3-bit quants (#5196)

ikawrakow committed 1 year ago

Verified f4d7e549

Vulkan Windows APU Memory Handling (#5199)

0cc4m committed 1 year ago

Verified 2256f36b

quantize : fix typo (#5211)

first-leon committed 1 year ago

Verified 7359016c

main : allow empty --prompt-cache file (#5176)

divinity76 committed 1 year ago

Verified 81341699

readme : minor (#5204)

romainneutron committed 1 year ago

Verified 5589921e

readme : update hot topics

ggerganov committed 1 year ago

Verified 49f44b5c

server : improve README (#5209)

wujjpp committed 1 year ago

Verified 6685cc41

ggml alloc: Fix for null dereference on alloc failure (#5200)

ptsochantaris committed 1 year ago

Verified ceebbb5b

kompute : fix fallback to CPU (#5201)

cebtenzzre committed 1 year ago

Verified 6daa69ee

Nomic Vulkan backend (#4456)

cebtenzzre committed 1 year ago

Verified fbf1ddec

Merge branch 'gg/flash-attn' of https://github.com/ggerganov/llama.cpp into flash-attn-cuda

FSSRepo committed 1 year ago

7980178a

fix compiler error

FSSRepo committed 1 year ago

a1d5a12b

metal : faster inner loop for C == 32

ggerganov committed 1 year ago

Verified 5fcb9c1c

ggerganov committed 1 year ago

Verified c6c1132e

metal : disable buffer allocation logs

ggerganov committed 1 year ago

Verified abeaf0d9

fix typo "RLIMIT_MLOCK" (#5175)

divinity76 committed 1 year ago

Verified 2aed77eb

tests : add ATTN tests

ggerganov committed 1 year ago

Verified 4794821a

server : embeddings compatibility for OpenAI (#5190)

wujjpp committed 1 year ago

Verified c82d18e8

py : fix except (#5194)

ggerganov committed 1 year ago

Verified 14fef85e

py : improve BPE tokenizer support (#5189)

Sang-Kil Park committed 1 year ago

Verified e76627bc

ggml : add max buffer sizes to opencl and metal backends (#5181)

slaren committed 1 year ago

Verified fbe7dfa5

cmake : fix Vulkan build (#5182)

netrunnereve committed 1 year ago

Verified 172ac826

metal : support Q > 8

ggerganov committed 1 year ago

Verified 1db22d70

ggerganov committed 1 year ago

Verified 134c81c7

Merge branch 'master' into gg/flash-attn

ggerganov committed 1 year ago

Verified 0ad44baf

Newer Older