ggml-org/llama.cpp

Pull Requests Commits

metal : try to improve batched decoding

ggerganov committed 1 year ago

Verified 3cb1c348

llama : pad KV cache size to 32

ggerganov committed 1 year ago

Verified 3e68df86

llama : avoid using "optional" keyword (#4283)

ggerganov committed 1 year ago

Verified 5a7d3125

llama : support optional tensors (#4283)

ggerganov committed 1 year ago

Verified d5a1cbde

swift : fix token_to_piece implementation (#4278)

ensan-hcl committed 1 year ago

Verified b220222a

build : enable libstdc++ assertions for debug builds (#4275)

cebtenzzre committed 1 year ago

Verified 511f52c3

llama : support attention bias on LLaMA architecture (#4283)

RealJosephus committed 1 year ago

Verified 03562f3a

llama : add Qwen support (#4281)

simonJJJ committed 1 year ago

Verified 37c746d6

llama : fix integer overflow during quantization (#4284)

ggerganov committed 1 year ago

Verified 880f5797

py : add requirements file for convert-hf-to-gguf.py (#4277)

danbev committed 1 year ago

Verified 8d6d9f03

ggml : add ggml_soft_max_ext (#4256)

ggerganov committed 1 year ago

Verified ef47ec18

server : add --log-disable to disable logging to file (#4260)

ziedbha committed 1 year ago

Verified 1d144112

server : add single-client multi-prompt support (#4232)

ziedbha committed 1 year ago

Verified f43f0936

make : fix Apple clang determination bug (#4272)

WillCorticesAI committed 1 year ago

Verified d2809a3b

build : fix build info generation and cleanup Makefile (#3920)

cebtenzzre committed 1 year ago

Verified 15f5d960

llava : ShareGPT4V compatibility (vision encoder only loading) (#4172)

cmp-nct committed 1 year ago

Verified 33c9892a

main : pass LOG_TEE callback to llama.cpp log (#4033)

AndrewGodfrey committed 1 year ago

Verified 8efa0f6e

readme : fix (#4135)

vodkaslime committed 1 year ago

Verified 524907aa

docker : add finetune option (#4211)

jooray committed 1 year ago

Verified 3bd2c7ce

batched.swift : update README.md (#4214)

ensan-hcl committed 1 year ago

Verified bde629bb

cmake : fix the metal file foder path (#4217)

tanliboy committed 1 year ago

Verified f7f9e062

readme : fix typo (#4253)

TortillaZHawaii committed 1 year ago

Verified 74daabae

llama : fix alignment of general.name in print meta (#4254)

danbev committed 1 year ago

Verified b18c66ca

convert.py : fix llama/llama2 conversion due to vocab_size=-1 (#4258)

slaren committed 1 year ago

Verified f4d973ce

llama : fix typical sampling (#4261)

tarcey committed 1 year ago

Verified 954e2285

py : fix oai proxy (#3972)

rhjdvsgsgks committed 1 year ago

Verified e2bd725f

examples : add readme files

ggerganov committed 1 year ago

Verified 1f5cd832

readme : add FreeChat (#4248)

psugihara committed 1 year ago

Verified 4fea3420

ggml : restore abort() in GGML_ASSERT (#4242)

cebtenzzre committed 1 year ago

Verified 64e64aa2

ggml : re-enable BLAS for CPU when src0 != F32 + remove redundant full offload checks in llama.cpp (#4240)

ggerganov committed 1 year ago

Verified 8406b092

Older