ggerganov/llama.cpp

Pull Requests Commits

docker : update CUDA images (#9213)

slaren committed 1 year ago

Verified 66b039a5

vulkan : fix build (#0)

ggerganov committed 1 year ago

20f1789d

ggerganov committed 1 year ago

231cff5f

Fix minicpm example directory (#9111)

xyb committed 1 year ago

Verified 3246fe84

llama : fix qs.n_attention_wv for DeepSeek-V2 (#9156)

compilade committed 1 year ago

Verified 78eb487b

server : add some missing env variables (#9116)

ngxson committed 1 year ago

Verified a77feb5d

llama : fix ChatGLM4 wrong shape (#9194)

RealJosephus committed 1 year ago

Verified 2e59d61c

llama : fix llama3.1 rope_freqs not respecting custom head_dim (#9141)

nyxkrage committed 1 year ago

Verified 75e1dbba

common : Update stb_image.h to latest version (#9161)

arch-btw committed 1 year ago

Verified ad76569f

ggml : do not crash when quantizing q4_x_x with an imatrix (#9192)

slaren committed 1 year ago

Verified 7d787ed9

metal : separate scale and mask from QKT in FA kernel (#9189)

ggerganov committed 1 year ago

Verified 06658ad7

ggml : add SSM Metal kernels (#8546)

ggerganov committed 1 year ago

Verified fc18425b

tests : fix compile warnings for unreachable code (#9185)

ggerganov committed 1 year ago

Verified 879275ac

ci : add VULKAN support to ggml-ci (#9055)

ggerganov committed 1 year ago

Verified 7a3df798

server : update deps (#9183)

ggerganov committed 1 year ago

Verified e5edb210

metal : gemma2 flash attention support (#9159)

slaren committed 1 year ago

Verified 0c41e03c

ggml-ci : try to improve build time (#9160)

slaren committed 1 year ago

Verified f12ceaca

llama : fix time complexity of string replacement (#9163)

jart committed 1 year ago

Verified 436787f1

common: fixed not working find argument --n-gpu-layers-draft (#9175)

GermanAizek committed 1 year ago

Verified 93bc3839

CUDA: fix Gemma 2 numerical issues for FA (#9166)

JohannesGaessler committed 1 year ago

Verified f91fc563

CPU/CUDA: Gemma 2 FlashAttention support (#8542)

JohannesGaessler committed 1 year ago

Verified e11bd856

quantize : fix typo in usage help of `quantize.cpp` (#9145)

joaodinissf committed 1 year ago

Verified 8f824ffe

lora : fix llama conversion script with ROPE_FREQS (#9117)

ngxson committed 1 year ago

Verified 3ba780e2

llama : use F32 precision in GLM4 attention and no FA (#9130)

piDack committed 1 year ago

Verified a07c32ea

Merge branch 'master' into compilade/bitnet-ternary

compilade committed 1 year ago

cb6d9962

[SYCL] Add a space to supress a cmake warning (#9133)

qnixsynapse committed 1 year ago

Verified 11b84eb4

[SYCL] Add oneDNN primitive support (#9091)

luoyu-intel committed 1 year ago

Verified 1731d423

llama : simplify Mamba with advanced batch splits (#8526)

compilade committed 1 year ago

Verified a1631e53

server : support reading arguments from environment variables (#9105)

ngxson committed 1 year ago

Verified fc54ef0d

llama : support for `falcon-mamba` architecture (#9074)

younesbelkada committed 1 year ago

Verified b40eb848

Newer Older