ggerganov/llama.cpp

Pull Requests Commits

convert : fix convert for refact models

ggerganov committed 1 year ago

Verified 03e940cd

ggml : require mask when using ALiBi

ggerganov committed 1 year ago

Verified 0faf92e7

vulkan : add dev notes

ggerganov committed 1 year ago

Verified 397b1f8f

ggml : fix assert message

ggerganov committed 1 year ago

Verified 536983b1

ggml : ggml_flash_attn_ext() support ALiBi (CUDA)

ggerganov committed 1 year ago

Verified 865af990

ggml : fix warning

ggerganov committed 1 year ago

Verified f7055d31

ggml : ggml_flash_attn_ext() support ALiBi (Metal)

ggerganov committed 1 year ago

Verified 97c27f59

ggml : ggml_flash_attn_ext() support ALiBi (CPU)

ggerganov committed 1 year ago

Verified 166e60bf

ggml : update ggml_soft_max_ext() CUDA, SYCL

ggerganov committed 1 year ago

Verified d0592d49

ggml : full ALiBi support

ggerganov committed 1 year ago

Verified 7fdca334

llava : fix moondream support (#7163)

abetlen committed 1 year ago

Verified d11afd66

Minor arithmetic improvement to mmvq wrapper kernel (#7172)

OuadiElfarouki committed 1 year ago

Verified 8c570c94

eval-callback : fix conversion to float (#7184)

slaren committed 1 year ago

Verified eaf4bd8b

Vulkan Bugfixes and Improvements (#7084)

0cc4m committed 1 year ago

Verified befddd0f

readme : add scheduled server workflow status badge

ggerganov committed 1 year ago

Verified d46dbc76

readme : add app (#6371)

l3utterfly committed 1 year ago

Verified 0961d866

llama3 custom regex split (#6965)

jaime-m-p committed 1 year ago

Verified 43248e55

CUDA: generalize FP16 fattn vec kernel (#7061)

JohannesGaessler committed 1 year ago

Verified a743d76a

Add warning if token is invalid (#7173)

Galunid committed 1 year ago

Verified f31ec120

llama : update llama_timings.n_p_eval setting (#7160)

danbev committed 1 year ago

Verified fd9f92b1

gguf-py : add special token modification capability (#7166)

CISC committed 1 year ago

Verified 22842164

opencl : alignment size converted from bits to bytes (#7090)

albertjin committed 1 year ago

Verified 47345248

TypoFix (#7162)

AhmedZeer committed 1 year ago

Verified 07cd41d0

cmake : fix typo (#7151)

cebtenzzre committed 1 year ago

Verified 4426e298

convert-hf : save memory with lazy evaluation (#7075)

compilade committed 1 year ago

Verified f98eb31c

Introduction of CUDA Graphs to LLama.cpp (#6766)

agray3 committed 1 year ago

Verified bc4bba36

JSON: [key] -> .at(key), assert() -> GGML_ASSERT (#7143)

JohannesGaessler committed 1 year ago

Verified c12452c7

Revert "llava : add support for moondream vision language model (#6899)"

ggerganov committed 1 year ago

Verified 9da243b3

server : add themes + favicon (#6848)

jboero committed 1 year ago

Verified bd1871fa

metal : use `vm_allocate` instead of `posix_memalign` on macOS (#7078)

giladgd committed 1 year ago

Verified 26458af1

Older