ggerganov/llama.cpp

Pull Requests Commits

convert-hf : add missing ftype to Baichuan and Xverse

compilade committed 2 years ago

65a1a585

convert-hf: add missing ftype

compilade committed 2 years ago

2b1e5ea3

convert-hf : support q8_0 conversion

compilade committed 2 years ago

d7e199e4

convert-hf : support bfloat16 conversion (#7158)

compilade committed 2 years ago

Verified 5a419926

ggerganov committed 2 years ago

fae9d234

feat: implemented sigmoid function (ggml/806)

justcho5 committed 2 years ago

f5ef34e4

build: fix and ignore msvc warnings (ggml/805)

iboB committed 2 years ago

ef0d5e3e

convert : skip unaccessible HF repos (#7210)

CrispStrobe committed 2 years ago

Verified 3292733f

server : free llama_batch on exit (#7212)

stevegrubb committed 2 years ago

Verified 98863133

llama : lookup word in vocab before doing BPE merges (#7193)

tonyfettes committed 2 years ago

Verified f99e1e45

server: fix reported top tokens for temperature 0 (#7203)

JohannesGaessler committed 2 years ago

Verified 5ae3426b

llama : add Jina Embeddings architecture (#6826)

JoanFM committed 2 years ago

Verified b83cc3f5

ggml : full ALiBi support (#7192)

ggerganov committed 2 years ago

Verified 9cb317f7

llama-bench : add pp+tg test type (#7199)

slaren committed 2 years ago

Verified e8496488

metal : fix flash attention kernel requirements (#7169)

ggerganov committed 2 years ago

Verified 18e43766

convert : print "ignore_merges" field

ggerganov committed 2 years ago

Verified 8c660242

llama : use n_vocab to differentiate between mistral 7B and llama3 8B (#7200)

slaren committed 2 years ago

Verified 25c6e82e

Fix memory bug in grammar parser (#7194)

jart committed 2 years ago

Verified 4e388097

Main+: optionally allow special tokens from user in interactive mode (#7097)

hanishkvc committed 2 years ago

Verified f89fe273

llava : fix moondream support (#7163)

abetlen committed 2 years ago

Verified d11afd66

Minor arithmetic improvement to mmvq wrapper kernel (#7172)

OuadiElfarouki committed 2 years ago

Verified 8c570c94

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

slaren committed 2 years ago

Verified eaf4bd8b

Vulkan Bugfixes and Improvements (#7084)

0cc4m committed 2 years ago

Verified befddd0f

readme : add scheduled server workflow status badge

ggerganov committed 2 years ago

Verified d46dbc76

readme : add app (#6371)

l3utterfly committed 2 years ago

Verified 0961d866

llama3 custom regex split (#6965)

jaime-m-p committed 2 years ago

Verified 43248e55

CUDA: generalize FP16 fattn vec kernel (#7061)

JohannesGaessler committed 2 years ago

Verified a743d76a

Add warning if token is invalid (#7173)

Galunid committed 2 years ago

Verified f31ec120

llama : update llama_timings.n_p_eval setting (#7160)

danbev committed 2 years ago

Verified fd9f92b1

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

CISC committed 2 years ago

Verified 22842164

Older