ochafik/llama.cpp

Pull Requests Commits

grammars: explore allowing disabling grammars to avoid fixed cost of token copies

ochafik committed 1 year ago

988fc66f

Update llama.cpp

ochafik committed 1 year ago

Verified 80736c55

Merge remote-tracking branch 'origin/master' into grammar-fast

ochafik committed 1 year ago

0148661a

metal : log more info on error (#6987)

bakkot committed 1 year ago

Verified a68a1e7e

ggml : add Flash Attention (#5021)

ggerganov committed 1 year ago

Verified 9c67c277

convert : use utf8 encoding (#7000)

ggerganov committed 1 year ago

Verified 952d03db

Improve usability of --model-url & related flags (#6930)

ochafik committed 1 year ago

Verified 8843a98c

Extending grammar integration tests (#6644)

HanClinto committed 1 year ago

Verified b8c1476e

main : fix typo in comment in main.cpp (#6985)

danbev committed 1 year ago

Verified 5539e6fd

build(cmake): simplify instructions (`cmake -B build && cmake --build build ...`) (#6964)

ochafik committed 1 year ago

Verified b8a7a5a9

ci : tmp disable gguf-split (#6983)

ggerganov committed 1 year ago

Verified d2c898f7

ggml : fix __MSC_VER -> _MSC_VER (#6977)

ggerganov committed 1 year ago

Verified 544f1f10

llava-cli : multiple images (#6969)

cpumaxx committed 1 year ago

Verified ffe66657

readme : update hot topics

ggerganov committed 1 year ago

Verified 24affa7d

llama : fix BPE pre-tokenization (#6920)

ggerganov committed 1 year ago

Verified f4ab2a41

sampling : use std::random_device{}() for default random seed (#6962)

dwrensha committed 1 year ago

Verified 3f167476

convert : fix conversion of some BERT embedding models (#6937)

christianazinn committed 1 year ago

Verified 3055a418

make : change GNU make default CXX from g++ to c++ (#6966)

przemoc committed 1 year ago

Verified 577277ff

ci : add building in MSYS2 environments (Windows) (#6967)

przemoc committed 1 year ago

Verified ca7f29f5

llama : fix typo LAMMAFILE -> LLAMAFILE (#6974)

JohannesGaessler committed 1 year ago

Verified c4f708a9

Fix more int overflow during quant (PPL/CUDA). (#6563)

dranger003 committed 1 year ago

Verified e00b4a8f

gguf : enforce that tensor names are unique (#6905)

ngxson committed 1 year ago

Verified 7bb36ccf

add device version in device list (#6959)

arthw committed 1 year ago

Verified ce023f6f

grammars: nits (revert const grammar sig, fix comment)

ochafik committed 1 year ago

c70037f2

grammar: nit (layout)

ochafik committed 1 year ago

d3425f5c

grammars: cache codepoints in llama_new_context_with_model

ochafik committed 1 year ago

49f0faaa

grammars: move token caches to llama_context

ochafik committed 1 year ago

d41f3147

flake.lock: Update

github-actions[bot] committed 1 year ago

6e472f58

Merge remote-tracking branch 'origin/master' into grammar-fast

ochafik committed 1 year ago

9f136231

Replace "alternative" boolean operator in conditional compilation directive (#6949)

mgroeber9110 committed 1 year ago

Verified 4dba7e81

Older