ochafik/llama.cpp

Pull Requests Commits

Update sampling.cpp

ochafik committed 1 year ago

10e3ad59

grammars: revert

ochafik committed 1 year ago

9912ff6f

grammars: define LLAMA_SKIP_EMPTY_REJECTS

ochafik committed 1 year ago

b93be200

grammars: revert more non-resample code

ochafik committed 1 year ago

80c15d07

grammars: revert llama_grammar_element.token_candidates

ochafik committed 1 year ago

fa0cbf50

grammars: disable buggy grammar candidate stubbing logic

Olivier Chafik committed 1 year ago

ca6de33c

grammars: assert original_logics if grammar & !apply_grammar

Olivier Chafik committed 1 year ago

b3bf5f5e

grammars: early exit from https://github.com/ggerganov/llama.cpp/pull/7370

Olivier Chafik committed 1 year ago

5d600599

Update sampling.cpp

Olivier Chafik committed 1 year ago

31cc6df9

Update main.cpp

Olivier Chafik committed 1 year ago

0a2ce59d

grammars: fix eog

Olivier Chafik committed 1 year ago

ae253706

grammars: stub candidates from grammar's (cached) potential next tokens

Olivier Chafik committed 1 year ago

f983d93f

grammars: fix original_logits logic

Olivier Chafik committed 1 year ago

05d0d52b

grammars: switch is_resampling logic around

Olivier Chafik committed 1 year ago

56e29f7c

ci : re-enable sanitizer runs (#7358)

ggerganov committed 1 year ago

Verified 059031b8

android : use "ci-android" branch for CI (#7341)

ggerganov committed 1 year ago

Verified 511182ea

CUDA: deduplicate FlashAttention code (#7352)

JohannesGaessler committed 1 year ago

Verified 133d99c5

server: correct --threads documentation [no ci] (#7362)

JohannesGaessler committed 1 year ago

Verified cb42c294

cuda : add half2 __shfl_xor() for ROCm 5.5 (#7263)

Engininja2 committed 1 year ago

Verified d233b507

llama : add support for larger Granite Code Models (20B, 34B) (#7324)

sroecker committed 1 year ago

Verified 0f98acfa

perplexity : ndot progress and show stats with < 100 tasks (#7348)

strawberrymelonpanda committed 1 year ago

Verified ca57e0f3

Update and fix Vulkan soft_max and argsort implementations (#7237)

0cc4m committed 1 year ago

Verified c1b295ee

github-actions-labeler: initial commit (#7330)

mofosyne committed 1 year ago

Verified de731963

convert : fix set_vocab_sentencepiece (#6866)

ggerganov committed 1 year ago

Verified b49a13dd

ggml : fix quants nans when all the group weights are very close to zero (#7313)

slaren committed 1 year ago

Verified 05834841

cmake : fix typo in AMDGPU_TARGETS (#7356)

Engininja2 committed 1 year ago

Verified ef277de2

Unicode codepoint flags for custom regexs (#7245)

jaime-m-p committed 1 year ago

Verified b43272af

CUDA: faster large batch FA without tensor cores (#7314)

JohannesGaessler committed 1 year ago

Verified 0fc1e820

ROCm: use native CMake HIP support (#5966)

GZGavinZhao committed 1 year ago

Verified 82ca83db

rpc : set SO_REUSEADDR for the server socket (#7320)

rgerganov committed 1 year ago

Verified f4bd8b3d

Older