ggml-org/llama.cpp

Pull Requests Commits

ggml : do not use _GNU_SOURCE gratuitously

ggerganov committed 2 years ago

Verified 78fafcaf

readme : add new roadmap + manifesto

ggerganov committed 2 years ago

Verified 447ccbe8

ggml : sync latest ggml (custom operators)

ggerganov committed 2 years ago

Verified bd34cdde

fix server sampling: top k sampler first (#1977)

anon998 committed 2 years ago

Verified c2a08f87

readme : add Azure CI discussion link

ggerganov committed 2 years ago

Verified 66a2555b

zig : upgrade build system support (#1981)

coderonion committed 2 years ago

Verified e65ca7e1

#1869 Fix null reference errors when training from scratch with CUDA (#1907)

robyngraf committed 2 years ago

Verified 5ec8dd5a

tests : sync test-grad0 from ggml

ggerganov committed 2 years ago

Verified 65bdd52a

flake : fix ggml-metal.metal path and run nixfmt (#1974)

novafacing committed 2 years ago

Verified fdd18609

convert : fix invalid params in write_vocab_only (#1975)

aisk committed 2 years ago

Verified c943d823

ggml : improve ggml_graph_dump_dot, add ggml_format_name (#1978)

slaren committed 2 years ago

Verified f2c754e1

readme : fix whitespaces

ggerganov committed 2 years ago

Verified 11da1a85

readme : fixed termux instructions (#1973)

albbus-stack committed 2 years ago

Verified 235b610d

llama : fix top-p sampling to match the canonical definition (#1953)

alexrenda committed 2 years ago

Verified b061ba9e

llama : make model stateless and context stateful (llama_state) (#1797)

didzis committed 2 years ago

Verified 527b6fba

Add OpenLLaMA instructions to the README (#1954)

eiery committed 2 years ago

Verified d7b7484f

rework convert.py to read hyper-parameters from config.json (#1958)

Green-Sky committed 2 years ago

Verified 74871372

cmake: revert CUDA arch default to 52, 61 if f16 (#1959)

JohannesGaessler committed 2 years ago

Verified bbca06e2

Fix typo in README.md (#1961)

RahulVivekNair committed 2 years ago

Verified fb98254f

readme : add link to p1

ggerganov committed 2 years ago

Verified 049aa16b

Fix typo (#1949)

Xiake Sun committed 2 years ago

Verified 2322ec22

llama : fix params struct slignment (#1936)

mudler committed 2 years ago

Verified aacdbd40

[Fix] Reenable server embedding endpoint (#1937)

SlyEcho committed 2 years ago

Verified 20568fe6

ggml : fix bug in LBFGS optimizer (found by ggml tests)

ggerganov committed 2 years ago

Verified 18b35625

llama : use aligned memory during ggml_init call from loading saved sessions (#1934)

l3utterfly committed 2 years ago

Verified ba4e85a8

cmake : fix trailing whitespaces

ggerganov committed 2 years ago

Verified 23fc5c21

llama : only use Q6_K for output weights if tensor size is multiple of 256 (#1932)

ikawrakow committed 2 years ago

Verified cb40dfca

cuda : faster k-quants on older GPUs (#1930)

ikawrakow committed 2 years ago

Verified ca7c3f4d

ggml : sync latest ggml repo (#1924)

ggerganov committed 2 years ago

Verified b97ca431

cmake : fix build shared ggml when CUDA is enabled (#1929)

howard0su committed 2 years ago

Verified 1e3abfce

Older