ggerganov/llama.cpp

Pull Requests Commits

metal : try to utilize more of the shared memory using smaller views

ggerganov committed 1 year ago

Verified 5cc672a9

ggml : avoid conv 2d kernel round up

ggerganov committed 1 year ago

Verified c824d2e3

ggml : add NUMA support (#1556)

zrm committed 1 year ago

Verified b853d456

k-quants : fix indentation

ggerganov committed 1 year ago

Verified 9225baef

tests : fix quantize perf (#1990)

katsu560 committed 1 year ago

Verified a84ab1da

k-quants : add AVX support to dot functions (#1916)

katsu560 committed 1 year ago

Verified 5743ca80

readme : add link to new k-quants for visibility

ggerganov committed 1 year ago

Verified 412c60e4

k-quants : support for super-block size of 64 (#2001)

ikawrakow committed 1 year ago

Verified 6769e944

Fix assert when free invalid cuda pointer (#2005)

howard0su committed 1 year ago

Verified cbebf61c

readme : add new roadmap + manifesto

ggerganov committed 1 year ago

Verified 447ccbe8

ggml : sync latest ggml (custom operators)

ggerganov committed 1 year ago

Verified bd34cdde

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

anon998 committed 1 year ago

Verified c2a08f87

readme : add Azure CI discussion link

ggerganov committed 1 year ago

Verified 66a2555b

zig : upgrade build system support (#1981)

coderonion committed 1 year ago

Verified e65ca7e1

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

robyngraf committed 1 year ago

Verified 5ec8dd5a

tests : sync test-grad0 from ggml

ggerganov committed 1 year ago

Verified 65bdd52a

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

novafacing committed 1 year ago

Verified fdd18609

convert : fix invalid params in write_vocab_only (#1975)

aisk committed 1 year ago

Verified c943d823

ggml : improve ggml_graph_dump_dot, add ggml_format_name (#1978)

slaren committed 1 year ago

Verified f2c754e1

readme : fix whitespaces

ggerganov committed 1 year ago

Verified 11da1a85

readme : fixed termux instructions (#1973)

albbus-stack committed 1 year ago

Verified 235b610d

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

alexrenda committed 1 year ago

Verified b061ba9e

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

didzis committed 1 year ago

Verified 527b6fba

Add OpenLLaMA instructions to the README (#1954)

eiery committed 1 year ago

Verified d7b7484f

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

Green-Sky committed 1 year ago

Verified 74871372

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

JohannesGaessler committed 1 year ago

Verified bbca06e2

Fix typo in README.md (#1961)

RahulVivekNair committed 1 year ago

Verified fb98254f

readme : add link to p1

ggerganov committed 1 year ago

Verified 049aa16b

Fix typo (#1949)

sammysun0711 committed 1 year ago

Verified 2322ec22

llama : fix params struct slignment (#1936)

mudler committed 1 year ago

Verified aacdbd40

Older