ggerganov/llama.cpp

Pull Requests Commits

ggml : relax contiguous constraints in activation function (#2371)

li-plus committed 2 years ago

Verified 875086bd

ggml : improve graph build time via hash table lookup (#2329)

slaren committed 2 years ago

Verified da188983

build : fix line breaking error in build-info.sh (#2349)

hesenp committed 2 years ago

Verified 82552b7f

main : add `--in-prefix-bos` to prefix BOS to user inputs; keep EOS (#2304)

jxy committed 2 years ago

Verified 0c06204f

ci : add non-AVX scalar build/test (#2356)

netrunnereve committed 2 years ago

Verified 1fed755b

k_quants : add AVX support to dot functions with QK_K as 64 (#2339)

katsu560 committed 2 years ago

Verified be2301bc

metal : concurrently dispatch commands (#2358)

lshzh-ww committed 2 years ago

Verified 1aa18ef9

Another speed gain for Q4_0 and Q4_1 on Metal (#2375)

ikawrakow committed 2 years ago

Verified 9a08eaf3

Fix Q4_K and Q5_K for QK_K = 64 on CUDA (#2359)

ikawrakow committed 2 years ago

Verified 129d844c

server: add rms_norm_eps parameter (#2380)

slaren committed 2 years ago

Verified d5512b78

[Server] Escape HTML in webchat (#2368)

SlyEcho committed 2 years ago

Verified c798308e

make rms_norm_eps a parameter (#2374)

slaren committed 2 years ago

Verified 41c67416

Chat UI extras (#2366)

akx committed 2 years ago

Verified b3f138d0

ggml : sync (unary ops refactor, static-correctness) (#2370)

ggerganov committed 2 years ago

Verified 5b2b2dc6

Fix scalar version of Q5_K when QK_K = 64 (#2362)

ikawrakow committed 2 years ago

Verified 42f70cb2

llama : add grammar-based sampling (#1773)

ejones committed 2 years ago

Verified 84e09a7d

Some more Q4_K and Q5_K speedup on CUDA (#2346)

ikawrakow committed 2 years ago

Verified 2f9cf974

Add gqa parameter support to the server (#2351)

IgnacioFDM committed 2 years ago

Verified 4f06592c

Fix __dp4a documentation (#2348)

JohannesGaessler committed 2 years ago

Verified 70d26ac3

common : n_threads == -1 uses std::thread::hardware_concurrency() (#2347)

Freed-Wu committed 2 years ago

Verified 57921ca6

fix n_tasks (#2342)

slaren committed 2 years ago

Verified 3602ac42

ggml: move op parameters from tensors to ggml_tensor::op_params (#2333)

slaren committed 2 years ago

Verified 95a6c595

llama : grouped-query attention + LLaMAv2 70B support (#2276)

ggerganov committed 2 years ago

Verified e76d630d

llama : print help to stdout (#2338)

maddes8cht committed 2 years ago

Verified 1d0824b2

flake : support `nix build '.#opencl'` (#2337)

Freed-Wu committed 2 years ago

Verified bc3ec2cd

llama : print max tensor size to stderr (#2336)

Christian Demsar committed 2 years ago

Verified a940458e

make : fix CLBLAST compile support in FreeBSD (#2331)

yukiteruamano committed 2 years ago

Verified 91171b80

examples : simplify vim plugin (#2327)

AustinMroz committed 2 years ago

Verified 355c80f4

metal : support bcast add & dup & cont op (#2323)

li-plus committed 2 years ago

Verified 83a00ce6

Speed up Q4_K (#2322)

ikawrakow committed 2 years ago

Verified d2a43664

Newer Older