ggml-org/llama.cpp

Pull Requests Commits

make : fix speculative build

ggerganov committed 2 years ago

Verified c79d130f

speculative : avoid grammar_mem

ggerganov committed 2 years ago

Verified 2db2471c

speculative : reuse grammar parser + better logs and comments

ggerganov committed 2 years ago

Verified e7dc5b08

speculative : print draft token pieces

ggerganov committed 2 years ago

Verified 6c150d76

common : warm-up with 2 tokens - seems to work better

ggerganov committed 2 years ago

Verified ebe41d49

grammar : remove one nested level

ggerganov committed 2 years ago

Verified 01345788

grammar : add comments to new grammar file

ggerganov committed 2 years ago

Verified 2d89da4f

grammars : add json_arr.gbnf

ggerganov committed 2 years ago

Verified e0a8658e

speculative : add grammar support

ggerganov committed 2 years ago

Verified 69f2fafe

llama-bench : make cpp file non-executable (#2999)

cebtenzzre committed 2 years ago

Verified 31035681

make : add speculative example (#3003)

leng-yue committed 2 years ago

Verified 5b8530d8

server : add a subtle loading animation to the edit box (#2466)

akx committed 2 years ago

Verified e4386f41

2x faster (rms) norm cuda kernels (3.7% e2e improvement) (#2985)

li-plus committed 2 years ago

Verified 35195689

ggml-alloc : use virtual memory for measurement (#2973)

slaren committed 2 years ago

Verified cf9b0848

speculative : PoC for speeding-up inference via speculative sampling (#2926)

ggerganov committed 2 years ago

Verified 47068e51

perplexity : fix ETA by warming up the model with an empty run

ggerganov committed 2 years ago

Verified 8f429fa5

gguf(python): Fix special vocab handling when id < 0 (#2984)

KerfuffleV2 committed 2 years ago

Verified 6519e9c9

metal : restore 363f0bf and fix reduce in F16_F32 kernels (#2986)

ggerganov committed 2 years ago

Verified b7f2aa9e

cov : disable comment in PRs (#2989)

alonfaraj committed 2 years ago

Verified 73a12a63

llama : fix bpe tokenize from byte (#2889)

opparco committed 2 years ago

Verified 37301347

metal : revert 6af0bab until we fix it

ggerganov committed 2 years ago

Verified d9151e6f

cov : add Code Coverage and codecov.io integration (#2928)

alonfaraj committed 2 years ago

Verified afc43d5f

opencl : fix a bug in ggml_cl_pool_malloc() for ggml_cl_mul_mat_f32() (#2955)

rchardx committed 2 years ago

Verified 6460f758

metal : more optimizations (#2959)

ikawrakow committed 2 years ago

Verified ca82cf7b

swift : add support for k-quants (#2983)

kchro3 committed 2 years ago

Verified 6a31a3bd

convert.py : BPE fixes (#2938)

KerfuffleV2 committed 2 years ago

Verified cff7b0bf

docs : add `catai` to `README.md` (#2967)

ido-pluto committed 2 years ago

Verified 340af42f

examples : fix gpt-neox (#2943)

mmnga committed 2 years ago

Verified c42f0ec6

swift : add missing c file to Package.swift (#2978)

kchro3 committed 2 years ago

Verified 2753415a

make : support overriding CFLAGS/CXXFLAGS/CPPFLAGS/LDFLAGS (#2886)

cebtenzzre committed 2 years ago

Verified bc054af9

Older