ggerganov/llama.cpp

Pull Requests Commits

Merge branch 'master' into compilade/lazy-convert-hf

compilade committed 2 years ago

bffdaf40

main : add --conversation / -cnv flag (#7108)

Dawid Potocki committed 2 years ago

Verified 83330d8c

sgemm : AVX Q4_0 and Q8_0 (#6891)

netrunnereve committed 2 years ago

Verified 465263d0

server : add_special option for tokenize endpoint (#7059)

JohanAR committed 2 years ago

Verified 911b3900

convert.py : --vocab-only generates false but valid params (#7027)

20kdc committed 2 years ago

Verified ad211ede

llama : add BPE pre-tokenization for Qwen2 (#7114)

jklj077 committed 2 years ago

Verified 229ffff8

clean up json_value & server_log (#7142)

ngxson committed 2 years ago

Verified 1fd9c174

convert : add BPE pre-tokenization for DBRX (#7132)

dranger003 committed 2 years ago

Verified 4cd621c2

py : also print the normalizers

ggerganov committed 2 years ago

Verified 7e0b6a7b

compare-llama-bench.py: add missing basicConfig (#7138)

mofosyne committed 2 years ago

Verified acdce3cd

ggml : introduce bfloat16 support (#6412)

jart committed 2 years ago

Verified 38554160

metal : fix unused warning

ggerganov committed 2 years ago

Verified c0e6fbf8

Further tidy on Android instructions README.md (#7077)

Jeximo committed 2 years ago

Verified c780e753

Fixed save_imatrix to match old behaviour for MoE (#7099)

jukofyork committed 2 years ago

Verified 48b2f9c1

server: fix incorrectly reported token probabilities (#7125)

JohannesGaessler committed 2 years ago

Verified af0a5b61

Fix OLMo HF to GGUF conversion (#6910)

nopperl committed 2 years ago

Verified b6aa6702

server : update readme with undocumented options (#7013)

K-Mistele committed 2 years ago

Verified 260b7c65

readme : update hot topics

ggerganov committed 2 years ago

Verified 53d6c52e

main : update log text (EOS to EOG) (#7104)

RhinoDevel committed 2 years ago

Verified 3af34c1d

docs: fix typos (#7124)

omahs committed 2 years ago

Verified 04976db7

ci : add GG_BUILD_EXTRA_TESTS_0 env (#7098)

ggerganov committed 2 years ago

Verified 947d3ad2

Add an option to build without CUDA VMM (#7067)

WilliamTambellini committed 2 years ago

Verified 858f6b73

flake.lock: Update (#7079)

ggerganov committed 2 years ago

Verified b3a995b4

gguf-py : add tqdm as a dependency

compilade committed 2 years ago

94e667a9

Merge branch 'master' into compilade/lazy-convert-hf

compilade committed 2 years ago

68c5ac62

minor : fix trailing whitespace

ggerganov committed 2 years ago

Verified bcdee0da

convert-hf : minor changes for consistency

compilade committed 2 years ago

62303e7f

convert-hf : faster model parts loading

compilade committed 2 years ago

bc78bf4c

Adding support for the --numa argument for llama-bench. (#7080)

kunnis committed 2 years ago

Verified 628b2991

Disable benchmark on forked repo (#7034)

CISC committed 2 years ago

Verified 8f8acc86

Older