ochafik/llama.cpp

Pull Requests Commits

add comment on limits to common_grammar_trigger.to/from json specializations

Olivier Chafik committed 1 year ago

71719a6e

avoid ggml_assert in server for grammar triggers inconsistency

Olivier Chafik committed 1 year ago

24010fe7

Merge remote-tracking branch 'origin/master' into tool-bench-prod

Olivier Chafik committed 1 year ago

00db4651

ci : fix xcframework artifact tag (#12191)

danbev committed 1 year ago

Verified fa31c438

ci : remove xframework upload (#12190)

danbev committed 1 year ago

Verified 3ccbfe5a

server : fix cache reuse logic (#12161)

Clauszy committed 1 year ago

Verified 06a92a19

llama : add xcframework build script (#11996)

danbev committed 1 year ago

Verified a057897a

Update src/llama-grammar.cpp

ochafik committed 1 year ago

Verified 01be080e

fix nit eol at eof

ochafik committed 1 year ago

92e9723d

disable slow hello Llama-3.1-8B (chopped unescaped string witin string problem)

ochafik committed 1 year ago

20a2f5f8

avoid returning optional from parse_json

ochafik committed 1 year ago

a6d78873

fix crashing typo

ochafik committed 1 year ago

ad3caa34

add common_grammar_trigger.{to_json,from_json}

ochafik committed 1 year ago

1317a35f

add llama_grammar_trigger_pattern

Olivier Chafik committed 1 year ago

5d43b726

common_grammar_trigger: always use string value (+ optional token)

ochafik committed 1 year ago

e6e9c138

Merge remote-tracking branch 'origin/master' into tool-bench-prod

Olivier Chafik committed 1 year ago

2470a1c1

ggml : portability fixes for VS 2017 (#12150)

mgroeber9110 committed 1 year ago

Verified 5bbe6a9f

readme : fix roadmap link (#12185)

ggerganov committed 1 year ago

Verified 20a9b8f5

main: allow preloading conversation with -p and add -st / --single-turn (#12145)

CISC committed 1 year ago

Verified 56d7a9f8

`server`: fix deadly typo in response_format.json_schema.schema handling (#12168)

ochafik committed 1 year ago

Verified 1a24c462

HIP: implement FlashAttention via rocWMMA for CDNA and RDNA3+ (#12032)

hjc4869 committed 1 year ago

Verified becade5d

ggerganov committed 1 year ago

dfd6b2c0

cuda: unary ops as float + de-duplicate (ggml/1130)

cmdr2 committed 1 year ago

b64d7cc2

ggerganov committed 1 year ago

3d1cf3cf

cuda/vulkan: specify fp32-only support for some operations in supports_op (ggml/1129)

cmdr2 committed 1 year ago

0cbee131

ggerganov committed 1 year ago

8371d445

cuda/cpu: Increase support for fp16 unary operations (ggml/1125)

cmdr2 committed 1 year ago

87abb7e9

whisper : support GGML_BACKEND_DL (whisper/2843)

slaren committed 1 year ago

6d4c23b8

cmake : fix compile assumptions for power9/etc (whisper/2777)

midnightmagic committed 1 year ago

6512a900

Told cmake to install ggml-cpp.h as a public header file. (ggml/1126)

petterreinholdtsen committed 1 year ago

45120557

Older