ochafik/llama.cpp

Pull Requests Commits

allow trivial stack advances (non lazy)

ochafik committed 1 year ago

7714cc85

only add pending char to stacks that will match them

ochafik committed 1 year ago

88c090a5

add asserts + stderr logs

ochafik committed 1 year ago

e69a77f2

log max stack size: seems to go in wrong direction for LAZY_STACKS=1 :-(

Olivier Chafik committed 1 year ago

6e9b715a

drop pending_chars after consuming them

Olivier Chafik committed 1 year ago

542a3851

BROKEN lazy stacks

ochafik committed 1 year ago

4d742329

Update llama-grammar.cpp

ochafik committed 1 year ago

3d659221

callback for llama_grammar_advance_stack

ochafik committed 1 year ago

30f213e8

Update llama-grammar.cpp

ochafik committed 1 year ago

a7233be7

ochafik committed 1 year ago

02e8f3d8

rename env to FASTTRACK

ochafik committed 1 year ago

0453734b

add DISABLE_FASTTRACK env var

ochafik committed 1 year ago

d9af10e7

add -jf / --json-schema-file flag

ochafik committed 1 year ago

0c90d967

use new llama_grammar_accept_candidate_for_stack when there's 1 candidate

ochafik committed 1 year ago

bd7739fb

prepare for less stacks / pending tokens

ochafik committed 1 year ago

ce5a116c

server : bump httplib to 0.19.0 (#11908)

ngxson committed 1 year ago

Verified 0f2bbe65

common : Fix a typo in help (#11899)

standby24x7 committed 1 year ago

Verified fe163d5b

ci : fix (again) arm64 build fails (#11895)

ngxson committed 1 year ago

Verified 818a340e

vulkan: support multi/vision rope, and noncontiguous rope (#11902)

jeffbolznv committed 1 year ago

Verified bf42a23d

metal : fix the crash caused by the lack of residency set support on Intel Macs. (#11904)

halechan committed 1 year ago

Verified c2ea16f2

scripts: fix compare-llama-bench commit hash logic (#11891)

JohannesGaessler committed 1 year ago

Verified 6dde1782

examples: fix typo in imatrix/README.md (#11884)

708-145 committed 1 year ago

Verified fc10c38d

metal : optimize dequant q6_K kernel (#11892)

akretz committed 1 year ago

Verified 22885105

readme : add notice about new package registry (#11890)

ggerganov committed 1 year ago

Verified c2cd24fb

repo : update links to new url (#11886)

ggerganov committed 1 year ago

Verified 68ff663a

server: fix type promotion typo causing crashes w/ --jinja w/o tools (#11880)

ochafik committed 1 year ago

Verified f3552296

vulkan: initial support for IQ1_S and IQ1_M quantizations (#11528)

remyoudompheng committed 1 year ago

Verified fc1b0d09

llguidance build fixes for Windows (#11664)

mmoskal committed 1 year ago

Verified 89daa256

opencl: Fix rope and softmax (#11833)

lhez committed 1 year ago

Verified 300907b2

cuda : add ampere to the list of default architectures (#11870)

slaren committed 1 year ago

Verified 94b87f87

Older