ggerganov/llama.cpp

Pull Requests Commits

quant : manual overrides of tensor types take precedence

ggerganov committed 65 days ago

Verified 8b407e39

convert : use n_groups instead of hardcoded values in reshape (#18929)

danbev committed 65 days ago

Verified 7dee9ff5

server : refactor oai_parser_opt, move it to server_chat_params (#18937)

ngxson committed 66 days ago

Verified 6df686be

convert : support Glm4MoeLite (#18936)

ddh0 committed 66 days ago

Verified 1706a6d7

jinja : fix undefined keys and attributes and int/float as bool (#18924)

CISC committed 66 days ago

Verified 959ecf7f

ci : run test-jinja -py on high perf [no ci] (#18916)

CISC committed 66 days ago

Verified 4037093c

server: fix memory reservations in populate_token_probs (#18787)

l-austenfeld committed 66 days ago

Verified 18361c57

ggml : add ggml_build_forward_select (#18550)

ggerganov committed 66 days ago

Verified 365a3e8c

model-conversion : add BUILD_DIR variable to run-converted-model scripts (#18927)

danbev committed 66 days ago

Verified 3d55846a

llama : Extend fallback, fix fileno for dio file, exclude case that mmap uses dio file (#18887)

JTischbein committed 67 days ago

Verified 287a3301

docs: add linux to index (#18907)

alosslessdev committed 67 days ago

Verified 293a1565

tests : add test-jinja -py option for cross-checking (#18906)

ngxson committed 67 days ago

Verified fe44d355

jinja : fix object item order (and properly implement dictsort) (#18904)

CISC committed 68 days ago

Verified bbcdac01

jinja : attribute support for join, map and sort (#18883)

CISC committed 68 days ago

Verified d03c45c9

jinja : add missing tojson filter for bool (#18900)

CISC committed 68 days ago

Verified 10c98cbd

jinja : fix lexing of float literals with sign (#18901)

CISC committed 68 days ago

Verified 420960ab

jinja: correct member access rule (#18905)

ngxson committed 68 days ago

Verified f55b033a

opencl: fix q6_K mv for m=1 (#18893)

lhez committed 68 days ago

Verified d1b4757d

ci : add label for jinja changes (#18903)

CISC committed 68 days ago

Verified 57c0beae

kv-cache : optimize KQ mask construction (#18842)

ggerganov committed 68 days ago

Verified 2fbde785

ggml webgpu: support for backend sampling (#18880)

reeselevine committed 69 days ago

Verified a89002f0

ggml : extend ggml_pool_1d + metal (#16429)

ThoreKoritzius committed 69 days ago

Verified 388ce822

docs : update ops.md for CANN backend (#18654)

hipudding committed 69 days ago

Verified 6ba6a3c7

ggml-blas: hide warnings from included BLAS headers (#18818)

DaAwesomeP committed 69 days ago

Verified 0802d4cf

mtmd : Fix ASR for LFM2.5-Audio-1.5B (#18876)

tdakhran committed 69 days ago

Verified c945aaae

common : implement new jinja template engine (#18462)

ngxson committed 69 days ago

Verified c15395f7

Setting mmap and direct_io to false as default in llama-bench.cpp (#18841)

JTischbein committed 69 days ago

Verified aa1dc377

CANN: Remove unused `ggml_cann_get_device` function (#18625)

rauletorresc committed 69 days ago

Verified 4ea2eaac

CANN: fix an issue where get_env was not fully renamed (#18796)

noemotiovon committed 69 days ago

Verified e20fa27a

CANN: support gated linear attn (#18653)

hipudding committed 69 days ago

Verified baa4ba0a

Older