ggml-org/llama.cpp

Pull Requests Commits

ggml-metal: fix 2D async copy to use row-by-row transfers

ggerganov committed 61 days ago

Verified c6a04cb5

pi: add rule to not force push branches unless asked

ggerganov committed 61 days ago

Verified f9e19a1f

ggml-metal: implement async 2D tensor copy functions

ggerganov committed 61 days ago

Verified c3a54d62

webui: fix slow mic stop and WAV encode (#22480)

ServeurpersoCom committed 61 days ago

Verified 59237bfb

ggml-cpu : disable tiled matmul on AIX to fix page boundary segfault (#22293)

shalinib-ibm committed 61 days ago

Verified 1cbc846e

ggml-cuda: refactor fusion code (#22468)

am17an committed 61 days ago

Verified 3142f1db

ggml-cpu: cmake: append xsmtvdotii march for SpacemiT IME (#22317)

qiurui144 committed 61 days ago

Verified b5c4227d

ggml-webgpu: Fix bug in FlashAttention support check (#22492)

reeselevine committed 61 days ago

Verified d6a50940

common: Intentionally leak logger instance to fix hanging on Windows (#22273)

rillomas committed 61 days ago

Verified 7b95ea5d

ggml : add sve tuned code for gemm_q8_0_4x8_q8_0() kernel (#21916)

hrushitfujitsu committed 61 days ago

Verified bdc9c743

TP: fix delayed AllReduce + zero-sized slices (#22489)

JohannesGaessler committed 61 days ago

Verified 739393be

ggml-cuda: Repost of 21896: Blackwell native NVFP4 support (#22196)

michaelw9999 committed 61 days ago

Verified fc2b0053

ggml-cuda: add flash-attn support for DKQ=320/DV=256 with ncols2=32 (… (#22286)

lnigam committed 61 days ago

Verified 7b8443ac

convert : add support for Nemotron Nano 3 Omni (#22481)

danbev committed 62 days ago

Verified 5d56effd

common : re-arm reasoning budget after DONE on new <think> (#22323)

BruceJillis committed 62 days ago

Verified 52e5f0a5

vulkan: Coalesce Q4_K/Q5_K scale loads (#21751)

TheBlueMatt committed 62 days ago

Verified f9f33654

ggml-webgpu: fix buffer aliasing for ssm_scan and refactor aliasing logic (#22456)

reeselevine committed 62 days ago

Verified 98bb5791

webui: Server tools (#21237)

allozaur committed 62 days ago

Verified f42e29fd

vulkan: add barrier after writetimestamp (#21865)

jeffbolznv committed 62 days ago

Verified 19821178

ggml: improve SPIR-V headers detection with __has_include (#21918)

EmilAskerov committed 62 days ago

Verified 698d19b9

ggml : skip already registered backends and devices (#22296)

angt committed 62 days ago

Verified 50494a28

ggml : revert to -lm linking instead of find_library (#22355)

angt committed 62 days ago

Verified d530d6e7

CANN: add new ops, optimize existing ops (#21204)

hipudding committed 62 days ago

Verified c3e08f47

spec : refactor params (#22397)

ggerganov committed 62 days ago

Verified 14e733e3

server: use pos_next instead of n_tokens for m-rope (#22439)

am17an committed 62 days ago

Verified 516e8d7a

ggml-webgpu: add Q1_0 support (#22374)

SharmaRithik committed 62 days ago

Verified 434b2a1f

server: (router) Forward form-data to model server (Fixes #22044) (#22118)

tha80 committed 62 days ago

Verified 983ca899

add fast mat-vec kernels for i-quants (#22344)

SharmaRithik committed 63 days ago

Verified 665abc60

Additional test for common/gemma4 : handle parsing edge cases (#22420)

hextriclosan committed 63 days ago

Verified 4414c04b

fix: rpc-server cache may not work in Windows environments (#22394)

unraido committed 63 days ago

Verified ceaf47c4

Older