ggml-org/llama.cpp

Pull Requests Commits

cont : fix `cur_buf_size` init after flushing a buffer

ggerganov committed 17 days ago

Verified 7c7be0fb

ggml : add `alloc_buffer_n` to buffer type interface

ggerganov committed 17 days ago

Verified e0d7afdf

mtmd : add post-decode callback (#24645)

ggerganov committed 17 days ago

Verified e3cab403

vulkan: support more CONCAT types (#24579)

jeffbolznv committed 17 days ago

Verified 9dbc6621

wasm : fix fallback symbol collision (#24639)

abetlen committed 17 days ago

Verified 6eab4718

SYCL: use native subgroup size for K-quant DMMV (#21700)

PMZFX committed 17 days ago

Verified e3bb1add

sycl: fix soft_max_f32 max reduction (#24451)

someoneinjd committed 17 days ago

Verified d8a3f523

sycl : fix reorder function; add fp32/fp16 in build script (#24578)

arthw committed 17 days ago

Verified 72be44f1

sycl : enhance set_rows to support q1_0, mxfp4, nvfp4 (#24564)

arthw committed 17 days ago

Verified 8872ab54

[SYCL] add to support pool_1d, move pool_1d/2d code to pool.cpp/hpp (#24584)

arthw committed 17 days ago

Verified 987fbd82

[SYCL]: Remove per-allocation Level Zero runtime checks (#23399)

sanmai committed 17 days ago

Verified c035ff49

metal : add repeat bf16 (#24638)

ggerganov committed 17 days ago

Verified 272088b9

chat: fix whitespace problems once and for all (#24624)

pwilkin committed 17 days ago

Verified a6dff712

UI/svg block rendering (#24080)

ServeurpersoCom committed 17 days ago

Verified 2a6c391a

CUDA: only support F32/F16 for GGML_OP_REPEAT (#24533)

leonardHONG committed 17 days ago

Verified 3686e9d6

ggml-webgpu: improve i-quants mul_mat performance and speed up prefill (#24530)

yomaytk committed 17 days ago

Verified 6e9007ae

convert : fix lora base model arch retrieval (#24621)

CISC committed 17 days ago

Verified dd4623a7

fix(ui): render thinking/reasoning block content as markdown (#24611)

franitel committed 17 days ago

Verified ef8268fe

ui: Add HEIC/HEIF image support (#24137)

NickM-27 committed 18 days ago

Verified 5f04dc7a

chat: add dedicated Cohere2MoE (North Code) parser (#24615)

pwilkin committed 18 days ago

Verified aedb2a5e

docs : fix typos in CUDA-FEDORA.md and grammars/README.md (#24459)

m-atharkhan committed 18 days ago

Verified 8edaca90

docker: specify registry to simplify Podman builds (#24607)

Minoru committed 18 days ago

Verified 20c5266f

UI/mobile keyboard and pwa popup fixes (#24610)

ServeurpersoCom committed 18 days ago

Verified fd5869fb

ui : fix ui clipping in mobile due to incorrect height setup (#24605)

amoshydra committed 18 days ago

Verified 1fd6dfe9

jinja : add count/d/e filter aliases (#24606)

CISC committed 18 days ago

Verified acd79d60

cli : fix not copying preserved tokens (#24258)

michaelw9999 committed 18 days ago

Verified 6e14286e

Add cohere2moe to llama-vocab for TINY_AYA (#24601)

bartowski1182 committed 18 days ago

Verified 8ed274ef

ci : use CUDA label for cuda backend (#24594)

CISC committed 18 days ago

Verified 46722116

add sycl to check-release (#24583)

CISC committed 18 days ago

Verified c2ba3e47

ui : fix llama-ui-embed crash when no asset dir is given (#24597)

aldehir committed 18 days ago

Verified 53bd47ea

Older