ggerganov/llama.cpp

Pull Requests Commits

models : dedup qwen35 graphs (#19660)

ggerganov committed 34 days ago

Verified 27326bfc

models : dedup Kimi Linear delta net implementation (#19668)

ymcki committed 34 days ago

Verified ad9f692f

Add Jinja support for "indent" string filter (#19529)

pwilkin committed 34 days ago

Verified 8a709735

ggml webgpu: Fix bug in dispatching large matrix-vector multiplication (#19535)

reeselevine committed 34 days ago

Verified e7f2f95c

server: save generated text for the /slots endpoint (for LLAMA_SERVER_SLOTS_DEBUG=1) (#19622)

matteoserva committed 35 days ago

Verified b55dcdef

model: support GLM-OCR (#19677)

ngxson committed 35 days ago

Verified eeef3cfc

docs: Fix broken links for preparing models in Backends (#19684)

MaciejDromin committed 35 days ago

Verified e99f1083

ggml webgpu: shader library organization (#19530)

reeselevine committed 35 days ago

Verified 238856ec

Pre-MCP UI and architecture cleanup (#19689)

allozaur committed 35 days ago

Verified ea003229

vulkan: split mul_mat into multiple dispatches to avoid overflow (#19509)

jeffbolznv committed 35 days ago

Verified d0061be8

common : make small string helpers as inline functions (#19693)

angt committed 35 days ago

Verified a569bda4

opencl: refactor expm1 and softplus (#19404)

shaofeiqi committed 35 days ago

Verified e2f19b32

opencl: optimize mean and sum_row kernels (#19614)

shaofeiqi committed 35 days ago

Verified 983559d2

model-conversion : add option to print tensor values (#19692)

danbev committed 36 days ago

Verified 2b089c77

Pre-MCP UI and architecture cleanup (#19685)

allozaur committed 36 days ago

Verified afa6bfe4

ggml: ggml-cpu: force-no-lto-for-cpu-feats (#19609)

talhaHavadar committed 36 days ago

Verified ae2d3f28

cuda : enable CUDA graphs for MMID 1 <= BS <= 4 (#19645)

ggerganov committed 36 days ago

Verified ad8207af

model-conversion : make printing of config values optional (#19681)

danbev committed 36 days ago

Verified 667b6942

ci : bump komac version (#19682)

CISC committed 36 days ago

Verified e48349a4

build : link ws2_32 as PUBLIC on Windows (#19666)

angt committed 36 days ago

Verified ae46a61e

build : cleanup library linking logic (#19665)

angt committed 36 days ago

Verified 65cede7c

convert : add JoyAI-LLM-Flash (#19651)

dranger003 committed 36 days ago

Verified 05fa625e

perplexity: add proper batching (#19661)

AesSedai committed 37 days ago

Verified d6129011

common : inline functions (#18639)

Nekotekina committed 37 days ago

Verified cceb1b4e

ggml : make `ggml_is_view` as API (#19539)

foldl committed 37 days ago

Verified d23a5599

model: Add support for Tiny Aya Models (#19611)

saurabhdash2512 committed 37 days ago

Verified 5f28c53d

build : rework llama_option_depr to handle LLAMA_CURL (#19658)

angt committed 37 days ago

Verified 44084941

Adjust workaround for ROCWMMA_FATTN/GFX9 to only newer ROCm veresions (#19591)

superm1 committed 37 days ago

Verified 2ba9adc0

models : deduplicate delta-net graphs for Qwen family (#19597)

ggerganov committed 37 days ago

Verified cc45f2ad

graph : fix KQ mask, lora, cvec reuse checks (#19644)

ggerganov committed 37 days ago

Verified d5dfc330

Newer Older