ggml-org/llama.cpp

Pull Requests Commits

RPC: improve loading

am17an committed 24 days ago

7d8a964e

cont : fix conflict

ggerganov committed 24 days ago

ee2ddd09

move graph_uids to rpc_dispatcher

am17an committed 24 days ago

043d5ea3

fix flush for apple rdma

am17an committed 24 days ago

7b1c77be

rpc: allow -sm tensor

am17an committed 24 days ago

e1d0dd6b

llama : fix tensor split for fused qkv with uneven K/V head sizes (#29294)

am17an committed 24 days ago

Verified f805c57a

hexagon: add q5_k quant type support (#29123)

jhen0409 committed 24 days ago

Verified 4de09265

hexagon: use DMA for contiguous dim1 CONCAT (#29404)

kurquhar committed 24 days ago

Verified ed319feb

metal : fix graph capture and handle empty graphs (#29390)

ggerganov committed 25 days ago

Verified 84e76d8a

metal : optimize sparse FA + clean-up (#29377)

ggerganov committed 25 days ago

Verified cdc06426

sync : ggml (#29396)

ggerganov committed 25 days ago

Verified bced4595

hexagon: handle multi-sequence in concat_2d (#29344)

jhen0409 committed 25 days ago

Verified a02c7f58

llama-grammar: fix numeric truncation for token_id parsing (#29382)

apach301 committed 25 days ago

Verified 5cf3a352

hexagon: dynamic quantizer improvements (#29395)

kurquhar committed 25 days ago

Verified 07fc586e

hexagon: support I32 CPY and CONT (#29379)

kurquhar committed 25 days ago

Verified 97a418bd

cuda : add F16 kernel support for CONV_2D_DW (#29064)

crion99 committed 25 days ago

Verified a72e04ab

test: flush status (#28352)

kurquhar committed 25 days ago

Verified 8212c780

ui : fix missing svg use and animation elements in preview and download (#28962)

nandan2003 committed 25 days ago

Verified 945064fc

llama: add llama_batch_ext (#24669)

ngxson committed 25 days ago

Verified fc343a84

server : change default pytest workers to 4 (#29376)

danbev committed 25 days ago

Verified 308883b3

ci : use hf-jobs-cpu-performance, disable pytest workers (#29369)

ggerganov committed 25 days ago

Verified 70596c4d

vulkan: int8 coopmat1 matmul implementation for AMD RDNA3 and RDNA4 (#27952)

0cc4m committed 25 days ago

Verified 70c4e158

vulkan: handle misalignment in conv_2d and conv_3d (#29365)

0cc4m committed 25 days ago

Verified 6b790a9c

vulkan: tune KHR cooperative matrix support for Adreno GPUs (#29328)

Raman-Raje committed 25 days ago

Verified 3423f940

cuda : add conv3d with implicit GEMM (#29137)

leejet committed 25 days ago

Verified 53ed051c

model : add Ling 3.0 VL support (#29151)

aetherbird committed 25 days ago

Verified f830688e

server,common : fix the GCC 12 stringop-overread false positive (again) (#29325)

angt committed 25 days ago

Verified 2b705839

test-save-load-state : print a per-model results table in --models mode (#29316)

ggerganov committed 25 days ago

Verified 4c5957c2

hexagon: reject MUL_MAT_ID when src1 precision is F32 (#29348)

jhen0409 committed 25 days ago

Verified 9710a321

scripts : make-release-desc - link previous release in changelog title (#29336)

ggerganov committed 25 days ago

Verified 013b31c0

Older