ggml-org/llama.cpp

Pull Requests Commits

minor : std::unordered_set over std::set

ggerganov committed 45 days ago

Verified 08b5d956

deprecate llama_adapter_lora_free

ngxson committed 57 days ago

cb5e0f87

lora: make sure model keep track of associated adapters

ngxson committed 57 days ago

f5e8bfdd

lora: count lora nodes in graph_max_nodes (#18469)

ngxson committed 57 days ago

Verified cd78e57c

sampling: reuse token data buffer in llama_sampler_sample (#18365)

JayZenith committed 57 days ago

Verified c32fa21d

server: fix files built redundantly (#18474)

jeffbolznv committed 57 days ago

Verified f14f4e42

kleidiai: add and integrate SVE 256-bit vector-length kernel (#18458)

chaxu01 committed 57 days ago

Verified 2d6c00a9

CUDA: add log line when mxfp4 acceleration is used (#18483)

am17an committed 58 days ago

Verified d77d7c5c

model-conversion : use CONVERTED_MODEL for compare-embeddings (#18461)

danbev committed 58 days ago

Verified a864fb1c

webui: fix prompt progress ETA calculation (#18468)

ngxson committed 58 days ago

Verified 51a48720

Webui/prompt processing progress (#18300)

ServeurpersoCom committed 58 days ago

Verified c9a3b40d

CUDA: fix replacment of bad archs in CMake (#18457)

JohannesGaessler committed 58 days ago

Verified 0bd1212a

server : Cmdline arg -to changes http read timeout from current 600sec default (#18279)

wbtek committed 58 days ago

Verified 5b1248c9

contributing: tighten AI usage policy (#18388)

ngxson committed 58 days ago

Verified 3595ae59

android: routine maintenance - Dec 2025 (#18338)

naco-siren committed 58 days ago

Verified c1366056

server : handle closed connection for tasks (#18459)

ggerganov committed 58 days ago

Verified 2a85f720

model-conversion : add device option to embd run orig model (#18386)

danbev committed 58 days ago

Verified 7cbec34a

retrieval : use at most n_seq_max chunks (#18400)

hectorem2 committed 59 days ago

Verified 0c898640

common: fix return value check for setpriority (#18412)

o7si committed 59 days ago

Verified daa242df

CUDA: Blackwell features for non-native builds (#18436)

JohannesGaessler committed 59 days ago

Verified e70e640d

cuda: fix race condition in cumsum (#18448)

am17an committed 59 days ago

Verified 5fa66c6e

ci : re-enable rocm build on amd64 (#18439)

TimNN committed 59 days ago

Verified 382808c1

HIP: Use mmq on MFMA devices for MUL_MAT_ID in cases where a lot of splits would be generated (#18202)

IMbackK committed 59 days ago

Verified 4ffc47cb

model : Plamo3 support (#17304)

mmnga committed 59 days ago

Verified 9c675c71

Revert "ggml-cuda: use CMAKE_CUDA_ARCHITECTURES if set when GGML_NATIVE=ON (#18413)" (#18426)

am17an committed 59 days ago

Verified 07a0c4ba

rpc: fix segfault on invalid endpoint format (#18387)

o7si committed 60 days ago

Verified 60f17f56

llama-fit-params: fix step size for last device (#18415)

JohannesGaessler committed 60 days ago

Verified f8d561eb

github: update issue templates [no ci] (#18410)

JohannesGaessler committed 60 days ago

Verified e59efe6a

mtmd: clarify that we no longer accept AI-generated PRs (#18406)

ngxson committed 60 days ago

Verified cffa5c46

cmake: Added more x86_64 CPU backends when building with `GGML_CPU_ALL_VARIANTS=On` (#18186)

bberberov committed 60 days ago

Verified 94de74e7

Older