ggml-org/llama.cpp

Pull Requests Commits

ggml-cpu: clean up s390x SIMD (#15855)

taronaeo committed 3 days ago

Verified d36e61c5

vulkan: Support pad_ext (#15794)

jeffbolznv committed 3 days ago

Verified c97b5e58

vulkan: Use larger loads in scalar/coopmat1 matmul (#15729)

jeffbolznv committed 3 days ago

Verified 267e9986

ggml WebGPU: remove userdata from request adapter callback (#15527)

danbev committed 3 days ago

Verified 3b15924d

CUDA: faster tile FA (Pascal/AMD), headsize 256 (#15769)

JohannesGaessler committed 4 days ago

Verified 79bc4292

kleidiai: generalize compute_forward_kv_cache to compute_forward_fp16 (#15817)

chaxu01 committed 4 days ago

Verified c4df49a4

server : speed up tests (#15836)

ngxson committed 4 days ago

Verified 3c3635d2

server : implement prompt processing progress report in stream mode (#15827)

ngxson committed 4 days ago

Verified 61bdfd52

ggml-cpu: document use of "free" memory [no ci] (#15834)

JohannesGaessler committed 4 days ago

Verified 01806e77

ggml-cpu: drop support for nnpa intrinsics (#15821)

taronaeo committed 4 days ago

Verified 186415d5

aLoRA Support (#15327)

gabe-l-hart committed 4 days ago

Verified fd621880

ci : exempt correct research label (#15825)

CISC committed 5 days ago

Verified 4281c7b3

Thinking model disabled assistant prefill (#15404)

gabe-l-hart committed 5 days ago

Verified 5fac79cb

Implement --log-colors with always/never/auto (#15792)

ericcurtin committed 5 days ago

Verified 408ff524

CUDA: fastdiv, launch bounds for mmvq + q8_1 quant (#15802)

JohannesGaessler committed 5 days ago

Verified 5143fa89

tests : add --list-ops and --show-coverage options (#15745)

danbev committed 5 days ago

Verified 3a550b5c

gguf: gguf_writer refactor (#15691)

Green-Sky committed 5 days ago

Verified a8128382

kv-cache : fix SWA checks + disable cacheless iSWA (#15811)

ggerganov committed 5 days ago

Verified c610b6c1

model-conversion : add --embeddings flag to modelcard.template [no ci] (#15801)

danbev committed 5 days ago

Verified 5d6688de

chat : fixed crash when Hermes 2 <tool_call> had a newline before it (#15639)

ExtReMLapin committed 6 days ago

Verified 4fd1242b

chat : nemotron thinking & toolcalling support (#15676)

pwilkin committed 6 days ago

Verified b2426e46

scripts : add Jinja tester PySide6 simple app (#15756)

pwilkin committed 6 days ago

Verified 9e2b1e83

llama : add support for EmbeddingGemma 300m (#15798)

danbev committed 6 days ago

Verified fb15d649

metal : Add template specialization for mul_mm_id w/ ne20 == 10 (#15799)

gabe-l-hart committed 6 days ago

Verified 856ed094

llama : set n_outputs to 1 to avoid 0 outputs mean-pooling (#15791)

danbev committed 6 days ago

Verified d1e2adba

CANN: Refactor ND to NZ workspace to be per-device (#15763)

noemotiovon committed 6 days ago

Verified c1c354e4

server: add exceed_context_size_error type (#15780)

ngxson committed 6 days ago

Verified a68d9144

Document the new max GPU layers default in help (#15771)

ericcurtin committed 6 days ago

Verified badb80ca

ggml: add ops for WAN video model (cuda && cpu) (#15669)

leejet committed 6 days ago

Verified 0a1b3982

CANN: Fix precision issue on 310I DUO multi-devices (#15784)

hipudding committed 6 days ago

Verified 5421f63a

Newer Older