ngxson/llama.cpp

Pull Requests Commits

test: remember to set n_thread for cpu backend

ngxson committed 122 days ago

bf9d14cd

ngxson committed 122 days ago

8549245c

ngxson committed 124 days ago

1cd87e07

fix single-thread case for init_tensor_uniform

ngxson committed 124 days ago

56d02f6f

test-backend-ops: force single thread on wasm

ngxson committed 124 days ago

5616b9c2

more debugging stuff

ngxson committed 124 days ago

990a98ae

webgpu : fix build on emscripten

ngxson committed 124 days ago

4529332e

Thinking model disabled assistant prefill (#15404)

gabe-l-hart committed 125 days ago

Verified 5fac79cb

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

ericcurtin committed 125 days ago

Verified 408ff524

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

JohannesGaessler committed 125 days ago

Verified 5143fa89

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

danbev committed 125 days ago

Verified 3a550b5c

gguf: gguf_writer refactor (#15691)

Green-Sky committed 125 days ago

Verified a8128382

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

ggerganov committed 125 days ago

Verified c610b6c1

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

danbev committed 125 days ago

Verified 5d6688de

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

ExtReMLapin committed 125 days ago

Verified 4fd1242b

chat : nemotron thinking & toolcalling support (#15676)

pwilkin committed 125 days ago

Verified b2426e46

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

pwilkin committed 125 days ago

Verified 9e2b1e83

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

danbev committed 126 days ago

Verified fb15d649

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

gabe-l-hart committed 126 days ago

Verified 856ed094

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

danbev committed 126 days ago

Verified d1e2adba

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

noemotiovon committed 126 days ago

Verified c1c354e4

server: add exceed_context_size_error type (#15780)

ngxson committed 126 days ago

Verified a68d9144

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

ericcurtin committed 126 days ago

Verified badb80ca

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

leejet committed 126 days ago

Verified 0a1b3982

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

hipudding committed 126 days ago

Verified 5421f63a

opencl: add hs=40 to FA (#15758)

rmatif committed 126 days ago

Verified 820bc985

CANN: fix acl_rstd allocation size in ggml_cann_rms_norm (#15760)

noemotiovon committed 126 days ago

Verified 239b60e8

vulkan: fix mmv subgroup16 selection (#15775)

0cc4m committed 126 days ago

Verified dff7551b

vulkan: don't use std::string in load_shaders, to improve compile time (#15724)

jeffbolznv committed 127 days ago

Verified 0fce7a12

vulkan : update ggml_vk_instance_validation_ext_available (#15666)

danbev committed 127 days ago

Verified 8227695d

Older