ggerganov/llama.cpp

Pull Requests Commits

Rewrite llama-run to use llama-server

ericcurtin committed 94 days ago

7b717fb4

gguf: gguf_writer refactor (#15691)

Green-Sky committed 94 days ago

Verified a8128382

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

ggerganov committed 94 days ago

Verified c610b6c1

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

danbev committed 95 days ago

Verified 5d6688de

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

ExtReMLapin committed 95 days ago

Verified 4fd1242b

chat : nemotron thinking & toolcalling support (#15676)

pwilkin committed 95 days ago

Verified b2426e46

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

pwilkin committed 95 days ago

Verified 9e2b1e83

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

danbev committed 95 days ago

Verified fb15d649

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

gabe-l-hart committed 95 days ago

Verified 856ed094

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

danbev committed 95 days ago

Verified d1e2adba

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

noemotiovon committed 95 days ago

Verified c1c354e4

server: add exceed_context_size_error type (#15780)

ngxson committed 95 days ago

Verified a68d9144

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

ericcurtin committed 95 days ago

Verified badb80ca

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

leejet committed 95 days ago

Verified 0a1b3982

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

hipudding committed 95 days ago

Verified 5421f63a

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

rmatif committed 95 days ago

Verified 820bc985

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

noemotiovon committed 96 days ago

Verified 239b60e8

vulkan: fix mmv subgroup16 selection (#15775)

0cc4m committed 96 days ago

Verified dff7551b

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

jeffbolznv committed 96 days ago

Verified 0fce7a12

vulkan : update ggml_vk_instance_validation_ext_available (#15666)

danbev committed 96 days ago

Verified 8227695d

ggml vulkan: add hardsigmoid and hardswish operations (#15762)

relent95 committed 96 days ago

Verified 0014fb4a

CUDA: Optimize `rms_norm_f32` kernel and its fused variants, giving 1-6% perf E2E (#15715)

ORippler committed 96 days ago

Verified 661ae31c

model-conversion : fix pyright errors (#15770)

danbev committed 96 days ago

Verified 407c2378

sampling : optimize dist sampler (#15704)

ggerganov committed 96 days ago

Verified cdedb70a

llama : fix incorrect model type for Gemma 270M (#15764)

danbev committed 96 days ago

Verified 2c8dac72

model-conversion : remove hardcoded /bin/bash shebangs [no ci] (#15765)

danbev committed 96 days ago

Verified 40a751ea

CANN: Add RoPE contiguous check for 310I DUP device (#15735)

hipudding committed 96 days ago

Verified 5eae9348

ggml-cpu : optimize RVV kernels (#15720)

xctan committed 96 days ago

Verified 05c0380f

model-conversion : add missing curl script [no ci] (#15761)

danbev committed 96 days ago

Verified 8c3fdf44

CANN: Mask unsupported TRANSPOSE_1D operator (#15733)

hipudding committed 96 days ago

Verified f6da8cb8

Older