ngxson/llama.cpp

Pull Requests Commits

add cont version for vulkan

ngxson committed 116 days ago

65b19a00

ngxson committed 116 days ago

0a2b23c2

extend test number range

ngxson committed 116 days ago

9f5d5cdf

ngxson committed 116 days ago

60e8f200

add non-cont test case

ngxson committed 116 days ago

01af2c2d

fix CPU non-cont

ngxson committed 116 days ago

fee65fa1

ngxson committed 116 days ago

4f57dda5

ngxson committed 116 days ago

d4f78de3

ggml: allow casting between f32 and i32

ngxson committed 117 days ago

f3b489b2

vulkan: fix mmv subgroup16 selection (#15775)

0cc4m committed 117 days ago

Verified dff7551b

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

jeffbolznv committed 117 days ago

Verified 0fce7a12

vulkan : update ggml_vk_instance_validation_ext_available (#15666)

danbev committed 117 days ago

Verified 8227695d

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

relent95 committed 117 days ago

Verified 0014fb4a

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

ORippler committed 117 days ago

Verified 661ae31c

model-conversion : fix pyright errors (#15770)

danbev committed 117 days ago

Verified 407c2378

sampling : optimize dist sampler (#15704)

ggerganov committed 117 days ago

Verified cdedb70a

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

danbev committed 117 days ago

Verified 2c8dac72

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

danbev committed 117 days ago

Verified 40a751ea

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

hipudding committed 117 days ago

Verified 5eae9348

ggml-cpu : optimize RVV kernels (#15720)

xctan committed 117 days ago

Verified 05c0380f

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

danbev committed 117 days ago

Verified 8c3fdf44

CANN: Mask unsupported TRANSPOSE_1D operator (#15733)

hipudding committed 117 days ago

Verified f6da8cb8

CANN: Fix type float_t to float (#15736)

noemotiovon committed 118 days ago

Verified 8a2234ea

fix: resolve unsigned int initialization warning for n_dims/size in gguf.cpp (#15754)

skrandy committed 118 days ago

Verified 3de00820

chore: Update `.clang-format` to use `BinPackArguments=true` (#15744)

ORippler committed 118 days ago

Verified 69db8a52

llama: -fa 1/0/-1 aliases for -fa on/off/auto (#15746)

JohannesGaessler committed 118 days ago

Verified c466abe1

vulkan: fix shaders gen when no integer dot is available (#15740)

0cc4m committed 118 days ago

Verified 0a2a3841

CANN: Resolve soft_max precision issue (#15730)

hipudding committed 118 days ago

Verified 9961d244

vulkan: Fix macro parameter order for f32 matmul shaders (#15716)

jeffbolznv committed 118 days ago

Verified 25f1045f

opencl: add attn sinks support for FA kernels (#15706)

rmatif committed 118 days ago

Verified 97669e40

Older