Commits

Commits on Aug 7, 2024

CUDA/HIP: fix tests/test-backend-ops (#8896 )
JohannesGaessler
committedAug 7, 2024

Commits on Aug 6, 2024

CUDA: fix padding logic for FP16/FP32 (#8884 )
JohannesGaessler
committedAug 6, 2024

Commits on Jul 28, 2024

cmake: use 1 more thread for non-ggml in CI (#8740 )
JohannesGaessler
committedJul 28, 2024

Commits on Jul 20, 2024

CUDA: MMQ code deduplication + iquant support (#8495 )
JohannesGaessler
committedJul 20, 2024

Commits on Jul 18, 2024

CUDA: fix partial offloading for ne0 % 256 != 0 (#8572 )
JohannesGaessler
committedJul 18, 2024

Commits on Jul 17, 2024

lookup: fibonacci hashing, fix crashes (#8548 )
JohannesGaessler
committedJul 17, 2024

Commits on Jul 16, 2024

make/cmake: add missing force MMQ/cuBLAS for HIP (#8515 )
JohannesGaessler
committedJul 16, 2024

Commits on Jul 11, 2024

CUDA: optimize and refactor MMQ (#8416 )
JohannesGaessler
committedJul 11, 2024

Commits on Jul 9, 2024

make/cmake: LLAMA_NO_CCACHE -> GGML_NO_CCACHE (#8392 )
JohannesGaessler
committedJul 9, 2024

Commits on Jul 5, 2024

CUDA: MMQ support for iq4_nl, iq4_xs (#8278 )
JohannesGaessler
committedJul 5, 2024
CUDA: fix MMQ stream-k rounding if ne00 % 128 != 0 (#8311 )
JohannesGaessler
committedJul 5, 2024

Commits on Jul 1, 2024

CUDA: refactor and optimize IQ MMVQ (#8215 )
JohannesGaessler
committedJul 1, 2024

Commits on Jun 27, 2024

CUDA: fix MMQ stream-k for --split-mode row (#8167 )
JohannesGaessler
committedJun 27, 2024

Commits on Jun 26, 2024

CUDA: fix misaligned shared memory read (#8123 )
JohannesGaessler
committedJun 26, 2024

Commits on Jun 24, 2024

Commits on Jun 20, 2024

common: fix warning (#8036 )

JohannesGaessler
and
slaren
committedJun 20, 2024
CUDA: stream-k decomposition for MMQ (#8018 )
JohannesGaessler
committedJun 20, 2024

Commits on Jun 14, 2024

CUDA: faster q2_K, q3_K MMQ + int8 tensor cores (#7921 )
JohannesGaessler
committedJun 14, 2024

Commits on Jun 12, 2024

CUDA: fix broken oob check for FA vec f32 kernel (#7904 )
JohannesGaessler
committedJun 12, 2024

Commits on Jun 11, 2024

llama-bench: more compact markdown tables (#7879 )
JohannesGaessler
committedJun 11, 2024
CUDA: int8 tensor cores for MMQ (q4_K, q5_K, q6_K) (#7860 )
JohannesGaessler
committedJun 11, 2024

Commits on Jun 10, 2024

CUDA: use tensor cores for MMQ (#7676 )
JohannesGaessler
committedJun 10, 2024

Commits on Jun 9, 2024

CUDA: revise q8_1 data layout for mul_mat_q (#7824 )
JohannesGaessler
committedJun 9, 2024

Commits on Jun 7, 2024

server: update cache_prompt documentation [no ci] (#7745 )
JohannesGaessler
committedJun 7, 2024

Commits on Jun 5, 2024

CUDA: refactor mmq, dmmv, mmvq (#7716 )
JohannesGaessler
committedJun 5, 2024

Commits on Jun 3, 2024

make: fix debug options not being applied to NVCC (#7714 )
JohannesGaessler
committedJun 3, 2024

Commits on Jun 1, 2024

Commits on May 31, 2024

scripts: update compare_llama_bench.py [no ci] (#7673 )
JohannesGaessler
committedMay 31, 2024

Commits on May 30, 2024

README: explain parallel build [no ci] (#7618 )
JohannesGaessler
committedMay 30, 2024

Commits on May 27, 2024

make: add --device-debug to NVCC debug flags (#7542 )
JohannesGaessler
committedMay 27, 2024