ggml-org/whisper.cpp

Pull Requests Commits

whisper : alternative way to handle the external encoders

ggerganov committed 1 year ago

Verified f25edade

whisper : fix usage of extenral encoders (e.g. CoreML)

ggerganov committed 1 year ago

Verified 74c260fe

talk-llama : sync llama.cpp

ggerganov committed 1 year ago

Verified 55152929

ggerganov committed 1 year ago

Verified 25a90ffa

ggml-backend : sync remnant

ggerganov committed 1 year ago

Verified 866b67ca

CUDA: mul_mat_vec_q tiling, refactor mul mat logic (llama/5434)

JohannesGaessler committed 1 year ago

Verified d7e9f58f

vulkan: only use M-sized matmul on Apple GPUs (llama/5412)

slp committed 1 year ago

Verified 04839bae

ggml : fix compile warnings (unused vars) (llama/4966)

ggerganov committed 1 year ago

Verified 3cc6e04a

ggml : add mmla kernels for quantized GEMM (llama/4966)

snadampal committed 1 year ago

Verified b7ef178b

metal : use autoreleasepool to avoid memory leaks (llama/5437)

irbull committed 1 year ago

Verified 47dfe9d4

ggml-alloc : v3 (ggml/727)

slaren committed 1 year ago

Verified 1d3270cc

examples : added audio_ctx argument to main and server (#1857)

dscripka committed 1 year ago

Verified a6fb6ab5

metal : option to embed MSL source into compiled binary (#1842)

didzis committed 1 year ago

Verified 163e74b6

examples : initialize context params properly (#1852)

ggerganov committed 1 year ago

Verified f273e66d

talk-llama : sync llama.cpp

ggerganov committed 1 year ago

Verified 02b4c52c

ggerganov committed 1 year ago

Verified 518199c0

src : relocate new backend sources

ggerganov committed 1 year ago

Verified 8b17a2f7

ggml : fix `error C2078: too many initializers` for MSVC ARM64 (llama/5404)

Xarbirus committed 1 year ago

Verified b6d28279

CUDA: more warps for mmvq on NVIDIA (llama/5394)

JohannesGaessler committed 1 year ago

Verified 9711bae0

CUDA: fixed mmvq kernel for bs 2,3,4 and -sm row (llama/5386)

JohannesGaessler committed 1 year ago

Verified eec38f63

Basic Vulkan Multi-GPU implementation (llama/5321)

0cc4m committed 1 year ago

Verified ef5e6b74

CUDA: mul_mat_vec_q max. batch size 8 -> 4 (llama/5370)

JohannesGaessler committed 1 year ago

Verified 77bf6b5f

Slight quantization improvement for Q4_K and Q5_K (llama/5361)

ikawrakow committed 1 year ago

Verified b562fff9

CUDA: mul_mat_vec_q for batch sizes > 1 (llama/5351)

JohannesGaessler committed 1 year ago

Verified b5dec374

ggml : make use of ggml-quants.h possible in C++ code (llama/5338)

ikawrakow committed 1 year ago

Verified fa0dc616

ggml : avoid duplicating function calls using MIN/MAX macros (llama/5325)

tom7 committed 1 year ago

Verified 55bcd62a

iq2_xxs: tune quantization (llama/5320)

ikawrakow committed 1 year ago

Verified 0ed762d6

cuda : fix LLAMA_CUDA_F16 (llama/5262)

slaren committed 1 year ago

Verified 1b5bb779

metal : add im2col F32 dst support (llama/5132)

ggerganov committed 1 year ago

Verified 9b735cea

llava : add MobileVLM support (llama/5132)

JidongZhang-THU committed 1 year ago

Verified 12c462d6

Older