ggml-org/whisper.cpp

Pull Requests Commits

ios : update SPM package

ggerganov committed 2 years ago

Verified 3ac05580

bench : variable n_past

ggerganov committed 2 years ago

Verified a1664574

metal : remove the "concurrent" flag

ggerganov committed 2 years ago

Verified bfcb2a2a

whisper : clean-up ggml_mul_mat_pad

ggerganov committed 2 years ago

Verified 0d5e4cdc

whisper : add description of ggml_mul_mat_pad

ggerganov committed 2 years ago

Verified 2b4160af

whisper : add comment for disabling mul-mat padding

ggerganov committed 2 years ago

Verified f3655438

bench : fix uninitialized vars

ggerganov committed 2 years ago

Verified c46167f8

whisper : add ggml_mul_mat_pad

ggerganov committed 2 years ago

Verified af947cb7

bench : start benching the decoder

ggerganov committed 2 years ago

Verified e81c67a1

bench : fix timings by running a pre-heat

ggerganov committed 2 years ago

Verified f408c645

coreml : add code to toggle Core ML config (CPU, ANE, GPU)

ggerganov committed 2 years ago

Verified d863f725

ios : update submodule

ggerganov committed 2 years ago

Verified d37f56e7

readme : add Metal info

ggerganov committed 2 years ago

Verified 23277d21

metal : sync latest llama.cpp kernels

ggerganov committed 2 years ago

Verified ecb23fb1

metal : speed-up KQ multiplication

ggerganov committed 2 years ago

Verified 8e8daa84

swiftui : fix build

ggerganov committed 2 years ago

Verified 16db4da3

ios : add Metal support

ggerganov committed 2 years ago

Verified 257d7942

objc : fix build (no Metal yet)

ggerganov committed 2 years ago

Verified 181bb8cb

whisper : add <functional> header

ggerganov committed 2 years ago

Verified 796f84cd

cmake : update to support Metal build

ggerganov committed 2 years ago

Verified 77f4bf49

whisper : factor out alloc init in a function

ggerganov committed 2 years ago

Verified b6f09669

whisper : add whisper_allocr to wrap ggml_allocr

ggerganov committed 2 years ago

Verified 254b6872

ggml-alloc : try to make CI happy by reducing vram to 128GB

ggerganov committed 2 years ago

Verified b19888cf

ggml : use simpler ggml_bytes() implementation

ggerganov committed 2 years ago

Verified 905c9441

whisper : offload the Encoder to Metal

ggerganov committed 2 years ago

Verified 3074a7ff

whisper : remove ggml_repeat in the encoder

ggerganov committed 2 years ago

Verified ec9a7db7

metal : run "cross" step on the GPU

ggerganov committed 2 years ago

Verified cd476375

ggml : fix ggml_nbytes (probably temp solution)

ggerganov committed 2 years ago

Verified 9fdd4153

metal : add multi-decoder support

ggerganov committed 2 years ago

Verified 79a88057

metal : decoder works on GPU!

ggerganov committed 2 years ago

Verified fbc9ddc5

Older