93935980f8
* metal : init * whisper : factor out graph builds * whisper : allocate encoder and decoder using ggml-alloc * whisper : ggml-alloc is now supported * whisper : CoreML support ggml-alloc * build : fix ggml-alloc * ios : update submodule * extra : update sync-ggml.sh script to also sync ggml-alloc * ci : see if this is causing the crash * whisper : refactor ggml-alloc init * whisper.android : try to fix build * whisper : initial Metal version * ci : try to debug vmem issue * metal : decoder works on GPU! * metal : add multi-decoder support * ggml : fix ggml_nbytes (probably temp solution) * metal : run "cross" step on the GPU * whisper : remove ggml_repeat in the encoder * whisper : offload the Encoder to Metal * ggml : use simpler ggml_bytes() implementation * ggml-alloc : try to make CI happy by reducing vram to 128GB * whisper : add whisper_allocr to wrap ggml_allocr * whisper : factor out alloc init in a function * cmake : update to support Metal build * whisper : add <functional> header * objc : fix build (no Metal yet) * ios : add Metal support * swiftui : fix build * metal : speed-up KQ multiplication * metal : sync latest llama.cpp kernels * readme : add Metal info * ios : update submodule * coreml : add code to toggle Core ML config (CPU, ANE, GPU) * bench : fix timings by running a pre-heat * bench : start benching the decoder * whisper : add ggml_mul_mat_pad * bench : fix uninitialized vars * whisper : add comment for disabling mul-mat padding * whisper : add description of ggml_mul_mat_pad * whisper : clean-up ggml_mul_mat_pad * metal : remove the "concurrent" flag * bench : variable n_past * ios : update SPM package
2.2 KiB
whisper.objc
Minimal Obj-C application for automatic offline speech recognition. The inference runs locally, on-device.
https://user-images.githubusercontent.com/1991296/197385372-962a6dea-bca1-4d50-bf96-1d8c27b98c81.mp4
Real-time transcription demo:
https://user-images.githubusercontent.com/1991296/204126266-ce4177c6-6eca-4bd9-bca8-0e46d9da2364.mp4
Usage
git clone https://github.com/ggerganov/whisper.cpp
open whisper.cpp/examples/whisper.objc/whisper.objc.xcodeproj/
// If you don't want to convert a Core ML model, you can skip this step by create dummy model
mkdir models/ggml-base.en-encoder.mlmodelc
Make sure to build the project in Release:
Also, don't forget to add the -DGGML_USE_ACCELERATE compiler flag for ggml.c in Build Phases.
This can significantly improve the performance of the transcription:
Core ML
If you want to enable Core ML support, you can add the -DWHISPER_USE_COREML -DWHISPER_COREML_ALLOW_FALLBACK compiler flag for whisper.cpp in Build Phases:
Then follow the Core ML support section of readme for convert the model.
In this project, it also added -O3 -DNDEBUG to Other C Flags, but adding flags to app proj is not ideal in real world (applies to all C/C++ files), consider splitting xcodeproj in workspace in your own project.
Metal
You can also enable Metal to make the inference run on the GPU of your device. This might or might not be more efficient compared to Core ML depending on the model and device that you use.
To enable Metal, just add -DGGML_USE_METAL instead off the -DWHISPER_USE_COREML flag and you are ready.
This will make both the Encoder and the Decoder run on the GPU.
If you want to run the Encoder with Core ML and the Decoder with Metal then simply add both -DWHISPER_USE_COREML -DGGML_USE_METAL flags. That's all!