Target Metal Performance Shader (MacOS and iOS)

mratsim / Arraymancer

A fast, ergonomic and portable tensor library in Nim with a deep learning focus for CPU, GPU and embedded devices via OpenMP, Cuda and OpenCL backends

https://mratsim.github.io/Arraymancer/

Apache License 2.0

1.33k stars 96 forks source link

Target Metal Performance Shader (MacOS and iOS) #32

Open mratsim opened 7 years ago

mratsim commented 7 years ago

The Metal framework is a low-level Apple framework to access GPU-accelerated graphics and compute features.

In particular Metal provides off-the-shelf functions for convolutional neural network.

Docs:

reference: https://developer.apple.com/documentation/metalperformanceshaders
comparison with Apple Accelerate Basic Neural Network: http://machinethink.net/blog/apple-deep-learning-bnns-versus-metal-cnn/

Cool videos:

mratsim commented 7 years ago

It might be that Metal only allows inference, not training.

Also see: Bender "Easily craft fast Neural Networks on iOS! Use TensorFlow models. Metal under the hood."

mratsim commented 6 years ago

Seems like CoreML is the new kid on the block for neural net on iOS: https://developer.apple.com/documentation/coreml