deepLearner

If Apple really wants to fully compete with Nvidia, they need to work with the Swift for TensorFlow folks to have Metal be a backend for TensorFlow on the Mac, using AMD GPUs. This would be extremely beneficial for both Apple and AMD. Right now, nearly all machine learning development is done using Nvidia GPUs, because of CUDA. I would be thrilled if I could the same code that trains my convolutional neural net on my Ubuntu box with my Nvidia GPU and run it, without modification, on a MacBook Pro. I really think Swift will eventually become the primary language for machine learning, so Metal coupled with TensorFlow is a no brainer! Also, might not hurt for Apple to also open source Metal, so it could truly become an industry standard.