TL;DR: TensorFlow for production (and probably work too, like Roman Trusov said), PyTorch for research and fun and Caffe2 for edge device inference (eg an ...
確定! 回上一頁