Closed zhangqianhui closed 7 years ago
@zhangqianhui : There was no special reason. We had a Caffe environment from a previous project which enables a smooth / faster transition to this project. Certainly, compared to other frameworks, Caffe allows one to run the code with in a different "condition" network (by dropping in just a binary and a prototxt) with minimum modification of the code. The Caffe Model Zoo also offers many models for quick "plug and play" reuse.
Depending on the purpose, I think implementing PPGN in Tensorflow is also a good idea if one wants to take advantage of the built-in multi-gpu training feature.
Can you explain it for special reason?