Closed BinhMinhs10 closed 3 years ago
It depends on how many iterations you want to train, if you use the same number of iteration as the paper (which is the default in my implementation), it takes 4 day on 16* TPU v3 s, according to the paper.
It depends on how many iterations you want to train, if you use the same number of iteration as the paper (which is the default in my implementation), it takes 4 day on 16* TPU v3 s, according to the paper.