regarding paper describing this model

[wenouyang]

Hi, thanks for sharing the code, Are there any research paper discussing this model?

Thanks.

[ASvyatkovskiy]

Following paper discusses some aspects of the model (mainly, mixed precision float training):
https://dl.acm.org/citation.cfm?id=3146358

It also discusses the distributed training algorithm, learning rate scheduling, and the neural network architecture.

We have another paper which is now submitted to a journal, it will become available soon.