Closed psyxusheng closed 5 years ago
Not sure how it is done in tensorflow, but in pytorch, you can write a fuction to share the dropout mask over time as in the following line.
https://github.com/Sunnydreamrain/IndRNN_pytorch/blob/master/action_recognition/Indrnn_action_network.py#L26
Anyway, the key is to share the dropout mask over time, which should be easily done.
Thank you
Looking forward to reply~~