some issue about pretrain

Thanks for your job! I have some questiones to ask:  In pretrain phase, you use the VideoEncoder + Videodecoder? 
Why you don not remove the temporal module as metioned in the paper ? Is there a portion of the code that hasn't been uploaded?