About the training of Transformer-Prior

[gyy01230]

Hello, I encountered an issue while using the code you provided to replicate the results of your paper: the .sh file you provided can only train the Transformer-VQ part, but it does not work for the Transformer-Prior part. After reading the source code you provided, I couldn't find the code for calculating the trajectory value as described in your paper. Is the source code you provided complete? If so, could you please explain how the Transformer-Prior is trained?