Hello, I had used your code to understand how recurrent attention model worked and implemented one on my own. I have modified your code a bit, removing some variables and structured it a bit better. One problem I had while looking was that tensorboard graph was not clear to understand, so I made this code keeping in mind to make a graph that is easy to understand, hoping this would help in future debugging. I have used the hyperparameters from the modified version as stop_gradient was not there in that one for mean_locs.
Hello, I had used your code to understand how recurrent attention model worked and implemented one on my own. I have modified your code a bit, removing some variables and structured it a bit better. One problem I had while looking was that tensorboard graph was not clear to understand, so I made this code keeping in mind to make a graph that is easy to understand, hoping this would help in future debugging. I have used the hyperparameters from the modified version as stop_gradient was not there in that one for mean_locs.