question about replay buffer and the number of BC_model

[8JasonStatham8]

In the code for generating the replay buffer mentioned in Issue 1, we need 'representations'(https://github.com/MLforHealth/rl_representations/blob/main/scripts/experiment.py#L493-L494). But the 'representations' is generated by (https://github.com/MLforHealth/rl_representations/blob/main/scripts/experiment.py#L427-L471) Does we also need it for replay buffer? And I have another question : Does we need generate each replay buffer (number of 7) for every kind of autoencoder, and use these 7 replay buffer generate 7 different BC_model?