Question: two decoder

[AprilXiaoyanLiu]

Hi willwhitney,

Thanks for providing this code. It's very useful to me. After I read the paper and code, I have two questions:

First:

• I think the step is we first train the embedding by combining VAE (encode decoder) and actiondecoder, and save embedding
• Then use saved embedding when training TD3 policy. During the training of TD3 policy, embedding is fixed. Am I correct?

Second:

• For the state action embedding method vae_dyne_sa.py , I don't quite understand why we have two decoder there. Can you help me understand it? Thanks