Closed idigitopia closed 4 years ago
There's no plan as of yet (unfortunately, I didn't store the best checkpoints and stored only the last checkpoints which severly overfit. I would have to retrain the models to store them). What's the use case you were thinking about?
I was trying to see if the penultimate layer representation of DQN can be used for planning. My initial experiments did not show any positive results so I just wanted to verify that.
Thank you for the update. kind regards,
I was wondering if there are any plans to release pretrained DQN Networks for the Atari Games.