Open krisbrud opened 2 years ago
The Encoder and decoder models of Dreamer expect the input to be a 3-channel 32x32 pixel image. To fix this, DreamerModel
in RLlib should be modified, such that it is compatible with the gym-auv
-environment.
TODO: Encoder:
Decoder:
To get familiar with Dreamer, a first experiment with the algorithm should be done, possibly trying to recreate a result from the original paper on the workstation in question.
DoD:
gym
-environment