Create an --encoder flag in _run_ppo.py that accepts a TF model ckpt to an AutoEncoder. Load the encoder, create an observation wrapper that applies the encoding to (pre-framestacked!) images, and then runs PPO.
Blocked on: #25
Deliverable: learning curve for PPO.
Create an --encoder flag in _run_ppo.py that accepts a TF model ckpt to an AutoEncoder. Load the encoder, create an observation wrapper that applies the encoding to (pre-framestacked!) images, and then runs PPO.
Blocked on: #25 Deliverable: learning curve for PPO.