This work combines Active Pre-Training with an On-Policy algorithm, Phasic Policy Gradient.
Is used to pre-train a model free algorithm before defining a downstream task. It calculates the reward based on an estimatie of the particle based entropy of states. This reduces the training time if you want to define various tasks - i.e. robots for a warehouse.
Improved Version of Proximal Policy Optimization, which uses auxiliary epochs to train shared representations between the policy and a value network.
Install via pip pip install unsupervised-on-policy.
Configure by changing values in the config at run_train.py.
Supports the key words load skip_pretrain and model_path.
Example usage:
python3 -m unsupervised_on_policy.run_train --skip_pretrain --prefix test --model_path /home/lars/apt_ppg