https://github.com/sophiaas/rlbase
Modular Deep RL infrastructure in PyTorch
https://github.com/sophiaas/rlbase
a2c deep-reinforcement-learning hierarchical-reinforcement-learning option-critic policy-gradient ppo proximal-policy-optimization pytorch-rl reinforcement-learning
Last synced: 20 days ago
JSON representation
Modular Deep RL infrastructure in PyTorch
- Host: GitHub
- URL: https://github.com/sophiaas/rlbase
- Owner: sophiaas
- Created: 2019-07-19T00:20:26.000Z (almost 7 years ago)
- Default Branch: master
- Last Pushed: 2024-07-25T10:59:54.000Z (almost 2 years ago)
- Last Synced: 2025-10-09T15:43:52.130Z (8 months ago)
- Topics: a2c, deep-reinforcement-learning, hierarchical-reinforcement-learning, option-critic, policy-gradient, ppo, proximal-policy-optimization, pytorch-rl, reinforcement-learning
- Language: Python
- Homepage:
- Size: 2.3 MB
- Stars: 3
- Watchers: 2
- Forks: 0
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# rlbase
Modular Deep RL infrastructure in PyTorch.
Currently implemented algorithms include PPO and PPOC (Option-Critic trained with Proximal Policy Optimization).
To train, first write a config file and save to `configs/ppo` or `configs/ppoc.` A config file specifies parameters in config objects (found in `core/config.py`) for each component of the model and experiment, such as optimization hyperparameters, network architectures, environments, logging behavior, etc. See example config at `configs/ppo/lightbot.py`
The specified configuration and algorithm can then be run by calling:
`python train.py --config [config file name] --algo [ppo or ppoc]`
See `train.py` for other options.
To evaluate a pre-trained model, run:
`python evaluate.py --model_dir [logging directory] --episode [episode to load checkpoint from] --n_eval_steps [number of steps to evaluate]`
Data from the evaluated model is saved to `[logging directory]/evaluate/`