Extended README

metastable_baselines2/README.md

@@ -6,13 +6,10 @@
 
 An extension to Stable Baselines 3. Based on Metastable Baselines 1.
 
-During training of a RL-Agent we follow the gradient of the loss, which leads us to a minimum. In cases where the found minimum is merely a local minimum, this can be seen as a _false vacuum_ in our loss space. Exploration mechanisms try to let our training procedure escape these _stable states_: Making them _metastable_.
-
-In order to archive this, this Repo contains some extensions for [Stable Baselines 3 by DLR-RM](https://github.com/DLR-RM/stable-baselines3)  
-These extensions include:
+This repo provides:
 
 - An implementation of ["Differentiable Trust Region Layers for Deep Reinforcement Learning" by Fabian Otto et al. (TRPL)](https://arxiv.org/abs/2101.09207)
-- Support for Prior Conditioned Annealing
+- Support for Prior Conditioned Annealing (WIP)
 - Support for Contextual Covariances (Planned)
 - Support for Full Covariances (Planned)
 
@@ -33,6 +30,23 @@ Then install this repo as a package:
 pip install -e .
 ```
 
+## Usage
+
+TRPL can be used just like SB3's PPO:
+
+```
+import gymnasium as gym
+from metastable_baselines2 import TRPL
+
+projection = 'Wasserstein' # or Frobenius or KL
+
+model = TRPL("MlpPolicy", env_id, n_steps=128, seed=0, policy_kwargs=dict(net_arch=[16]), projection_class=projection, projection_kwargs={'mean_bound': mean_bound, 'cov_bound': cov_bound}, verbose=1)
+
+model.learn(total_timesteps=100)
+```
+
+For avaible projection_kwargs have a look at [Metastable Projections](https://git.dominik-roth.eu/dodox/metastable-projections).
+
 ## License
 
 Since this Repo is an extension to [Stable Baselines 3 by DLR-RM](https://github.com/DLR-RM/stable-baselines3), it contains some of it's code. SB3 is licensed under the [MIT-License](https://github.com/DLR-RM/stable-baselines3/blob/master/LICENSE).