Some documentation and comment improvement suggestions

README.md

@@ -2,6 +2,14 @@
 
 Code for the paper [Reducing Exploitability with Population Based Training](https://arxiv.org/abs/2208.05083). We reduce exploitability by *adversarial RL policies* by training against a diverse population of opponents.
 
+## Development Status
+
+This repository was created for the associated research paper and is not actively maintained. No automated tests exist, and the population-based training (PBT) components are still experimental.
+
+## Target Audience
+
+This repository is intended for practitioners with prior reinforcement-learning experience. Familiarity with [RLlib](https://docs.ray.io/en/latest/rllib/index.html) and [gin-config](https://github.com/google/gin-config) is assumed. The codebase is geared towards advanced researchers rather than complete beginners.
+
 
 ## Setup
 
@@ -43,7 +51,7 @@ The settings intended to be configured with gin are:
 - `RLSettings` (`aprl_defense.trial.settings.RLSettings`)
 - Additionally, depending on whether one of these modes is used
     - `selfplay` (`aprl_defense.training_managers.simple_training_manager.SelfplayTrainingManager`)
-    - `single-agent` - no additonal arguments
+    - `single-agent` - no additional arguments
     - `attack` (`aprl_defense.training_managers.simple_training_manager.AttackManager`)
     - `pbt` (`aprl_defense.training_managers.pbt_manager.PBTManager`)
 

src/aprl_defense/common/utils.py

@@ -146,8 +146,8 @@ def spaces_from_env(env):
 
 def get_base_train_config(alg):
     # Normalize actions is not available for multi-agent envs, at least according to
-    # this: https://github.com/ray-project/ray/issues/8518k
-    # For me this problem only occured with SAC, DDPG
+    # this: https://github.com/ray-project/ray/issues/8518
+    # For me this problem only occurred with SAC, DDPG
     config = {"normalize_actions": False}
     return config