A framework for writing training for Rocket League bots.
Customizable training for Rocket League bots of any language.
Tutorial: Click Here
- Install Python 3.7 or later
pip install rlbottraining
Features at a glance
RLBotconfig files to support bots in any programming language
- Reproducable training setups with seeded randomness
- Customizable Pass/Fail criteria
- Importing of shots/playlists from BakkesMod training
- Playlists of exercises
- Automatic reloading of both bot and exercise code
- Imports from bakkesmod training
- At the entry point (
run_exercises.py) we decide which
GraderExerciseshould be run and the
config_paths define the RLBot setup to be used (which bots, which mode).
GraderExercisecomposes together the initial game state and how the bot is judged. The responsibility of judging the bot is handled by
Graderdecides whether to continue the exercise or to
Failthe exercise by looking at each tick. A
Gradermay optionally gather metrics (e.g. "time until goal") to help track bot performance as the bot is improved.
- Separation of responsiblity: This repository is designed to make it nice to define new exercises whereas the training API of the
RLBotframework provides the minimum features for any training to occur.
Tips for writing your own exercises:
- Always subclass
Graders will allows you to share termination conditions, safely store state across ticks and provide metrics.
- Compose your
Graders. For instance, by using
- Provide meaningful error messages in your
Graders by subclassing
- Provide an example of striker vs goalie
- Put metrics (
Grader.get_metrics()) into a database
- A way of browsing past runs, including metrics
- Reproducing failed exercises
- Visualization of metrics
- Continous integration of different bots
Release history Release notifications | RSS feed
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.