When starting the bot, there will be run folders "rlgym-ppo-run-[X]". Pay attention to the folders within, move run folders with the lower numbered folders into ./checkpoint_reversion befor running the bot again. This allows the learning continuation.
NOTE The bigger the number in the run folder, the newer the iteration is. -potential scripting advantage
Keep track of rewards and other settings between each run (and commit) here. Follow a file format of log-[yyymmdd]-[revision].md.
Delete everything in ./data and /wandb before running again.
- Learner Settings: learner_settings.md
- Rewards: rewards.md
Change render inside learner settings in example.py to True to watch the sim in realtime