Users can train CartPole via:
python train_ppo.py --config ppo.yaml
To train with Dual-clip PPO:
python train_ppo.py --config dual_clip_ppo.yaml
To train with A2C algorithm:
python train_a2c.py
If you want to evaluate the agent during training and save the best model and save checkpoints, try to train with callbacks:
python train_ppo.py --config callbacks.yaml
More details about callbacks can be found in Callbacks.