cartPPOle

PPO implementation for CartPole-v1 using Gymnasium and PyTorch.

Development

Enter the project environment, then use the Makefile targets below.

Train

make train

Smaller/larger training runs:

make train_xs
make train_sm
make train_md
make train_lg
make train_xl
make train_xxl

Train with rendering enabled:

make train_visual

The default checkpoint is saved to:

checkpoints/policy.pt

Play a checkpoint

make play

This loads checkpoints/policy.pt and runs one rendered episode.

Aim UI

Training metrics are tracked with Aim. Start the UI with:

make aim