Yorha-7/spider_bot_training ? reverse-engineered prompt

Build me a Python project for training a 12 DOF spider robot in Isaac Lab with reinforcement learning. I want it to support velocity tracking and trot gait control, with PPO training, multiple parallel simulation environments, and a simple keyboard teleop mode using WASD and QE for movement. It should also include scripts to train, play back a policy, and export checkpoints to JIT and ONNX.

Please make it work for both the smaller spider and the bigger spider version, and keep the robot setup in URDF and USD so it’s easy to update the model. Add clear docs for setup, running training, converting the robot description, and comparing sim and hardware behavior. If there are any Isaac Lab details that change often, look up the current docs online before wiring it up.

I’d also like basic test and utility scripts, plus a clean repo structure so the training code, robot configs, and keyboard control are easy to find and extend later.