This repo provides a custom task to be used with Isaac Gym as well as scripts for training and exporting models.

• Nvidia GPU with 470+ drivers (nvidia-smi to make sure these are set up)
• Isaac Gym (tested with version 1.0.preview3)
• IsaacGymEnvs (tested with commit 9656bac7e59b96382d2c5040b90d2ea5c227d56d)

• Download Isaac Gym from Nvidia’s developer portal: https://developer.nvidia.com/isaac-gym
• cd isaacgym/python && pip install -e . to install Python package
• More detailed installation guide in: isaacgym/docs/install.html

• git clone https://github.com/NVIDIA-Omniverse/IsaacGymEnvs (tested with commit 9656bac7e59b96382d2c5040b90d2ea5c227d56d)
• cd IsaacGymEnvs && pip install -e . to install Python package

• train.py: Main script used for training/testing models
• export.py: Script for exporting models as ONNX files
• cfg/task/: YAML configs for each task defining environment parameters and domain randomization
• cfg/train/: YAML configs for each task defining RL model and hyperparameters
• tasks/: Class definitions and assets for each task
• runs/: Model checkpoints and summaries will be saved here

Create a new folder called tasks/{task_name}. The following files will be needed:

• tasks/{task_name}/{task_name}.py: Class defining the task. Rewards, observations, actions are all defined here.
• tasks/{task_name}/assets/: URDF and meshes should be placed here.
• cfg/task/{TaskName}.yaml: Config file defining environment parameters and domain randomization
• cfg/train/{TaskName}PPO.yaml: Config gile defining hyperparameters for training

• python train.py task={task_name} to begin training
• Models are saved as runs/{TaskName}/nn/{checkpoint_name}.pth

• python export.py task={task_name} checkpoint="runs/{TaskName}/nn/{checkpoint_name}.pth" to export
• Model is exported to runs/{TaskName}/nn/{checkpoint_name}.pth.onnx

This repo was designed to be used after installing IsaacGymEnvs (https://github.com/NVIDIA-Omniverse/IsaacGymEnvs) as a Python library. IsaacGymEnvs comes close to being usable as a library, but requires adding your custom task to a map.

The following files are taken from IsaacGymEnvs and modified so that this isn't necessary:

• cfg/config.yaml (minimal changes)
• utils/rlgames_utils.py (minimal changes)
• base/vec_task.py (minimal changes)
• train.py

Instead of looking for tasks in a static map, this code uses pydoc.locate to dynamically load the task class.