Tutorial Door Env

This tutorial will walk you through collecting, training, and evaluating a data set for the door environment. This is intended to help users get accustom to the repository and be able to extrapolate this information to new environments.

Install and Setup

Follow the installation guide.

Collecting Data

Want to skip this step? Download our pre-gathered door environment data set.

Follow the dataset guide, setting up --env=door.

Training

To start training, you can run the training script.

python scripts/train.py  --env=door --data=/path/to/datasets/demo250

Common Issue: If you are running out of memory, try changing the batch size in the configuration file.

This will generate the checkpoint files inside the save/ directory.

Evaluation

Want to skip this step? Download our pre-trained door Visuomotor Policies.

Evaluation is useful for seeing the status of your visuomotor policy. The following script will run the door opening environment with the generated policy.

python3 scripts/evaluate.py --env=door --path=/path/to/checkpoints/door_