This is an example RLDS dataset builder adapted from
https://github.com/kpertsch/rlds_dataset_builder for training Interleave-VLAs. This repo aims to illustrate the data format of Interleave-VLA.
We assume interleave Open X-Embodiment data is first extracted and stored in pickles. The folder to pickles should be set in config.py.
Next, select dataset_name in openx_interleave.py to your desired dataset.
Finally, run
tfds build --overwrite
