Datasets:

---

vipra-project

/

cotrain-vqgan-vision-cache

like 0

Follow

ViPRA 1

This PR significantly enhances the dataset card for the vipra-project/cotrain-dynamics14 dataset by adding comprehensive information derived from the paper, project page, and GitHub repository.

Key changes include:

• Metadata: Added task_categories as robotics and image-feature-extraction to improve discoverability.
• Links: Incorporated direct links to the ViPRA paper, project page, and GitHub repository.
• Abstract & Overview: Included the paper's abstract and the project overview from the GitHub README to provide a quick understanding of ViPRA.
• Dataset Description: Provided detailed information about the cotrain-dynamics14 dataset's composition, purpose in ViPRA pretraining, and the optional VQGAN vision cache.
• Dataset Structure Example: Added a "Sample JSONL Entry" to clearly demonstrate the format of data within the dataset.
• Sample Usage: Integrated a Python code snippet from the GitHub README's deployment section, showcasing how to use the ViPRAClient to obtain robot actions.
• Citation: Included the official BibTeX citation for the ViPRA paper.
• License: Explicitly stated the Apache License 2.0 at the end of the card.

These updates make the dataset card much more informative, discoverable, and user-friendly for the community.