Release GeneralVLA models and robotic datasets on Hugging Face #1
Description
Hi @SteveZeyuZhang 🤗
Niels here from the open-source team at Hugging Face. I discovered your work through Hugging Face's daily papers as yours got featured: https://huggingface.co/papers/2602.04315.
The paper page lets people discuss your paper and find related artifacts. I noticed on your project page that you've already set up a Hugging Face repository for the model (https://huggingface.co/AIGeeksGroup/GeneralVLA), which is awesome!
It'd be great to make the pre-trained checkpoints (the ASM, 3DAgent, and control policy) and the generated trajectory datasets for the 14 tasks mentioned available on the 🤗 Hub to improve their discoverability and visibility. We can add tags so that people find them when filtering https://huggingface.co/models and https://huggingface.co/datasets.
Uploading models
If the model weights are ready for release, you can find a guide here: https://huggingface.co/docs/hub/models-uploading. For custom VLA models, leveraging hf_hub_download or the PyTorchModelHubMixin is a great way to let users download and use the weights directly.
Uploading datasets
Since GeneralVLA can automatically generate robotic data, it would be amazing to host those demonstrations on 🤗 as well. This allows researchers to easily load them:
from datasets import load_dataset
dataset = load_dataset("AIGeeksGroup/your-dataset-name")Besides that, our dataset viewer allows people to explore the trajectories directly in the browser.
Let me know if you're interested or need any help regarding this! We can also help link the repositories to the paper page so they show up for everyone following your work.
Cheers,
Niels
ML Engineer @ HF 🤗