Modelpublic

quiet-storm/univla-lerobot-ppnb-eu5-step5000

sourceHugging Faceupdated 4mo agoView on Hugging Face

0likes5downloads

Model Card

joon-stack/univla-lerobot-ppnb-eu5-step5000

Full merged UniVLA checkpoint for LeRobot cache fine-tuning.

—dataset: pick_place_nanobanana
—model: euclidean_plain_5tokens_k16
—base VLA: hf_univla_bridge_euclidean_plain_5tokens_k16_lam30k_8gpu_g256_step050000
—step: 5000
—latent action token length: 5
—codebook size: 16
—predicted action horizon: 10

This repository contains full merged VLA weights and the matching action_decoder-5000.pt.