sean1295/libero_all

This dataset is associated with the paper TBD-VLA: Temporal Block Diffusion Vision Language Action Model.

TBD-VLA is a discrete token-based Vision-Language-Action (VLA) framework that incorporates block diffusion to enable temporal action generation. It partitions action sequences into temporal blocks and performs masked discrete diffusion within each block, while maintaining autoregressive generation across blocks.

• Project Page: https://tbd-vla.github.io
• GitHub Repository: https://github.com/TBD-VLA/lerobot

Sample Usage

You can use this dataset to train a TBD-VLA policy using the LeRobot library. Below is an example training command for the LIBERO environment:

python src/lerobot/scripts/lerobot_train.py \
  --policy.type=tbdvla \
  --output_dir=/$OUTPUT_DIR \
  --dataset.repo_id=sean1295/libero_all \
  --job_name=tbdvla_experiment \
  --steps=150000 \
  --batch_size=4 \
  --save_freq=20000 \
  --log_freq=1000 \
  --policy.device=cuda \
  --policy.n_bins=512 \
  --policy.block_temporal_size=4 \
  --policy.n_diffusion_steps=2 \
  --policy.gripper_dims=[-1] \
  --policy.chunk_size=16 \
  --policy.n_action_steps=16 \
  --policy.gradient_checkpointing=true \
  --policy.push_to_hub=false \
  --wandb.enable=false

Citation

@article{lee2026tbdvlatemporalblockdiffusion,
      title={TBD-VLA: Temporal Block Diffusion Vision Language Action Model},
      author={Lee, Sung-Wook and Kang, Xuhui and Kuo, Yen-Ling},
      journal={arXiv preprint},
      year={2026},
}