Instructions to use maniguard-review/smolvla-stack with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LeRobot
How to use maniguard-review/smolvla-stack with LeRobot:
# See https://github.com/huggingface/lerobot?tab=readme-ov-file#installation for more details git clone https://github.com/huggingface/lerobot.git cd lerobot pip install -e .[smolvla]
# Launch finetuning on your dataset python lerobot/scripts/train.py \ --policy.path=maniguard-review/smolvla-stack \ --dataset.repo_id=lerobot/svla_so101_pickplace \ --batch_size=64 \ --steps=20000 \ --output_dir=outputs/train/my_smolvla \ --job_name=my_smolvla_training \ --policy.device=cuda \ --wandb.enable=true
# Run the policy using the record function python -m lerobot.record \ --robot.type=so101_follower \ --robot.port=/dev/ttyACM0 \ # <- Use your port --robot.id=my_blue_follower_arm \ # <- Use your robot id --robot.cameras="{ front: {type: opencv, index_or_path: 8, width: 640, height: 480, fps: 30}}" \ # <- Use your cameras --dataset.single_task="Grasp a lego block and put it in the bin." \ # <- Use the same task description you used in your dataset recording --dataset.repo_id=HF_USER/dataset_name \ # <- This will be the dataset name on HF Hub --dataset.episode_time_s=50 \ --dataset.num_episodes=10 \ --policy.path=maniguard-review/smolvla-stack - Notebooks
- Google Colab
- Kaggle
SmolVLA — Stack
ICLR 2027 — Review-only mirror
This repository is an anonymized copy of the ManiGuard resources prepared for peer review.
Citation: If you use these resources in your research, please cite the ManiGuard paper. Full citation details will be provided after the double-blind review period.
ManiGuard main-evaluation checkpoint for stack_retrieve.
Selected training step: 10,360.
Download and evaluate
hf download maniguard-review/smolvla-stack --local-dir /path/to/smolvla-stack
python -m maniguard.serve.smolvla_native --checkpoint /path/to/smolvla-stack --port 8000
Use the included processor configuration and normalization statistics. Use the ManiGuard family evaluation configuration for camera and action decoding.
License
The base model lerobot/smolvla_base
is released under the Apache License, Version 2.0. ManiGuard's fine-tuning
contributions are also released under Apache-2.0.
- Downloads last month
- 45
Model tree for maniguard-review/smolvla-stack
Base model
lerobot/smolvla_base