| library_name: pytorch | |
| tags: | |
| - robotics | |
| - manipulation | |
| - imitation-learning | |
| - act | |
| - lerobot | |
| # ACT-Pick | |
| ACT-Pick is an **Action Chunking Transformer (ACT)** policy trained using | |
| **LeRobot** for a tabletop **object pickup task**. | |
| ## Model Details | |
| - Policy Type: ACT | |
| - Learning Method: Imitation Learning | |
| - Framework: PyTorch | |
| ## Files | |
| - `model.safetensors` — trained policy weights | |
| - `config.json` — model configuration | |
| - `policy_preprocessor.json` — observation normalization config | |
| - `policy_preprocessor_step_3_normalizer_processor.safetensors` — observation normalizer | |
| - `policy_postprocessor.json` — action unnormalization config | |
| - `policy_postprocessor_step_0_unnormalizer_processor.safetensors` — action unnormalizer | |
| - `train_config.json` — training configuration | |
| ## Usage | |
| ```bash | |
| lerobot-eval \ | |
| --policy.repo_id=NEW_HF_USERNAME/ACT-Pick \ | |
| --policy.device=cuda | |
| ``` | |