| tags: | |
| - robotics | |
| - imitation-learning | |
| - act | |
| - isaac-lab | |
| - galaxea | |
| library_name: generic | |
| # ACT Policy for Gearbox Assembly | |
| This model is an **Action Chunking with Transformers (ACT)** policy trained for a gearbox assembly task in Isaac Lab. | |
| ## Model Details | |
| - **Architecture**: ACT (Action Chunking with Transformers) | |
| - **Task**: Gearbox Assembly | |
| - **Framework**: Isaac Lab / PyTorch | |
| - **Dataset**: `sim_gearbox_assembly_demos_filtered` | |
| ## Training Data Usage | |
| The model was trained using **all available episodes** from the provided dataset. | |
| ### Important Data Characteristics | |
| - **Process Simplification**: The **pin insertion step was omitted** in all training episodes. The task focuses on the steps following pin insertion (e.g., gear mounting, cover installation). | |
| - **Data Quality Note**: The training set deliberately includes one known **failed episode** where a pin fell over during the process, causing the subsequent cover installation to fail. This episode was **not filtered out** and was used in training. | |
| ## Hyperparameters | |
| The model was trained with the following configuration: | |
| | Parameter | Value | | |
| |:---|:---| | |
| | **Policy Class** | ACT | | |
| | **KL Weight** | 10 | | |
| | **Chunk Size** | 100 | | |
| | **Hidden Dim** | 512 | | |
| | **Feedforward Dim** | 3200 | | |
| | **Batch Size** | 32 | | |
| | **Learning Rate** | 1e-5 | | |
| | **Epochs** | 10000 | | |
| | **Seed** | 0 | | |
| | **Save Interval** | 200 | | |
| | **Early Stopping** | 3000 | | |
| ## Usage | |
| This model is designed to be loaded with the ACT policy wrapper in the Galaxea Lab framework for the `GalaxeaLab-GearboxAssembly` environment. | |