| --- |
| library_name: lerobot |
| license: apache-2.0 |
| base_model: lerobot/pi05_base |
| pipeline_tag: robotics |
| tags: |
| - robotics |
| - lerobot |
| - pi0_5 |
| - pi05 |
| - manipulation |
| - block-transfer |
| - trossen-ai |
| datasets: |
| - TrossenRoboticsCommunity/trossen_ai_stationary_block_transfer |
| --- |
| |
| # pi0.5 — Block Transfer (batch size 8) |
|
|
| Fine-tuned [pi0.5](https://huggingface.co/lerobot/pi05_base) vision-language-action policy |
| for the **block transfer** task on the **Trossen AI Stationary** bimanual platform. |
|
|
| ## Model details |
|
|
| | | | |
| |---|---| |
| | **Architecture** | pi0.5 (PaliGemma `gemma_2b` backbone + `gemma_300m` action expert) | |
| | **Base model** | [`lerobot/pi05_base`](https://huggingface.co/lerobot/pi05_base) | |
| | **Framework** | [LeRobot](https://github.com/huggingface/lerobot) | |
| | **Robot platform** | Trossen AI Stationary (bimanual, 14-DoF) | |
| | **Cameras** | `cam_high`, `cam_low`, `cam_left_wrist`, `cam_right_wrist` (480×640 RGB) | |
| | **Action chunk** | 50 steps, 10 inference steps | |
| | **Precision** | bfloat16 | |
| | **License** | Apache 2.0 | |
|
|
| ## Training |
|
|
| | Hyperparameter | Value | |
| |---|---| |
| | **Batch size** | **8** | |
| | Training steps | 100,000 | |
| | Learning rate | 2e-5 (cosine decay to 2.5e-6 over 30k steps, 1k warmup) | |
| | Optimizer | AdamW (β=(0.9, 0.95), wd=0.01, grad clip 1.0) | |
| | Gradient checkpointing | Enabled | |
| | Hardware | 1× NVIDIA H200 (141 GB VRAM) via [Trossen Cloud Service](https://www.trossenrobotics.com/) | |
| | Dataset | [`TrossenRoboticsCommunity/trossen_ai_stationary_block_transfer`](https://huggingface.co/datasets/TrossenRoboticsCommunity/trossen_ai_stationary_block_transfer) | |
|
|
| ## Usage |
|
|
| Load the policy with LeRobot: |
|
|
| ```python |
| from lerobot.policies.pi0_5.modeling_pi0_5 import PI05Policy |
| |
| policy = PI05Policy.from_pretrained( |
| "TrossenRoboticsCommunity/pi05-block-transfer-bs8" |
| ) |
| ``` |
|
|
| See the [LeRobot pi0.5 docs](https://github.com/huggingface/lerobot) for the full |
| inference / rollout setup on real hardware. |
|
|
| ## Files |
|
|
| | File | Purpose | |
| |---|---| |
| | `config.json` | Policy configuration | |
| | `model.safetensors` | Model weights (~7.5 GB) | |
| | `policy_preprocessor.json` + `*_normalizer_processor.safetensors` | Input normalizer (state/action quantiles) | |
| | `policy_postprocessor.json` + `*_unnormalizer_processor.safetensors` | Output unnormalizer | |
|
|
| ## Related models |
|
|
| - [`TrossenRoboticsCommunity/pi05-block-transfer-bs8`](https://huggingface.co/TrossenRoboticsCommunity/pi05-block-transfer-bs8) — this model |
| - [`TrossenRoboticsCommunity/pi05-block-transfer-bs32`](https://huggingface.co/TrossenRoboticsCommunity/pi05-block-transfer-bs32) — same task, batch size 32 |
|
|
