---
datasets: arclabmit/xarm7_beavrsim_shellgame_dataset
library_name: lerobot
license: apache-2.0
pipeline_tag: robotics
tags:
- robotics
- lerobot
- act
model-index:
- name: xarm7_act_beavrsim_shellgame_model
  results:
  - task:
      type: robotics
      name: Robotic Manipulation
    dataset:
      name: beavr_sim
      type: simulation
    metrics:
    - type: success_rate
      value: 10.299999999999999
      name: Success Rate
    - type: reward
      value: -0.05
      name: Avg Reward
---

# Model Card for act

<!-- Provide a quick summary of what the model is/does. -->


[Action Chunking with Transformers (ACT)](https://huggingface.co/papers/2304.13705) is an imitation-learning method that predicts short action chunks instead of single steps. It learns from teleoperated data and often achieves high success rates.


This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).

---

## How to Get Started with the Model

For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
Below is the short version on how to train and run inference/eval:

### Train from scratch

```bash
lerobot-train \
  --dataset.repo_id=${HF_USER}/<dataset> \
  --policy.type=act \
  --output_dir=outputs/train/<desired_policy_repo_id> \
  --job_name=lerobot_training \
  --policy.device=cuda \
  --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
  --wandb.enable=true
```

_Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`._

### Evaluate the policy/run inference

```bash
lerobot-record \
  --robot.type=so100_follower \
  --dataset.repo_id=<hf_user>/eval_<dataset> \
  --policy.path=<hf_user>/<desired_policy_repo_id> \
  --episodes=10
```

Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.

---

## Model Details

- **License:** apache-2.0


<!-- EVAL_RESULTS_START -->
## Evaluation Results
*Evaluated on 2026-02-05 09:45*

| Metric | Value |
| :--- | :--- |
| **Success Rate** | 10.3% |
| **Average Reward** | -0.050 |
| **Max Reward (Avg)** | 1.030 |
| **Episodes** | 1000 |
| **Eval Speed** | 2.47 s/ep |
| **Seed** | 26 |

> [!TIP]
> Detailed per-episode results can be found in [eval/eval_info.json](./eval/eval_info.json).

<!-- EVAL_RESULTS_END -->