README.md

---
datasets: lerobot/piper-collect
library_name: lerobot
license: apache-2.0
model_name: act
pipeline_tag: robotics
tags:
- act
- robotics
- lerobot
---
# The lerobot/piper-collect dataset has only 5 recorded joints + Gripper || Piper Arms have 6 + Gripper
# Training Data
Dataset: lerobot/piper-collect

GPU: 5070ti

Batch_Size: 8

Steps: 100k

Final stats: step:100K || smpl:800K || ep:1K || epch:6.57 || loss:0.100 || grdn:6.494 || lr:1.0e-05 || updt_s:0.106 || data_s:0.002

# 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