---
library_name: lerobot
tags:
- act
- robotics
- lerobot
- bimanual
---

# act_insert_candle

Standard LeRobot ACT trained from scratch on bimanual SO101 data.

| | |
|---|---|
| **Policy** | ACT (`policy.type=act`) |
| **Dataset** | [villekuosmanen/armnetbench_insert_candle](https://huggingface.co/datasets/villekuosmanen/armnetbench_insert_candle) |
| **Task** | `insert_candle` |
| **Action dim** | 12 (bimanual) |
| **Cameras** | `top`, `left_wrist`, `right_wrist` |
| **Training** | GCloud A100-80GB (`alpha-act-a100-3`), batch 16, action horizon 30 |
| **Loss @ 10k** | 0.149 |
| **W&B project** | [act_insert_candle](https://wandb.ai/pravsels/act_insert_candle) |
| **W&B run** | [7dc5r8fi](https://wandb.ai/pravsels/act_insert_candle/runs/7dc5r8fi) |

## Checkpoints

| Step | Path |
|------|------|
| 010000 | `checkpoints/010000/pretrained_model/` |
| 008000 | `checkpoints/008000/pretrained_model/` |

## Usage

```python
from lerobot.policies.act.modeling_act import ACTPolicy
policy = ACTPolicy.from_pretrained(
    "pravsels/act_insert_candle",
    subfolder="checkpoints/010000/pretrained_model",
)
```