---
library_name: lerobot
tags:
  - lerobot
  - robotics
  - imitation-learning
  - diffusion-policy
  - aloha
  - bimanual
datasets:
  - lerobot/aloha_sim_insertion_human
---

# Decoupled Bimanual Diffusion — ALOHA Insertion — Shared State

EMA inference checkpoint for a decoupled bimanual diffusion policy trained on
[`lerobot/aloha_sim_insertion_human`](https://huggingface.co/datasets/lerobot/aloha_sim_insertion_human).

## Configuration

- Training steps: 20,000
- Batch size: 32
- Training seed: 1000
- Communication mode: `none`
- State routing: `shared` (both arm branches receive the full 14-dimensional state)
- Actions: two 7-dimensional arm actions
- Shared visual input: `observation.images.top` at 480 x 640
- Observation steps: 2
- Prediction horizon: 64
- Action steps: 32
- Checkpoint: EMA weights plus LeRobot preprocessor and postprocessor

## Evaluation

Evaluated in `AlohaInsertion-v0` with 400-step episodes and synchronous batches
of 50 rollouts. Across seeds 1000 and 2000 (100 episodes total):

- Success: **16/100 (16%)**
- Average reward sum: **218.80**
- Average maximum reward: **2.36**

Per-seed success was 7/50 (seed 1000) and 9/50 (seed 2000).

## Notes

This is a research checkpoint from a custom LeRobot fork implementing decoupled
bimanual diffusion. It is intended for the matching policy implementation and
ALOHA simulation setup. Results may vary with environment and dependency versions.