# LingBot-VA

LingBot-VA is an **autoregressive video-action world-model policy** built on the **Wan2.2**
video-diffusion stack. It interleaves, in one autoregressive sequence, the prediction of
future **video latents** and **robot actions** ("VA" = Video-Action). The LeRobot
integration wires LingBot-VA into the standard training, evaluation and processor
interfaces.

## Model Overview

LingBot-VA is a **dual-stream "mixture-of-transformers"**: a video/latent stream
(`patch_embedding_mlp → blocks → proj_out`) and an action stream
(`action_embedder → blocks → action_proj_out`) share the same 30 transformer blocks and
text conditioning. Actions are produced by the dedicated `action_proj_out` head — they are
**not** decoded from predicted pixels, though video and action are co-trained.

| Component | Class | Role |
|---|---|---|
| DiT backbone (trainable) | `WanTransformer3DModel` | ~5B-param dual-stream transformer (the only weights stored in the LeRobot checkpoint). |
| VAE (frozen) | `AutoencoderKLWan` | Wan2.2 VAE, `z_dim=48`. Lazy-pulled from the source repo. |
| Text encoder (frozen) | `UMT5EncoderModel` | UMT5-XXL, `d_model=4096`. Lazy-pulled from the source repo. |

At inference the policy runs an autoregressive loop per chunk: it denoises the video-latent
stream (CFG, ~20 steps) and the action stream (~50 steps) with two independent
flow-matching schedulers, maintaining a KV cache across chunks. Real observed keyframes are
fed back into the KV cache as the chunk is executed (closed-loop world modeling).

### What the LeRobot Integration Covers

- Standard `policy.type=lingbot_va` configuration through LeRobot.
- Checkpoint conversion from the released HuggingFace checkpoints.
- Autoregressive dual-stream inference behind the standard `select_action` interface
  (single-environment eval, `--eval.batch_size=1`).
- Opt-in saving of the policy's **predicted (imagined) videos** during eval / training.
- Evaluation with `lerobot-eval` on the LIBERO benchmark.

Training (the flow-matching dual-stream loss + latent dataset) is part of a follow-up
training port and is not yet wired into `lerobot-train`.

## Installation

1. Install LeRobot by following the [Installation Guide](./installation).
2. Install the LingBot-VA extra (brings in `diffusers>=0.36` for the Wan2.2 stack):

```bash
pip install -e ".[lingbot_va]"
# For LIBERO evaluation (Linux only):
pip install -e ".[lingbot_va,libero]"
```

## Checkpoint Conversion

The released checkpoints are diffusers-style directories
(`robbyant/lingbot-va-base`, `robbyant/lingbot-va-posttrain-robotwin`,
`robbyant/lingbot-va-posttrain-libero-long`). Convert one to LeRobot format with:

```bash
python -m lerobot.policies.lingbot_va.convert_lingbot_va_checkpoints \
    --checkpoint robbyant/lingbot-va-posttrain-libero-long \
    --variant libero \
    --output_dir outputs/lingbot_va_libero_long
```

**Packaging:** only the trainable ~5B transformer is stored in the LeRobot
`model.safetensors`. The frozen VAE + UMT5 + tokenizer (~20 GB) are **lazily pulled** from
`config.wan_pretrained_path` at load time (defaults to the source repo). Pass
`--bundle-frozen` to copy those sub-folders next to the converted checkpoint instead.

Run conversion on a Linux machine with a CUDA GPU and enough RAM/VRAM to materialize the
transformer.

## Evaluation (LIBERO)

```bash
lerobot-eval \
    --policy.path=outputs/lingbot_va_libero_long \
    --env.type=libero --env.task=libero_10 \
    --eval.n_episodes=50 --eval.batch_size=1 \
    --output_dir=outputs/eval/lingbot_va_libero
```

LingBot-VA's streaming inference (KV cache + observed-keyframe feedback) is implemented for
single-environment eval; use `--eval.batch_size=1`.

### Saving predicted (imagined) videos

Set `--policy.save_predicted_video=true` to additionally VAE-decode the predicted video
latents and write `pred_episode_*.mp4` next to the env-rendered `eval_episode_*.mp4` videos.
The same flag works for the periodic eval during `lerobot-train`.

## Inference Hyperparameters (LIBERO)

| Key | Value |
|---|---|
| height × width | 128 × 128 |
| cameras | `observation.images.image` (agentview), `observation.images.image2` (eye-in-hand) |
| action channels used | 0–6 (7-DoF arm + gripper) |
| action_per_frame / frame_chunk_size | 4 / 4 |
| attn_window | 30 |
| video / action denoising steps | 20 / 50 |
| guidance_scale / action_guidance_scale | 5 / 1 |
| snr_shift / action_snr_shift | 5.0 / 0.05 |

These are the defaults of `LingBotVAConfig`; override any of them via `--policy.<name>=...`.

## Notes & Limitations

- **Correctness gate:** matching the upstream LIBERO success rate requires validating the
  converted checkpoint on a GPU and tensor-diffing intermediate activations against the
  upstream implementation. The most sensitive parts are the action quantile normalization,
  the camera ordering, the `action_per_frame`/`frame_chunk_size` alignment, and `attn_mode`.
- **Attention backend:** inference uses the `torch` SDPA backend (always available). The
  `flashattn` and `flex` backends are optional; `flex` is only needed for training.
- **Model size:** the DiT is ~5B params and the frozen VAE+UMT5 add ~20 GB; inference needs
  roughly 18–24 GB of VRAM.

## License

LingBot-VA is released under Apache-2.0. See the
[upstream repository](https://github.com/Robbyant/lingbot-va).