---
title: MaskWAM-real-ckpt
canonical_url: "https://www.modelscope.ai/models/hanyangyu1021/MaskWAM-real-ckpt"
md_url: "https://www.modelscope.ai/models/hanyangyu1021/MaskWAM-real-ckpt.md"
repository: hanyangyu1021/MaskWAM-real-ckpt
last_updated: 2026-10-13
license: apache-2.0
base_model:
  - Wan-AI/Wan2.2-TI2V-5B
base_model_relation: finetune
parameters: 5.6B
tensor_type:
  - BF16
library_name:
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 32
stars: 0
---

# MaskWAM-real-ckpt

> MaskWAM-real-ckpt - An open-source model by hanyangyu1021 on ModelScope. MaskWAM fine-tuned on real-world XTrainer data for mask-guided control.

hanyangyu1021/MaskWAM-real-ckpt is a 5.6B-parameter machine learning model on ModelScope. licensed under apache-2.0. derived from Wan-AI/Wan2.2-TI2V-5B.

- **Repository**: hanyangyu1021/MaskWAM-real-ckpt
- **License**: apache-2.0
- **Parameters**: 5.6B
- **Base model**: Wan-AI/Wan2.2-TI2V-5B
- **Downloads**: 32
- **Stars**: 0
- **Last updated**: 2026-10-13

Source: https://www.modelscope.ai/models/hanyangyu1021/MaskWAM-real-ckpt

---

# MaskWAM Real-Robot Checkpoint

This is the real-robot/XTrainer checkpoint for
[MaskWAM](https://github.com/hanyangyu1021/maskwam).

The checkpoint uses:

- model type: `mask_wan_mot_real`
- image size: `384 × 320`
- action chunk: 48 steps
- first-frame mask prompting
- future RGB and mask prediction

## Files

- `config.json`: model architecture and inference configuration
- `model.safetensors`: MaskWAM policy weights

WAN2.2, SAM3, normalization statistics, and example data are distributed
separately. Follow the
[MaskWAM installation and Playground instructions](https://github.com/hanyangyu1021/maskwam#installation)
to configure these external assets.

## License

This checkpoint contains weights derived from
[Wan2.2-TI2V-5B](https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B)
and is distributed under the Apache License 2.0. The MaskWAM source code is
distributed separately under the MIT License. See `LICENSE` for the checkpoint
license terms.

## Download

```bash
pip install -U modelscope-hub
ms download hanyangyu1021/MaskWAM-real-ckpt \
  --local-dir "$MASKWAM_CHECKPOINT_ROOT/MaskWAM-real-ckpt"
```

Set `MASKWAM_CHECKPOINT_ROOT` before launching the Playground:

```bash
export MASKWAM_CHECKPOINT_ROOT=/path/to/checkpoints
scripts/playground.sh --host 127.0.0.1 --port 7860
```