---
title: NAFNet-REDS-width64
canonical_url: "https://www.modelscope.cn/models/mlx-community/NAFNet-REDS-width64"
md_url: "https://www.modelscope.cn/models/mlx-community/NAFNet-REDS-width64.md"
repository: mlx-community/NAFNet-REDS-width64
last_updated: 2026-06-04
license: mit
pipeline_tag: image-to-image
tasks:
  - image-to-image
base_model:
  - megvii-research/NAFNet
base_model_relation: finetune
parameters: 67.9M
tensor_type:
  - F32
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 10
stars: 0
tags:
  - mlx
  - image-restoration
  - deblurring
  - denoising
  - nafnet
---

# NAFNet-REDS-width64

> NAFNet-REDS-width64 - mlx-community 在 ModelScope 开源的模型。NAFNet width64 (MLX) — Image deblurring

mlx-community/NAFNet-REDS-width64 是 ModelScope 魔搭社区上的 67.9M 参数image-to-image模型，采用 mit 许可，基于 megvii-research/NAFNet 构建。

- **Repository**: mlx-community/NAFNet-REDS-width64
- **License**: mit
- **Tasks**: image-to-image
- **Parameters**: 67.9M
- **Base model**: megvii-research/NAFNet
- **Tags**: mlx, image-restoration, deblurring, denoising, nafnet
- **Downloads**: 10
- **Stars**: 0
- **Last updated**: 2026-06-04

Source: https://www.modelscope.cn/models/mlx-community/NAFNet-REDS-width64

---

# NAFNet width64 (MLX) — Image deblurring

Apple MLX port of **[NAFNet](https://github.com/megvii-research/NAFNet)** (Simple Baselines for
Image Restoration, ECCV 2022). Runs on Apple Silicon via [MLX](https://github.com/ml-explore/mlx).

This checkpoint: **REDS** (Image deblurring). width64.

## Usage
```python
from nafnet_mlx import NAFNetConfig
from nafnet_mlx.pipeline import load_model, restore_to_file
m = load_model("model.safetensors", NAFNetConfig.reds_width64())
restore_to_file(m, "input.png", "output.png")
```

## Validation
Faithful NHWC port (SimpleGate, Simplified Channel Attention, channel-axis LayerNorm2d,
UNet + PixelShuffle). PT-vs-MLX full-model parity on a real image ~1e-6.
Uses NAFNetLocal (TLC) local pooling.

## License & attribution
MIT. Derived from megvii-research/NAFNet (MIT). See `NOTICE`.
