| --- |
| license: mit |
| library_name: mage_flow |
| pipeline_tag: image-to-image |
| base_model: microsoft/Mage-Flow-Edit |
| base_model_relation: quantized |
| tags: |
| - xpo3 |
| - mage-flow |
| - mage-flow-edit |
| - image-editing |
| - image-to-image |
| - nvfp4 |
| - fp8 |
| - blackwell |
| - quantization |
| --- |
| |
| <h1 align="center">Mage-Flow Edit 路 XPO3 NVFP4</h1> |
|
|
| <p align="center">A 30-step native Blackwell W4A4 release of Mage-Flow Edit.</p> |
|
|
| <p align="center"> |
| <a href="https://huggingface.co/microsoft/Mage-Flow-Edit">Original model</a> |
| 路 <a href="https://huggingface.co/Comfy-Org/Mage-Flow">ComfyUI source weights</a> |
| 路 <a href="https://huggingface.co/starsfriday/Mage-Flow-Edit-FP8">FP8 text encoder source</a> |
| 路 <a href="https://huggingface.co/ajh-code/Mage-Flow-Turbo-XPO3-NVFP4">XPO3 text-to-image sibling</a> |
| 路 <a href="https://huggingface.co/ajh-code/Mage-Flow-Edit-Turbo-XPO3-NVFP4">XPO3 Edit-Turbo sibling</a> |
| </p> |
|
|
| --- |
|
|
| ## Download |
|
|
| | File | Purpose | Size | |
| |:--|:--|--:| |
| | `Mage-Flow-Edit-XPO3-NVFP4.safetensors` | Self-contained Edit transformer | 4.65 GB | |
| | `qwen3vl_4b_fp8_scaled.safetensors` | Shared scaled-FP8 text encoder | 5.24 GB | |
| | `Mage-Flow-VAE.safetensors` | Shared Mage VAE | 345 MB | |
|
|
| The root `config.json` is included for Hugging Face model/download accounting. |
|
|
| ## Quick start |
|
|
| Linux, Python 3.11, CUDA 13, and an NVIDIA Blackwell SM120 GPU are required. |
|
|
| ```bash |
| hf download ajh-code/Mage-Flow-Edit-XPO3-NVFP4 --local-dir mage-edit-xpo3 |
| cd mage-edit-xpo3 |
| python3.11 -m venv .venv |
| . .venv/bin/activate |
| pip install -r requirements.txt |
| python edit.py reference.jpg "Change the background; keep the subject unchanged." --output edited.png |
| ``` |
|
|
| The promoted profile is 30 steps, CFG 5, image-only fused GELU, calibrated |
| FP4 bridge on all 12 blocks, direct-HND attention on steps 7-29, and exact |
| SDPA fallback on steps 0-6. Every optimization has a CLI disable switch. |
|
|
| ## Measured performance |
|
|
| | GPU 路 512-long-edge edit | BF16 hot denoise | XPO3 hot denoise | Speedup | |
| |:--|--:|--:|--:| |
| | RTX 5080 路 Dog background | 4.2491 s | 2.5709 s | 1.65脳 | |
| | RTX 5080 路 Fruit tablecloth | 4.4839 s | 2.6815 s | 1.67脳 | |
| | RTX 5060 Ti 路 Dog background | 8.6019 s | 4.7614 s | 1.81脳 | |
| | RTX 5060 Ti 路 Fruit tablecloth | 9.3282 s | 5.1300 s | 1.82脳 | |
|
|
| Hot denoise timings are local matched measurements, not universal end-to-end claims. The RTX 5060 Ti runs also used 31-33% less peak allocation than BF16. |
|
|
| ## Validated scope |
|
|
| - Same-seed dog and fruit edits were deterministic across fresh and hot runs. |
| - The standalone package reproduced both approved RTX 5080 composed outputs pixel-exactly. |
| - Each release-only edit recorded 552 direct routes, 168 exact fallback routes, all 12 bridge blocks, and restored every temporary patch. |
|
|
| This is an XPO3 runtime package, not a generic portable quantization format. |
| The bundled native libraries target Linux x86-64, Python 3.11, PyTorch |
| 2.13.0+cu130, and SM120. |
|
|
| MIT applies to the XPO3 package and Mage-derived runtime. The scaled-FP8 |
| Qwen3-VL component and bundled SpargeAttn code are Apache-2.0; see |
| `THIRD_PARTY_NOTICES.md`. |
|
|
