Add pipeline tag
Browse filesHi! I'm Niels from the Hugging Face team. This PR adds the `pipeline_tag: unconditional-image-generation` to the model card's metadata. This will help users discover the model when filtering by task on the Hub.
README.md
CHANGED
|
@@ -1,11 +1,12 @@
|
|
| 1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
tags:
|
| 3 |
- flow-matching
|
| 4 |
- pixel-diffusion
|
| 5 |
- pixel-generation
|
| 6 |
-
datasets:
|
| 7 |
-
- ILSVRC/imagenet-1k
|
| 8 |
-
license: apache-2.0
|
| 9 |
---
|
| 10 |
|
| 11 |
# Asymmetric Flow Models
|
|
@@ -27,6 +28,9 @@ Stanford University
|
|
| 27 |
|
| 28 |
|
| 29 |
|
|
|
|
|
|
|
|
|
|
| 30 |
## Citation
|
| 31 |
```
|
| 32 |
@article{chen2026asymmetric,
|
|
|
|
| 1 |
---
|
| 2 |
+
datasets:
|
| 3 |
+
- ILSVRC/imagenet-1k
|
| 4 |
+
license: apache-2.0
|
| 5 |
+
pipeline_tag: unconditional-image-generation
|
| 6 |
tags:
|
| 7 |
- flow-matching
|
| 8 |
- pixel-diffusion
|
| 9 |
- pixel-generation
|
|
|
|
|
|
|
|
|
|
| 10 |
---
|
| 11 |
|
| 12 |
# Asymmetric Flow Models
|
|
|
|
| 28 |
|
| 29 |
|
| 30 |
|
| 31 |
+
## Abstract
|
| 32 |
+
Flow-based generation in high-dimensional spaces is difficult because velocity prediction requires modeling high-dimensional noise. Asymmetric Flow Modeling (AsymFlow) is a rank-asymmetric velocity parameterization that restricts noise prediction to a low-rank subspace while keeping data prediction full-dimensional. On ImageNet 256$\times$256, AsymFlow achieves a leading 1.57 FID, outperforming prior DiT/JiT-like pixel diffusion models by a large margin.
|
| 33 |
+
|
| 34 |
## Citation
|
| 35 |
```
|
| 36 |
@article{chen2026asymmetric,
|