Update README.md
Browse files
README.md
CHANGED
|
@@ -28,7 +28,7 @@ us humans when looking at faces, hands, etc. We are working on making these reco
|
|
| 28 |
### Image Sizes
|
| 29 |
Würstchen was trained on image resolutions between 1024x1024 & 1536x1536. We sometimes also observe good outputs at resolutions like 1024x2048. Feel free to try it out.
|
| 30 |
We also observed that the Prior (Stage C) adapts extremely fast to new resolutions. So finetuning it at 2048x2048 should be computationally cheap.
|
| 31 |
-
<img src="https://cdn-uploads.huggingface.co/production/uploads/634cb5eefb80cc6bcaf63c3e/
|
| 32 |
|
| 33 |
## How to run
|
| 34 |
This pipeline should be run together with a prior https://huggingface.co/warp-ai/wuerstchen-prior:
|
|
@@ -55,6 +55,11 @@ output = pipeline(
|
|
| 55 |
).images
|
| 56 |
```
|
| 57 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
## Model Details
|
| 59 |
- **Developed by:** Pablo Pernias, Dominic Rampas
|
| 60 |
- **Model type:** Diffusion-based text-to-image generation model
|
|
|
|
| 28 |
### Image Sizes
|
| 29 |
Würstchen was trained on image resolutions between 1024x1024 & 1536x1536. We sometimes also observe good outputs at resolutions like 1024x2048. Feel free to try it out.
|
| 30 |
We also observed that the Prior (Stage C) adapts extremely fast to new resolutions. So finetuning it at 2048x2048 should be computationally cheap.
|
| 31 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/634cb5eefb80cc6bcaf63c3e/5pA5KUfGmvsObqiIjdGY1.jpeg" width=1000>
|
| 32 |
|
| 33 |
## How to run
|
| 34 |
This pipeline should be run together with a prior https://huggingface.co/warp-ai/wuerstchen-prior:
|
|
|
|
| 55 |
).images
|
| 56 |
```
|
| 57 |
|
| 58 |
+
### Image Sampling Times
|
| 59 |
+
The figure shows the inference times (on an A100) for different batch sizes (`num_images_per_prompt`) on Würstchen compared to [Stable Diffusion XL](https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0) (without refiner).
|
| 60 |
+
The left figure shows inference times (using torch > 2.0), whereas the right figure applies `torch.compile` to both pipelines in advance.
|
| 61 |
+
|
| 62 |
+
|
| 63 |
## Model Details
|
| 64 |
- **Developed by:** Pablo Pernias, Dominic Rampas
|
| 65 |
- **Model type:** Diffusion-based text-to-image generation model
|