Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,41 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- anime
|
| 5 |
+
- diffusion
|
| 6 |
+
- text-to-image
|
| 7 |
+
- image-generation
|
| 8 |
+
library_name: diffusers
|
| 9 |
+
pipeline_tag: text-to-image
|
| 10 |
+
language:
|
| 11 |
+
- en
|
| 12 |
---
|
| 13 |
+
|
| 14 |
+
|
| 15 |
+
# Aniimage-2
|
| 16 |
+
|
| 17 |
+
Aniimage-2 is the second public model in the Aniimage series.
|
| 18 |
+
The model is a 512x512 anime image generation model trained from scratch using a UNet + VAE + CLIP architecture at around 400 million parameters.
|
| 19 |
+
Aniimage-2 has been trained on 1.2M anime images from [Danbooru](https://danbooru.donmai.us/) and other sources.
|
| 20 |
+
It is not based on any existing models, the UNet is trained from scratch.
|
| 21 |
+
|
| 22 |
+
## Model Details
|
| 23 |
+
|
| 24 |
+
| | |
|
| 25 |
+
|---|---|
|
| 26 |
+
| **Resolution** | 512×512 |
|
| 27 |
+
| **Architecture** | Latent Diffusion (UNet + VAE + CLIP) |
|
| 28 |
+
| **Parameters** | ~430M |
|
| 29 |
+
| **Training Steps** | 70,000 |
|
| 30 |
+
| **Batch Size** | 64 |
|
| 31 |
+
| **Dataset** | ~1.2 million curated anime images from Danbooru |
|
| 32 |
+
| **GPU** | NVIDIA RTX 5060 Ti 16GB |
|
| 33 |
+
| **Scheduler** | DPM ++ SDE |
|
| 34 |
+
| **Step count** | 50 using DPM ++ SDE|
|
| 35 |
+
|
| 36 |
+
## Requirements
|
| 37 |
+
|
| 38 |
+
- **GPU**: ~6 GB VRAM minimum (recommend 8+ GB)
|
| 39 |
+
- **CPU**: ~7 GB RAM. Image generation is extremely slow on CPU. (recommend 10+ GB)
|
| 40 |
+
|
| 41 |
+
|