|
|
--- |
|
|
license: mit |
|
|
tags: |
|
|
- deepdanbooru |
|
|
- safetensors |
|
|
- pytorch |
|
|
pipeline_tag: image-to-text |
|
|
--- |
|
|
|
|
|
# Deepdanbooru-PyTorch Model Card |
|
|
|
|
|
Deepdanbooru-pytorch is a modular, weight-compatible rewrite of [TorchDeepDanbooru](https://github.com/AUTOMATIC1111/TorchDeepDanbooru), with all original weights migrated into a `model.safetensors` file for efficient loading. |
|
|
|
|
|
## Model Details |
|
|
|
|
|
The complete implementation is available in the [Pearisli/VisGen repository](https://github.com/Pearisli/VisGen/blob/main/models/deepdanbooru.py). You don’t need to clone the entire repo—simply copy the contents of `deepdanbooru.py` and you’re all set. |
|
|
|
|
|
## Examples |
|
|
|
|
|
```python |
|
|
# import or define class DeepDanbooruModel |
|
|
from PIL import Image |
|
|
|
|
|
model = DeepDanbooruModel.from_pretrained("pearisli/deepdanbooru-pytorch") |
|
|
model = model.to("cuda") |
|
|
|
|
|
image = Image.open("example.jpg").convert("RGB") |
|
|
model.tag(image) |
|
|
``` |
