|
--- |
|
license: other |
|
--- |
|
|
|
# This is a Custom Diffusion model fine-tuned from the Stable Diffusion v1-4. |
|
|
|
Custom Diffusion allows you to fine-tune text-to-image diffusion models, such as Stable Diffusion, given a few images of a new concept (~4-20). |
|
|
|
Here we give an example model fine-tuned using 5 images of a cat downloaded from UnSplash. The example code of inference is shown below. |
|
|
|
|
|
## Example code of inference |
|
```python |
|
import os |
|
import sys |
|
import torch |
|
|
|
os.system("git clone https://github.com/adobe-research/custom-diffusion") |
|
sys.path.append("custom-diffusion") |
|
from diffusers import StableDiffusionPipeline |
|
from src import diffuser_training |
|
|
|
device = 'cuda' |
|
pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16) |
|
pipe = pipe.to(self.device) |
|
|
|
weight_path = 'custom_diffusion_cat.bin' |
|
diffuser_training.load_model(pipe.text_encoder, pipe.tokenizer, pipe.unet, weight_path, '<new1>') |
|
prompt = "<new1> cat swimming in a pool" |
|
images = pipe(prompt, num_inference_steps=200, guidance_scale=6., eta=1.).images |
|
``` |
|
|