README.md

---
library_name: diffusers
license: other
license_name: flux-1-dev-non-commercial-license
language:
- ko
base_model:
- black-forest-labs/FLUX.1-dev
---

# Model Card for Model ID

한국어 적용 가능성 **테스트용** (원래 모델에 비해서 이미지 생성 성능이 낮음)


## Model Details


### 한국어 적용 방법

- black-forest-labs/FLUX.1-dev를 기반으로 사용
- Bingsu/clip-vit-large-patch14-ko를 적용
- T5 Encoder에 한국어 vocab 확장을 적용 + fine-tuning (코랩 쿼타 이슈로 인해서, 1 X A100을 2시간 학습)

### example
```
import torch
from diffusers import FluxPipeline
pipe = FluxPipeline.from_pretrained("jaeyong2/FLUX.1-dev-ko", torch_dtype=torch.bfloat16)

pipe.enable_model_cpu_offload() #save some VRAM by offloading the model to CPU. Remove this if you have enough GPU power

prompt = "하늘 위를 달리는 고양이'"
image = pipe(
    prompt,
    height=1024,
    width=1024,
    guidance_scale=3.5,
    num_inference_steps=50,
    max_sequence_length=512,
    generator=torch.Generator("cpu").manual_seed(0)
).images[0]
```


<div style="max-width: 350px; margin: 0 auto;">
<img src='https://huggingface.co/jaeyong2/FLUX.1-dev-ko/resolve/main/flux-dev3.png' />
</div>