---
library_name: diffusers
pipeline_tag: text-to-image
inference: true
base_model: stabilityai/sd-turbo
---
# DPO LoRA Stable Diffusion v2-1
Model trained with LoRA implementation of Diffusion DPO Read more [here](https://github.com/huggingface/diffusers/tree/main/examples/research_projects/diffusion_dpo)


Base Model: https://huggingface.co/stabilityai/stable-diffusion-2-1

Without DPO Lora  
![image/png](https://cdn-uploads.huggingface.co/production/uploads/6064e095abd8d3692e3e2ed6/3o1UBMgrBTT5SFfgYFmaq.png)

With DPO Lora
![image/png](https://cdn-uploads.huggingface.co/production/uploads/6064e095abd8d3692e3e2ed6/chu29lik7TVGEyVddWyqb.png)

## Running

```python
from diffusers import DiffusionPipeline
from diffusers.utils import make_image_grid
import torch

pipe = DiffusionPipeline.from_pretrained(
    # "stabilityai/stable-diffusion-2-1", # or SD-2.1
    "stabilityai/sd-turbo", # or SD-TURBO

    torch_dtype=torch.float16, variant="fp16"
)
pipe.load_lora_weights("radames/dpo-lora-sd2.1", adapter_name="dpo-lora-sd21", weight_name="pytorch_lora_weights-10000.safetensors")
pipe.to("cuda");
prompt = "portrait headshot professional of elon musk"
negative_prompt = "3d render, cartoon, drawing, art, low light"
images = pipe(
    prompt=prompt,
    negative_prompt=negative_prompt,
    width=512,
    height=512,
    num_inference_steps=2,
    guidance_scale=1.0,
    num_images_per_prompt=4
).images
make_image_grid(images, 1, 4)

```