---
license: openrail++
base_model: stabilityai/stable-diffusion-xl-base-1.0
tags:
- stable-diffusion-xl
- stable-diffusion-xl-diffusers
- text-to-image
- diffusers
- instruct-pix2pix
inference: false
datasets:
- timbrooks/instructpix2pix-clip-filtered
---

# SDXL InstructPix2Pix (768768)

Instruction fine-tuning of [Stable Diffusion XL (SDXL)](https://hf.co/papers/2307.01952) à la [InstructPix2Pix](https://huggingface.co/papers/2211.09800). Some results below:


**Edit instruction**: *"Turn sky into a cloudy one"*

![](https://huggingface.co/datasets/sayakpaul/sample-datasets/resolve/main/sdxl-instructpix2pix-release/0_0_mountain_gs%403.0_igs%401.5_steps%4050.png)

**Edit instruction**: *"Make it a picasso painting"*

![](https://huggingface.co/datasets/sayakpaul/sample-datasets/resolve/main/sdxl-instructpix2pix-release/1_1_cyborg_gs%403.0_igs%401.5_steps%4050.png)

**Edit instruction**: *"make the person older"*

![](https://huggingface.co/datasets/sayakpaul/sample-datasets/resolve/main/sdxl-instructpix2pix-release/image_three_2.png)

## Usage in 🧨 diffusers

Make sure to install the libraries first:

```bash 
pip install accelerate transformers
pip install git+https://github.com/huggingface/diffusers
```

```python 
import torch
from diffusers import StableDiffusionXLInstructPix2PixPipeline
from diffusers.utils import load_image

resolution = 768
image = load_image(
    "https://hf.co/datasets/diffusers/diffusers-images-docs/resolve/main/mountain.png"
).resize((resolution, resolution))
edit_instruction = "Turn sky into a cloudy one"

pipe = StableDiffusionXLInstructPix2PixPipeline.from_pretrained(
    "diffusers/sdxl-instructpix2pix-768", torch_dtype=torch.float16
).to("cuda")

edited_image = pipe(
    prompt=edit_instruction,
    image=image,
    height=resolution,
    width=resolution,
    guidance_scale=3.0,
    image_guidance_scale=1.5,
    num_inference_steps=30,
).images[0]
edited_image.save("edited_image.png")
```

To know more, refer to the [documentation](https://huggingface.co/docs/diffusers/main/en/api/pipelines/pix2pix). 

🚨 Note that this checkpoint is experimental in nature and there's a lot of room for improvements. Please use the "Discussions" tab of this repository to open issues and discuss. 🚨

## Training
We fine-tuned SDXL using the InstructPix2Pix training methodology for 15000 steps using a fixed learning rate of 5e-6 on an image resolution of 768x768. 

Our training scripts and other utilities can be found [here](https://github.com/sayakpaul/instructpix2pix-sdxl/tree/b9acc91d6ddf1f2aa2f9012b68216deb40e178f3) and they were built on top of our [official training script](https://huggingface.co/docs/diffusers/main/en/training/instructpix2pix). 

Our training logs are available on Weights and Biases [here](https://wandb.ai/sayakpaul/instruct-pix2pix-sdxl-new/runs/sw53gxmc). Refer to this link for details on all the hyperparameters. 

### Training data
We used this dataset: [timbrooks/instructpix2pix-clip-filtered](https://huggingface.co/datasets/timbrooks/instructpix2pix-clip-filtered).

### Compute
one 8xA100 machine

### Batch size
Data parallel with a single gpu batch size of 8 for a total batch size of 32.

### Mixed precision
FP16