README.md

---
license: cc-by-nc-4.0
---

# InstaFlow-0.9B fine-tuned from 2-Rectified Flow

InstaFlow-0.9B is a **one-step** text-to-image generative model fine-tuned from [2-Rectified Flow](https://huggingface.co/XCLiu/2_rectified_flow_from_sd_1_5).

It is trained with text-conditioned reflow and distillation as described in [our paper](https://arxiv.org/abs/2309.06380). 

Rectified Flow has interesting theoretical properties. You may check [this ICLR paper](https://arxiv.org/abs/2209.03003) and [this arXiv paper](https://arxiv.org/abs/2209.14577).

## 512-Resolution Images Generated from InstaFlow-0.9B

![image/png](https://cdn-uploads.huggingface.co/production/uploads/646b0bbdec9a61e871799339/Q3RF8B3tMVL2Zx7QDwYjL.png)

# Usage

Please refer to the [official github repo](https://github.com/gnobitab/InstaFlow).

## Training

Training pipeline:
1. Distill (Stage 1):
   Starting from the [2-Rectified Flow](https://huggingface.co/XCLiu/2_rectified_flow_from_sd_1_5) checkpoint, we fix the time t=0 for the neural network,
   and fine-tune it using the distillation objective with a batch size of 1024 for 21,500 iterations.
   The guidance scale of the teacher model, 2-Rectified Flow, is set to 1.5 and the similarity loss is L2 loss. (54.4 A100 GPU days) 
2. Distill (Stage 2):
   We switch the similarity loss to LPIPS loss, then we continue to train the model using the distillation objective
   and a batch size of 1024 for another 18,000 iterations. (53.6 A100 GPU days) 
   
The final model is **InstaFlow-0.9B**.

**Total Training Cost:**  It takes 199.2 A100 GPU days in total (data generation + reflow + distillation) to get InstaFlow-0.9B.

## Evaluation Results - Metrics

The following metrics of InstaFlow-0.9B are measured on MS COCO 2017 with 5,000 images and 1-step Euler solver:

*FID-5k = 23.4, CLIP score = 0.304*

Measured on MS COCO 2014 with 30,000 images and 1-step Euler solver:

*FID-30k = 13.1*

## Citation
```
@article{liu2023insta,
  title={InstaFlow: One Step is Enough for High-Quality Diffusion-Based Text-to-Image Generation},
  author={Liu, Xingchao and Zhang, Xiwen and Ma, Jianzhu and Peng, Jian and Liu, Qiang},
  journal={arXiv preprint arXiv:2309.06380},
  year={2023}
}
```