Improved Distribution Matching Distillation for Fast Image Synthesis [4-step demo][1-step demo][ComfyUI]

Few-step Text-to-Image Generation.

Improved Distribution Matching Distillation for Fast Image Synthesis,
Tianwei Yin, Michaël Gharbi, Taesung Park, Richard Zhang, Eli Shechtman, Frédo Durand, William T. Freeman
arXiv technical report (arXiv 2405.14867)

Contact

Feel free to contact us if you have any questions about the paper!

Tianwei Yin tianweiy@mit.edu

Abstract

Recent approaches have shown promises distilling diffusion models into efficient one-step generators. Among them, Distribution Matching Distillation (DMD) produces one-step generators that match their teacher in distribution, without enforcing a one-to-one correspondence with the sampling trajectories of their teachers. However, to ensure stable training, DMD requires an additional regression loss computed using a large set of noise-image pairs generated by the teacher with many steps of a deterministic sampler. This is costly for large-scale text-to-image synthesis and limits the student's quality, tying it too closely to the teacher's original sampling paths. We introduce DMD2, a set of techniques that lift this limitation and improve DMD training. First, we eliminate the regression loss and the need for expensive dataset construction. We show that the resulting instability is due to the fake critic not estimating the distribution of generated samples accurately and propose a two time-scale update rule as a remedy. Second, we integrate a GAN loss into the distillation procedure, discriminating between generated samples and real images. This lets us train the student model on real data, mitigating the imperfect real score estimation from the teacher model, and enhancing quality. Lastly, we modify the training procedure to enable multi-step sampling. We identify and address the training-inference input mismatch problem in this setting, by simulating inference-time generator samples during training time. Taken together, our improvements set new benchmarks in one-step image generation, with FID scores of 1.28 on ImageNet-64x64 and 8.35 on zero-shot COCO 2014, surpassing the original teacher despite a 500X reduction in inference cost. Further, we show our approach can generate megapixel images by distilling SDXL, demonstrating exceptional visual quality among few-step methods.

Environment Setup

# In conda env 
conda create -n dmd2 python=3.8 -y 
conda activate dmd2 

pip install --upgrade anyio
pip install -r requirements.txt
python setup.py  develop

Inference Example

ImageNet

python -m demo.imagenet_example  --checkpoint_path IMAGENET_CKPT_PATH

Text-to-Image

# Note: on the demo page, click ``Use Tiny VAE for faster decoding'' to enable much faster speed and lower memory consumption using a Tiny VAE from [madebyollin](https://huggingface.co/madebyollin/taesdxl)

# 4 step (much higher quality than 1 step)
python -m demo.text_to_image_sdxl --checkpoint_path SDXL_CKPT_PATH --precision float16

# 1 step 
python -m demo.text_to_image_sdxl --num_step 1 --checkpoint_path SDXL_CKPT_PATH --precision float16 --conditioning_timestep 399

We can also use the standard diffuser pipeline:

4-step generation

import torch
from diffusers import DiffusionPipeline, UNet2DConditionModel, LCMScheduler
from huggingface_hub import hf_hub_download
from safetensors.torch import load_file
base_model_id = "stabilityai/stable-diffusion-xl-base-1.0"
repo_name = "tianweiy/DMD2"
ckpt_name = "dmd2_sdxl_4step_unet_fp16.bin"
# Load model.
unet = UNet2DConditionModel.from_config(base_model_id, subfolder="unet").to("cuda", torch.float16)
unet.load_state_dict(torch.load(hf_hub_download(repo_name, ckpt_name), map_location="cuda"))
pipe = DiffusionPipeline.from_pretrained(base_model_id, unet=unet, torch_dtype=torch.float16, variant="fp16").to("cuda")
pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
prompt="a photo of a cat"

# LCMScheduler's default timesteps are different from the one we used for training 
image=pipe(prompt=prompt, num_inference_steps=4, guidance_scale=0, timesteps=[999, 749, 499, 249]).images[0]

1-step generation

import torch
from diffusers import DiffusionPipeline, UNet2DConditionModel, LCMScheduler
from huggingface_hub import hf_hub_download
from safetensors.torch import load_file
base_model_id = "stabilityai/stable-diffusion-xl-base-1.0"
repo_name = "tianweiy/DMD2"
ckpt_name = "dmd2_sdxl_1step_unet_fp16.bin"
# Load model.
unet = UNet2DConditionModel.from_config(base_model_id, subfolder="unet").to("cuda", torch.float16)
unet.load_state_dict(torch.load(hf_hub_download(repo_name, ckpt_name), map_location="cuda"))
pipe = DiffusionPipeline.from_pretrained(base_model_id, unet=unet, torch_dtype=torch.float16, variant="fp16").to("cuda")
pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
prompt="a photo of a cat"
image=pipe(prompt=prompt, num_inference_steps=1, guidance_scale=0, timesteps=[399]).images[0]

Pretrained models can be found in ImageNet and SDXL.

Training and Evaluation

ImageNet-64x64

Please refer to ImageNet-64x64 for details.

SDXL

Please refer to SDXL for details.

SDv1.5

Please refer to SDv1.5 for details.

License

Improved Distribution Matching Distillation is released under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Known Issues

Current FSDP for SDXL training is really slow; help is greatly appreciated!
Current LORA training is actually slower than the full finetuning and takes the same amount of memory; help is greatly appreciated!

Citation

If you find DMD2 useful or relevant to your research, please kindly cite our papers:

@article{yin2024improved,
    title={Improved Distribution Matching Distillation for Fast Image Synthesis},
    author={Yin, Tianwei and Gharbi, Micha{\"e}l and Park, Taesung and Zhang, Richard and Shechtman, Eli and Durand, Fredo and Freeman, William T},
    journal={arXiv 2405.14867},
    year={2024}
}

@inproceedings{yin2024onestep,
    title={One-step Diffusion with Distribution Matching Distillation},
    author={Yin, Tianwei and Gharbi, Micha{\"e}l and Zhang, Richard and Shechtman, Eli and Durand, Fr{\'e}do and Freeman, William T and Park, Taesung},
    booktitle={CVPR},
    year={2024}
}

Third-part Code

EDM for dnnlib, torch_utils and edm folders.

Acknowledgments

This work was done while Tianwei Yin was a full-time student at MIT. It was developed based on our reimplementation of the original DMD paper. This work was supported by the National Science Foundation under Cooperative Agreement PHY-2019786 (The NSF AI Institute for Artificial Intelligence and Fundamental Interactions, http://iaifi.org/), by NSF Grant 2105819, by NSF CISE award 1955864, and by funding from Google, GIST, Amazon, and Quanta Computer.

lmxyy/DMD2