复现论文 Denoising Diffusion Probabilistic Model, in Pytorch。
-
论文中的【Algorithm 1 Training】和【Algorithm 2 Sampling】公式推导 | https://blog.csdn.net/u010006102/article/details/134648877
-
复现代码 | https://github.com/FMsunyh/denoising-diffusion-pytorch
- torch 1.13.0
- python 3.10
百度云盘下载: CelebA/Img/img_align_celeba.zip
cp img_align_celeba.zip ./data/celebA/
cd ./data/celebA/
unzip img_align_celeba.zip
- MNIST
python train.py --dataset mnist --epochs 6 --channels 1
- celebA
python train.py --dataset CelebA --epochs 100 --channels 3
可以查看每一轮的预测结果
./outputs
- MNIST数据集,训练6轮后的测试效果
- celebA数据集,训练50轮后的测试效果
- Denoising Diffusion Probabilistic Models (DDPM) | https://nn.labml.ai/diffusion/ddpm/index.html
- labmlai | https://github.com/labmlai/annotated_deep_learning_paper_implementations
- CelebA Dataset | https://mmlab.ie.cuhk.edu.hk/projects/CelebA.html
- U-Net model for Denoising Diffusion Probabilistic Models (DDPM) | https://nn.labml.ai/diffusion/ddpm/unet.html