search

andreas128 / RePaint

Official PyTorch Code and Models of "RePaint: Inpainting using Denoising Diffusion Probabilistic Models", CVPR 2022
1.99k stars 160 forks source link

Great work, When will you relase the train code? #8

Closed zhangbaijin closed 2 years ago

ExponentialML commented 2 years ago

Get the training parameters here: https://github.com/openai/improved-diffusion Use the repository here to train: https://github.com/openai/guided-diffusion

You don't need to train a classifier, but you can if you wish. The image_train.py exists in both repositories, but I believe guided-diffusion has better optimizations for training than the improved-diffusion one, so use that repository instead.

andreas128 commented 2 years ago

@ExponentialML thanks for answering.

pfeducode commented 1 year ago

Do you mean to replace the training code of this project with the training file in guided-diffusion directly？

Hydrageno commented 1 year ago

Do you mean to replace the training code of this project with the training file in guided-diffusion directly？

I have the same question, do u solve it?

pfeducode commented 1 year ago

你的意思是直接用guided-diffusion中的训练文件替换本项目的训练代码？

我有同样的问题，你解决了吗？

I have only tested the code for guided diffusion. The author of this repository has provided the parameter configuration for diffusion in document face_example.yml, which I believe can be replicated. However, you need to organize the code according to the methods in the RePaint paper, such as adding masks to images.

Hydrageno commented 1 year ago

你的意思是直接用guided-diffusion中的训练文件替换本项目的训练代码？

我有同样的问题，你解决了吗？

I have only tested the code for guided diffusion. The author of this repository has provided the parameter configuration for diffusion in document face_example.yml, which I believe can be replicated. However, you need to organize the code according to the methods in the RePaint paper, such as adding masks to images.

thank u, I failed to train my diffusion model yesterday.

pfeducode commented 1 year ago

你的意思是直接用guided-diffusion中的训练文件代替本项目的训练代号？

我有同样的问题，你解决了吗？

我只测试了引导扩散的代码。本仓库作者在文档face_example.yml中提供了diffusion的参数配置，相信可以复制。但是，你需要按照RePaint论文中的方法来组织代码，比如给图像添加遮罩。

谢谢你，我昨天没有训练我的扩散模型。 you are welcome

jbnjvc10000 commented 1 year ago

你的意思是直接用guided-diffusion中的训练文件代替本项目的训练代号？

我有同样的问题，你解决了吗？

我只测试了引导扩散的代码。本仓库作者在文档face_example.yml中提供了diffusion的参数配置，相信可以复制。但是，你需要按照RePaint论文中的方法来组织代码，比如给图像添加遮罩。

谢谢你，我昨天没有训练我的扩散模型。 you are welcome

Excuse me, I actually need to train two models if i want repaint another dataset?