yang-song / score_sde_pytorch

PyTorch implementation for Score-Based Generative Modeling through Stochastic Differential Equations (ICLR 2021, Oral)
https://arxiv.org/abs/2011.13456
Apache License 2.0
1.58k stars 295 forks source link

Rescaling of ouput #12

Open cutecows opened 2 years ago

cutecows commented 2 years ago

Thanks for the great repo!

I noticed that you scale your images to be between [0,1] or [-1,1]. I'd like to run the model with coordinate data, which doesn't lend itself well to scaling. I'm wondering if this is expected to significantly impact performance?

zihaozou commented 1 year ago

I think the main reason to scale the image to [-1,1] is that we start from a normal distribution