Hongkai Zheng

Can you first check your data path? "lmdb.Error: lmdb: No such file or directory", lmdb didn't find your file.

> Hi, I'm also trying to train DDPM from scratch to reproduce the paper's results. Any update?

你好！上面的代码我看了，现在这个代码各种东西杂糅在一起，很难debug。有没有可能把不必要的东西删掉，找到能够复现错误的最小的代码片段呢？

可以啊 微信直接发这里不好，发我邮箱[email protected] 吧

I will push a new code in like one week. You can try that one. However, for the current version, I would suggest add use nn.BCEwithLogits and L2 weight penalty...

Hey, 1. If you have your own dataset, 86-91 is not needed. 'sub', 'pde_sub' is the subsampling ratio. If you work with your data, no need to worry about that....

Thanks for your response! I agree that they are not guaranteed to produce the same results. But I expect they are approximately the same up to a small error. What...

BTW, I tested it on A100 but didn't enable TF32.