train loss goes very large

[imdoublecats]

Thank you for your work, but i have a problem. When i trained with cifar10.yml, the loss went down at first, and the sample result looked normal. But the loss becomed very large after a few steps, and the sample result was full of noise. i also tried to train my custom data but did not meet this problem

[yinyin-llll]

hi ,i also trained the model ,but i met the next question

and the prompt is python main.py --config cifar10.yml --exp "D:\Anaconda\jupyter notebook\diffusion model\ddim-main" --doc "D:\Anaconda\jupyter notebook\diffusion model\ddim-main\logs" --ni do you know how to solve the problem?

[imdoublecats]

hi ,i also trained the model ,but i met the next question

and the prompt is python main.py --config cifar10.yml --exp "D:\Anaconda\jupyter notebook\diffusion model\ddim-main" --doc "D:\Anaconda\jupyter notebook\diffusion model\ddim-main\logs" --ni do you know how to solve the problem?

seems a path setting problem，but main.py only has 232 lines, how did you get line 300+ traceback?

[yinyin-llll]

hh,i added some annotations in main.py.but i have set the path,and i have cretaed the logs file in the catalog,like this and i used the model-790000.ckpt as my train file(put it in logs file and train) what's your operation about training?

[jiamings]

Sorry, the codebase is only used as a reference for sampling, please use some other codebase if you want to achieve high quality training.

[imdoublecats]

这是来自QQ邮箱的假期自动回复邮件。   您好，我最近正在休假中，无法亲自回复您的邮件。我将在假期结束后，尽快给您回复。