search

BPHO-Salk / PSSR

Deep Learning-Based Point-Scanning Super-Resolution (PSSR)
BSD 3-Clause "New" or "Revised" License
81 stars 29 forks source link

Something wrong with running Inference using our pretrained models #2

Closed coolwind97 closed 3 years ago

coolwind97 commented 3 years ago

Thanks for your works to supporting the data and code. Here I tried to reproduce scenario 1. I'm sure there's no problem with the environment settings, but there's a problem at runtime, as shown below. RuntimeError:“Input type (torch.cuda.FloatTensor) and weight type (torch.cuda.HalfTensor) should be the same” I guarantee that only the test images file path is changed, as shown in the readme.md

lynnf375 commented 3 years ago

Thanks for trying! We are aware of this issue now thanks to @WillGiang. I am in the process of updating the pretrained models in 3DEM.org server, which would take some time. Meanwhile, if you can send me an email to lfang@salk.edu, I'll send you the updated pretrained EM model.

Hopefully this helps!

coolwind97 commented 3 years ago

This is the best news I've heard since this holiday. I'm very glad to get the reply from the author. I've solved this problem with the tips. Thanks to @lynnf375 and @WillGiang for your efforts.

Have a nice day!

lynnf375 commented 3 years ago

This is the best news I've heard since this holiday. I'm very glad to get the reply from the author. I've solved this problem with the tips. Thanks to @lynnf375 and @WillGiang for your efforts.

Have a nice day!

That's awesome to hear 👍 Please let us know if you need help with any other issues - we are here to help!

coolwind97 commented 3 years ago

Hi,Linjing. I've reproduced the inference in scenario one, and it works great. Now I'm trying to train my own data set, and everything is going well, but I have a doubt. Can I only input 8 nm pixel data set or try to input other pixel data set? I look forward to hearing from you.

------------------ 原始邮件 ------------------ 发件人: "BPHO-Salk/PSSR" @.>; 发送时间: 2021年4月6日(星期二) 凌晨2:58 @.>; @.>;"State @.>; 主题: Re: [BPHO-Salk/PSSR] Something wrong with running Inference using our pretrained models (#2)

This is the best news I've heard since this holiday. I'm very glad to get the reply from the author. I've solved this problem with the tips. Thanks to @lynnf375 and @WillGiang for your efforts.

Have a nice day!

That's awesome to hear 👍 Please let us know if you need help with any other issues - we are here to help!

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub, or unsubscribe.

lynnf375 commented 3 years ago

Hi,Linjing. I've reproduced the inference in scenario one, and it works great. Now I'm trying to train my own data set, and everything is going well, but I have a doubt. Can I only input 8 nm pixel data set or try to input other pixel data set? I look forward to hearing from you. …

Hi! I am sorry that I missed this message after the issue was closed. The 8nm constraint is only valid if using our pre-trained EM model. If you are going to train a model with your own data, there won't be any pixel size constraint.