Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

question about resnet18 PIM accuracy #11

Open
cosineAlh opened this issue Jan 8, 2025 · 7 comments
Open

question about resnet18 PIM accuracy #11

cosineAlh opened this issue Jan 8, 2025 · 7 comments

Comments

@cosineAlh
Copy link

cosineAlh commented Jan 8, 2025

Hi! This is a very clear and useful tool for relative researchers. However I have some questions when i use this tool.

I used the default given .pth model to test resnet18, the original accuracy is about 86%, however the pim-based result is about 26% without any non-idealties. Can anyone give me some tips?

@shisirun
Copy link

shisirun commented Jan 9, 2025

I have the same problem,Do you have any ideas?

@cosineAlh
Copy link
Author

I have the same problem,Do you have any ideas?

Hi, I am still confusing about this, can we exchange contact for better discussion? You can find my email on my github homepage.

@cosineAlh
Copy link
Author

I have the same problem,Do you have any ideas?

Hi, I found that the version of MNSIM_Python do not have this problem, you can use that version.

@cosineAlh
Copy link
Author

cosineAlh commented Jan 10, 2025

I have the same problem,Do you have any ideas?

Well, I find out what is wrong! You should modify this line in quantize.py like this!
Hope the authors could correct it. @Zhu-Zhenhua @ILTShade
image

@Zhu-Zhenhua
Copy link
Contributor

Hi all, many thanks for figuring out this problem, we'll fix the bugs as soon as possible!!

@shisirun
Copy link

我有同样的问题,你有什么想法吗?

好吧,我找出了问题所在!你应该像这样修改这行 quantize.py!希望作者能纠正它。 图像

Thank you very much for your reply, I changed it according to your reply, but the Original accuracy became 0.4616477272727273, do you know why

@cosineAlh
Copy link
Author

我有同样的问题,你有什么想法吗?

好吧,我找出了问题所在!你应该像这样修改这行 quantize.py!希望作者能纠正它。 图像

Thank you very much for your reply, I changed it according to your reply, but the Original accuracy became 0.4616477272727273, do you know why

I am not sure if you use the most recent updated .pth file, you can try to re-download the latest one from the link provided by the authors.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants