Skip to content

Update acceleration tutorials for MetaTensor #786

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

Merged
merged 7 commits into from
Jul 16, 2022

Conversation

yuchen-xu
Copy link
Contributor

Fixes #782 .

Status

Work in progress

Checks

  • Notebook runs automatically ./runner [-p <regex_pattern>]

@review-notebook-app
Copy link

Check out this pull request on  ReviewNB

See visual diffs & provide feedback on Jupyter Notebooks.


Powered by ReviewNB

@yuchen-xu yuchen-xu force-pushed the 782-acceleration-metatensor branch from 8956d19 to 5983bbb Compare July 16, 2022 01:44
@Nic-Ma
Copy link
Contributor

Nic-Ma commented Jul 16, 2022

Hi @yuchen-xu ,

Did you see any performance drop in the fast training pipeline when you switched from 0.9.0 to 0.9.1?

Thanks.

@yuchen-xu
Copy link
Contributor Author

@Nic-Ma Thanks a lot for the suggestion. I tried it and here's what I got:
0.9.1rc3 Original: 314.19s Optimized: 21.78s
0.9.0 Original: 278.79s Optimized: 13.00s

@yuchen-xu yuchen-xu marked this pull request as ready for review July 16, 2022 02:54
@yuchen-xu yuchen-xu requested review from Nic-Ma and wyli July 16, 2022 02:54
@Nic-Ma
Copy link
Contributor

Nic-Ma commented Jul 16, 2022

Hi @wyli ,

Seems 0.9.1rc3 is much slower, especially for the optimized pipeline?

Thanks.

@wyli
Copy link
Contributor

wyli commented Jul 16, 2022

yes I've addressed some deepcopy issues in Project-MONAI/MONAI#4709, will tag a new release candidate for testing. this PR looks good.

@wyli wyli changed the title [WIP] Update acceleration tutorials for MetaTensor Update acceleration tutorials for MetaTensor Jul 16, 2022
@wyli wyli merged commit 6863062 into Project-MONAI:main Jul 16, 2022
@yuchen-xu yuchen-xu mentioned this pull request Jul 18, 2022
@Nic-Ma
Copy link
Contributor

Nic-Ma commented Jul 19, 2022

I tested the training speed on V100 with MONAI 0.9.0 and latest dev (after 0.9.1rc3):
average speed of 1 epoch: 0.9.0=1.74s, dev=1.8s
Maybe we can continue to optimize MetaTensor operations in the future versions.

Thanks.

boneseva pushed a commit to boneseva/MONAI-tutorials that referenced this pull request Apr 21, 2024
* updated AMP tutorial for 0.9.1rc2

* updated for 0.9.1rc2 compatibility

* modification done, pending final checks

* tests run on 0.9.1rc3, ready for review

* updated for 0.9.1rc3

Co-authored-by: Yuchen Xu <[email protected]>
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

Successfully merging this pull request may close these issues.

Update acceleration tutorials for MetaTensor
3 participants