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.

Sign up for GitHub

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

Jump to bottom

Split learning with CIFAR-10 #1168

Merged
merged 1 commit into from
Feb 16, 2023
Merged

Split learning with CIFAR-10 #1168

merged 1 commit into from
Feb 16, 2023

Conversation

holgerroth
Copy link
Collaborator

@holgerroth holgerroth commented Nov 29, 2022 •
edited
Loading

  • Adds split learning example with CIFAR-10 (1 client has images, and 1 client has labels)
  • Shows end-to-end workflow, including private set intersection (PSI) workflow in the notebook.

TODOs

  • Currently not runnable, Requires FCI

@holgerroth holgerroth force-pushed the cifar10_splitnn branch 2 times, most recently from e6b2dd6 to 15cbb6b Compare January 20, 2023 20:05
@holgerroth holgerroth marked this pull request as ready for review February 1, 2023 17:12
@holgerroth
Copy link
Collaborator Author

/build

1 similar comment
@holgerroth
Copy link
Collaborator Author

/build

@holgerroth holgerroth force-pushed the cifar10_splitnn branch 2 times, most recently from b27012e to 5d3bc5b Compare February 2, 2023 00:12
@holgerroth
Copy link
Collaborator Author

/build

@holgerroth holgerroth enabled auto-merge (squash) February 2, 2023 00:13
@holgerroth holgerroth disabled auto-merge February 2, 2023 00:13
@holgerroth
Copy link
Collaborator Author

/build

@holgerroth holgerroth force-pushed the cifar10_splitnn branch 2 times, most recently from 51c84ca to e43eeef Compare February 2, 2023 15:53
@holgerroth holgerroth force-pushed the cifar10_splitnn branch 2 times, most recently from 0a5c781 to 6b91ea8 Compare February 6, 2023 16:22
@holgerroth
Copy link
Collaborator Author

/build

@holgerroth
Copy link
Collaborator Author

/build

@holgerroth
Copy link
Collaborator Author

/build

@holgerroth
Copy link
Collaborator Author

/build

@holgerroth
Copy link
Collaborator Author

/build

@holgerroth
Copy link
Collaborator Author

/build

@holgerroth
Copy link
Collaborator Author

/build

1 similar comment
@holgerroth
Copy link
Collaborator Author

/build

ZiyueXu77
ZiyueXu77 previously approved these changes Feb 16, 2023
View reviewed changes
Copy link
Collaborator

@ZiyueXu77 ZiyueXu77 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Logic looks good, we can merge it now and double check the performance after the dependency functionality added

@holgerroth
Cifar10 split learning
f8192c9 
run training with higher timeout value

rename classes

add training accuracy

remove printouts

add figure

update printouts

update readme

vertical split data

run psi

add notebook

update notebook

update notebook

take intersection.txt as input for split-learning

configure overlap

add todo

refactor to use FCI

update requirements

formatting

add validation

unify gitignore

revert network; remove unnecessary check; use stats pool for computation time

introduce cifar10 data utils

move splitnn example to vertical_federated_learning

move more files

deleted moved files

move to tutorials

address comments

move CIFAR10SplitNN to splitnn src folder

fix some import errors

use system panic

use psi from apt_opt folder
@holgerroth holgerroth requested a review from ZiyueXu77 February 16, 2023 20:18
@holgerroth
Copy link
Collaborator Author

/build

@holgerroth holgerroth merged commit 3111a23 into NVIDIA:dev Feb 16, 2023
chesterxgchen
chesterxgchen reviewed Feb 16, 2023
View reviewed changes
examples/tutorial/vertical_federated_learning/cifar10-splitnn/cifar10_split_data_vertical.py

splitter = Cifar10VerticalDataSplitter(split_dir=args.split_dir, overlap=args.overlap)

# set up a dummy context for logging
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

is possible to hide FLARE specific concept in the executor, I felt we should hide as much as we can the FLARE constructs. and expose as little as possible, so that user felt easy to transition from single machine learning to FL learning. In this case particularly, can we not introduce FLContext(), at this point, As we are just doing data splitting.

Copy link
Collaborator Author

@holgerroth holgerroth Feb 16, 2023 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This way, it can also be used as an FL component similar to the cifar10 FedAvg example for simulation purposes.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For example, you can automate the runs by adding the Cifar10VerticalDataSplitter to your job config and run experiments with different overlaps in parallel.

chesterxgchen
chesterxgchen reviewed Feb 16, 2023
View reviewed changes
Copy link
Collaborator

@chesterxgchen chesterxgchen left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I just started looking at this and look its already merged,

@holgerroth holgerroth mentioned this pull request Feb 16, 2023
Closed
@holgerroth
Copy link
Collaborator Author

I just started looking at this and look its already merged,

Thanks for the comments, @chesterxgchen. I created a new issue related to the vertical data splitter you mentioned #1379.

@holgerroth holgerroth deleted the cifar10_splitnn branch February 16, 2023 21:19
guopengf pushed a commit to holgerroth/NVFlare that referenced this pull request Mar 9, 2023
@holgerroth @guopengf
Cifar10 split learning (NVIDIA#1168)
e3968a7 
run training with higher timeout value

rename classes

add training accuracy

remove printouts

add figure

update printouts

update readme

vertical split data

run psi

add notebook

update notebook

update notebook

take intersection.txt as input for split-learning

configure overlap

add todo

refactor to use FCI

update requirements

formatting

add validation

unify gitignore

revert network; remove unnecessary check; use stats pool for computation time

introduce cifar10 data utils

move splitnn example to vertical_federated_learning

move more files

deleted moved files

move to tutorials

address comments
holgerroth added a commit to holgerroth/NVFlare that referenced this pull request May 15, 2023
@holgerroth
Cifar10 split learning (NVIDIA#1168)
71746c5 
run training with higher timeout value

rename classes

add training accuracy

remove printouts

add figure

update printouts

update readme

vertical split data

run psi

add notebook

update notebook

update notebook

take intersection.txt as input for split-learning

configure overlap

add todo

refactor to use FCI

update requirements

formatting

add validation

unify gitignore

revert network; remove unnecessary check; use stats pool for computation time

introduce cifar10 data utils

move splitnn example to vertical_federated_learning

move more files

deleted moved files

move to tutorials

address comments
holgerroth added a commit to holgerroth/NVFlare that referenced this pull request Dec 4, 2023
@holgerroth
Cifar10 split learning (NVIDIA#1168)
468e0b6 
run training with higher timeout value

rename classes

add training accuracy

remove printouts

add figure

update printouts

update readme

vertical split data

run psi

add notebook

update notebook

update notebook

take intersection.txt as input for split-learning

configure overlap

add todo

refactor to use FCI

update requirements

formatting

add validation

unify gitignore

revert network; remove unnecessary check; use stats pool for computation time

introduce cifar10 data utils

move splitnn example to vertical_federated_learning

move more files

deleted moved files

move to tutorials

address comments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@YuanTingHsieh YuanTingHsieh YuanTingHsieh left review comments

@chesterxgchen chesterxgchen chesterxgchen left review comments

@ZiyueXu77 ZiyueXu77 ZiyueXu77 approved these changes

@yanchengnv yanchengnv Awaiting requested review from yanchengnv

@yhwen yhwen Awaiting requested review from yhwen

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@holgerroth @YuanTingHsieh @chesterxgchen @ZiyueXu77