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

Auto3DSeg continue training (skip trained algos) #6310

Merged
merged 11 commits into from
Apr 6, 2023

Conversation

myron
Copy link
Collaborator

@myron myron commented Apr 5, 2023

Second PR for issue #6291
fixes #6291

Since the previous PR #6290 was reverted #6295

Allows to skip the already trained algos, and continue training only for the non-trained ones.

after this PR, the default option AutoRunner(train=None) will have this behavior, whereas manually setting AutoRunner(train=True/False) will always train all or skip all training. Previously we can only train all or skip all (without any option to resume)

I changed import_bundle_algo_history() to return a better algo_dict

previously it returned "list[dict(name: algo)]" - a list of dict, but each dict must have a single key name "name => algo". Not it returns a list of dicts, each with several keys dict(AlgoEnsembleKeys.ID: name, AlgoEnsembleKeys.ALGO, algo, "is_trained": bool, etc).
this allows to put additional metadata inside of each algo_dict, and it's easier to read it back.

previously, to get a name we had to use "name = history[0].keys()[0]", now it's more elegant "name = history[0][AlgoEnsembleKeys.ID]".

this however required to change many files, everywhere where import_bundle_algo_history and export_bundle_algo_history was used.

All the tests have passed, except for "integration GPU utilization tests" , but those errors seems unrelated

After this PR, tutorials need to be updated too Project-MONAI/tutorials#1288

@myron myron added this to the Auto3DSeg enhancement [P0 v1.2] milestone Apr 5, 2023
@myron myron changed the title Auto3DSeg continue training (skipp trained algos) Auto3DSeg continue training (skip trained algos) Apr 5, 2023
Signed-off-by: myron <[email protected]>
Signed-off-by: myron <[email protected]>
Copy link
Contributor

@mingxin-zheng mingxin-zheng left a comment

Choose a reason for hiding this comment

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

Looks good to me. Thanks!

@wyli
Copy link
Contributor

wyli commented Apr 6, 2023

/build

@wyli
Copy link
Contributor

wyli commented Apr 6, 2023

/build

Signed-off-by: Wenqi Li <[email protected]>
@wyli
Copy link
Contributor

wyli commented Apr 6, 2023

/build

@wyli wyli enabled auto-merge (squash) April 6, 2023 12:03
Copy link
Contributor

@wyli wyli left a comment

Choose a reason for hiding this comment

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

thanks, verified that the integration tests work fine with this PR

@wyli wyli merged commit e4b313d into Project-MONAI:dev Apr 6, 2023
wyli pushed a commit to Project-MONAI/tutorials that referenced this pull request Apr 13, 2023
algo_dict key updates
this should be merged after updates in
Project-MONAI/MONAI#6310

fixes #1306

---------

Signed-off-by: myron <[email protected]>
Signed-off-by: Mingxin Zheng <[email protected]>
boneseva pushed a commit to boneseva/MONAI-tutorials that referenced this pull request Apr 21, 2024
algo_dict key updates
this should be merged after updates in
Project-MONAI/MONAI#6310

fixes Project-MONAI#1306

---------

Signed-off-by: myron <[email protected]>
Signed-off-by: Mingxin Zheng <[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.

Auto3DSeg skip trained algos
3 participants