Way to fix bug with validation frequency #1601
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
hakuryuu96
requested review from
shaydeci,
ofrimasad,
BloodAxe and
Louis-Dupont
as code owners
Louis-Dupont
reviewed
Nov 2, 2023
|
After a discussion with @BloodAxe we decided:
Corresponding changes have been added in the last commit. |
BloodAxe
reviewed
Nov 3, 2023
BloodAxe
reviewed
Nov 3, 2023
BloodAxe
reviewed
Nov 3, 2023
BloodAxe
reviewed
Nov 3, 2023
BloodAxe
reviewed
Nov 4, 2023
BloodAxe
reviewed
Nov 4, 2023
BloodAxe
reviewed
Nov 5, 2023
Louis-Dupont
pushed a commit
that referenced
this pull request
Nov 9, 2023
* Way to fix bug with validation frequency * Fixed test, the state of net was rewritten * Added validating the latest epoch and epochs from save_ckpt_epoch_list * Added one more testcase to check wether latest notdivisible epoch has valid in metrics * Following the SRP recommendation... * Which inference time exactly * Fixed incorrect keyword in writing function * Missing brackets around epoch+1 in valid run check function. * Final fixes hopefully :) * Fixed trainer to add scalars only in main process --------- Co-authored-by: Eugene Khvedchenya <ekhvedchenya@gmail.com>
Louis-Dupont
added a commit
that referenced
this pull request
Nov 13, 2023
* first working version without test * add * add tests * move to _get_pipeline * fix * fix test * Way to fix bug with validation frequency (#1601) * Way to fix bug with validation frequency * Fixed test, the state of net was rewritten * Added validating the latest epoch and epochs from save_ckpt_epoch_list * Added one more testcase to check wether latest notdivisible epoch has valid in metrics * Following the SRP recommendation... * Which inference time exactly * Fixed incorrect keyword in writing function * Missing brackets around epoch+1 in valid run check function. * Final fixes hopefully :) * Fixed trainer to add scalars only in main process --------- Co-authored-by: Eugene Khvedchenya <ekhvedchenya@gmail.com> * add images and update autopadding responsability * add example of visualization w/o resizing * add docstring * remove unwanted prints * add explicit auto_paddign --------- Co-authored-by: hakuryuu96 <marchenkophilip@gmail.com> Co-authored-by: Eugene Khvedchenya <ekhvedchenya@gmail.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
https://github.com/Deci-AI/super-gradients/issues/1324
This PR addresses the issue above. The behavior of saving checkpoints changes when max_epochs is not divisible by val_frequency. In this case, ckpt_latest won't be checked whether it is the best. A corresponding warning was added.