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

[Fix] Fix the resuming error caused by HistoryBuffer #1078

Merged
merged 6 commits into from
Apr 21, 2023

Conversation

HAOCHENYE
Copy link
Collaborator

@HAOCHENYE HAOCHENYE commented Apr 16, 2023

Thanks for your contribution and we appreciate it a lot. The following instructions would make your pull request more healthy and more easily get feedback. If you do not understand some items, don't worry, just make the pull request and seek help from maintainers.

Motivation

After PR #972, HistoryBuffer.statistics will be called with the custom_keys_copy. However, the class attributed HistoryBuffer._statistics_methods will not be dumped during saving the checkpoint. It means the resumed HistoryBuffers can not call the statistic methods since they have an empty _statistic_methods

This error can be trigger follow these steps:

  1. set log_with_hierarchy=True for the log_processor and start training
  2. Stop the training until the next checkpoint has been saved after the prior validation. At this time, All HistoryBuffers including val/time and val/data_time will be dumped without _statistics_methods.
  3. Resume the training. The HistoryBuffers will be loaded without _statistics_methods.
  4. During validation, the HistoryBuffer of val/data_time and val/time will call statistics method without _statistics_methods, and an error will be raised.

Modification

  1. Make HistoryBuffer Dump the _statistics_methods.
  2. Refine the logic of log_processor

BC-breaking (Optional)

Does the modification introduce changes that break the backward-compatibility of the downstream repos?
If so, please describe how it breaks the compatibility and how the downstream projects should modify their code to keep compatibility with this PR.

Use cases (Optional)

If this PR introduces a new feature, it is better to list some use cases here, and update the documentation.

Checklist

  1. Pre-commit or other linting tools are used to fix the potential lint issues.
  2. The modification is covered by complete unit tests. If not, please add more unit test to ensure the correctness.
  3. If the modification has potential influence on downstream projects, this PR should be tested with downstream projects, like MMDet or MMCls.
  4. The documentation has been modified accordingly, like docstring or example tutorials.

@HAOCHENYE HAOCHENYE added P0 Highest priority Bug:P0 and removed P0 Highest priority labels Apr 19, 2023
@zhouzaida zhouzaida linked an issue Apr 20, 2023 that may be closed by this pull request
2 tasks
Co-authored-by: Zaida Zhou <58739961+zhouzaida@users.noreply.github.com>
@HAOCHENYE HAOCHENYE changed the title [Fix] Fix the resuming error caused by not dumping the class attribut… [Fix] Fix the resuming error caused by HistoryBuffer Apr 20, 2023
@zhouzaida zhouzaida merged commit 17c5414 into open-mmlab:main Apr 21, 2023
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

[Bug] the first val loop report KeyError after resuming training
2 participants