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 loss smooth when loss name doesn't start with loss #539

Merged
merged 2 commits into from
Oct 8, 2022

Conversation

liuyanyi
Copy link
Contributor

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

fix loss smooth when loss name doesn't start with loss, such as Cascade RCNN in mmdet and RefineSingleStageDetector in mmrotate.

For example, loss may have prefix 's0.loss_xxx'.

cascade rcnn loss before:

09/19 11:31:38 - mmengine - INFO - Epoch(train) [1][500/6400]  lr: 2.5000e-03  eta: 6:32:10  time: 0.3258  data_time: 0.0035  memory: 4946  grad_norm: 6.8527  loss_rpn_cls: 0.1451  loss_rpn_bbox: 0.0861  s0.loss_cls: 0.2497  s0.acc: 94.9219  s0.loss_bbox: 0.0781  s1.loss_cls: 0.1206  s1.acc: 95.5078  s1.loss_bbox: 0.0738  s2.loss_cls: 0.0415  s2.acc: 97.5586  s2.loss_bbox: 0.0226  loss: 1.2572

total loss sum = 0.8175, total loss in log = 1.2572

after:

09/19 11:25:44 - mmengine - INFO - Epoch(train) [1][500/6400]  lr: 2.5000e-03  eta: 6:28:44  time: 0.3342  data_time: 0.0035  memory: 4042  grad_norm: 6.7206  loss_rpn_cls: 0.1482  loss_rpn_bbox: 0.0601  s0.loss_cls: 0.3946  s0.acc: 89.4531  s0.loss_bbox: 0.1871  s1.loss_cls: 0.1690  s1.acc: 92.3828  s1.loss_bbox: 0.1161  s2.loss_cls: 0.0619  s2.acc: 96.5820  s2.loss_bbox: 0.0268  loss: 1.1638

total loss sum = 1.1638, total loss in log = 1.1638

Modification

Please briefly describe what modification is made in this PR.

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.

@zytx121
Copy link

zytx121 commented Sep 19, 2022

@HAOCHENYE Please have a look.

ZwwWayne
ZwwWayne previously approved these changes Sep 19, 2022
HAOCHENYE
HAOCHENYE previously approved these changes Sep 19, 2022
@liuyanyi liuyanyi dismissed stale reviews from HAOCHENYE and ZwwWayne via 22a013d September 21, 2022 12:37
@ZwwWayne ZwwWayne added this to the 0.2.0 milestone Sep 22, 2022
@ZwwWayne ZwwWayne merged commit ccb32c7 into open-mmlab:main Oct 8, 2022
C1rN09 pushed a commit to C1rN09/mmengine that referenced this pull request Nov 1, 2022
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.

5 participants