fix eval hook save wrong best checkpoint bug #5341
Closed
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Outdated suggestions cannot be applied.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.
Motivation
When we enable save best in eval hook, it will link last ckpt as best ckpt. Related code as follow:
/mmdetection/mmdet/core/evaluation/eval_hooks.py
But, when we add checkpoint hook to runner, we use priority 70. We add eval hook use priority 50. Related code:
/mmcv/runner/base_runner.py
/mmdetection/mmdet/api/train.py
So, eval hook with a higher pirority than checkpoint hook, and it will be called before checkpoint hook. Therefore, it will save the last epoch ckpt instead of cur epoch ckpt when save best ckpt. Example, cur epoch is 9 epoch, and eval hook will test model with 9 epoch's weight but link 8 epoch's ckpt as best ckpt.
Modification
change
/mmdetection/mmdet/api/train.py
to
/mmdetection/mmdet/api/train.py