Fix char error rate calculation #14
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
When decoding outputs from a model to preview or compute the char error rate, the part of the output that corresponds to the padding region of the input needs to be ignored. The loss function only takes into consideration the un-padded part of the input. This fixes an issue where predictions had spurious extra characters on the end and the calculated char error rate was higher than it should have been.
Since the model output downsamples the input image width by 4x, the size increment for target vectors can be smaller than the size increment for inputs. Also make the image width step a power of 2, since such sizes are generally more optimal in various layers of the ML runtime.
- Add `lr` flag to set initial LR, making it easier to experiment with different values. - Remove use of deprecated `verbose` kwarg for `ReduceLROnPlateau` and instead of `get_last_lr` to log the learning rate.
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Outdated suggestions cannot be applied.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.
This fixes an issue with calculating the character error rate and generating model previews during recognition model training, due to mis-handling of padding in the inputs. The reported rate was higher than it should have been. See commits for details.
Also fix a warning about use of deprecated
verboseflag for PyTorch schedulers.