Add support for masked input in TrainingSampler #546
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
Update from upstream
Update from head
Update from master
kazemnejad
changed the title
|
@seanpmorgan Could you please force run the CI tests on this PR? |
guillaumekln
previously approved these changes
Sep 30, 2019
There was a problem hiding this comment.
Looks good to me.
@qlzh727 should also validate before merging.
qlzh727
requested changes
Sep 30, 2019
It will throw an error if both sequence_length and mask are provided.
|
@qlzh727 Did you have a chance to take a look at the new commits? |
qlzh727
requested changes
Oct 3, 2019
qlzh727
reviewed
Oct 7, 2019
qlzh727
previously approved these changes
Oct 7, 2019
qlzh727
approved these changes
Oct 7, 2019
|
@qlzh727 @guillaumekln Thank you very much for your feedbacks. |
|
Thank you for these improvements. Looking forward to the seq2seq tutorial! |
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Outdated suggestions cannot be applied.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.
By applying this patch,
TrainingSamplerwill be able to support bothmaskandsequnce_lengtharguments. Butsequnce_lengthwill have higher priority. So if both are provided, themaskargument will be ignored. I had to add this priority because Keras implicitly passes the mask information to layers upon the invocation (even in model sub-classing). Without this priority, the user will face an error by using the following snippet (due to the automatically passed mask argument):To overcome this issue, the user had to manually delete the
_keras_maskattribute from thesome_masked_inputvariable.( e.g.delattr(some_masked_input, '_keras_mask')) to be able to apply thesequence_lengthparameter only.It Fixes (#534) by adding support for masked inputs. Also, this will bring better Keras integration.
Fixes #534