add truncation in text translator #472
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
Signed-off-by: David Dale <[email protected]>
facebook-github-bot
added
the
CLA Signed
Signed-off-by: David Dale <[email protected]>
cbalioglu
reviewed
Apr 26, 2024
src/fairseq2/generation/text.py
Outdated
| @@ -200,6 +205,9 @@ def __call__(self, source_text: str) -> Tuple[str, Seq2SeqGeneratorOutput]: | |||
| """ | |||
| source_seq = self._source_text_encoder(source_text) | |||
|
|
|||
| if self.max_src_len and source_seq.shape[0] > self.max_src_len: | |||
There was a problem hiding this comment.
Although it would be a very edge case; if max_src_len is 0, then the first condition if self.max_src_len will be false, but maybe the user deliberately wants to trim to zero length (although does not make much sense, still expected since our parameter contract says that this parameter is ignored only if it is None). The "right" check would be if self.max_src_len is not None and ...
There was a problem hiding this comment.
If the source is trimmed to 0 elements, the translation is going to fail, because the encoder would consume an empty tensor. Therefore, I don't want to apply truncation to zero, and instead will change the parameter contract to explicitly truncate only to a positive number of tokens.
Signed-off-by: David Dale <[email protected]>
cbalioglu
approved these changes
May 3, 2024
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Outdated suggestions cannot be applied.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.
What does this PR do? Please describe:
This PR allows bypassing the problem of
TextTranslatorfailure when the number of source tokens exceeds the maximal length supported by the translation model. As tokenization happens within theTextTranslator, truncation, if implemented, should also happen there. The PR therefore adds amax_src_lenargument to it, and enables truncation if this argument is not empty.Does your PR introduce any breaking changes? If yes, please list them:
None
Check list: