Conversation
| res.update(Template._data_collator(self, batch, padding_to=padding_to)) | ||
|
|
||
| # image_bound and audio_bounds are computed before batching. Keep them aligned with input_ids when the | ||
| # base collator left-pads shorter samples. In inference, the base collator always uses left padding; |
Collaborator
There was a problem hiding this comment.
Thanks for solving this issue! I think other minicpm templates share the same issue, this function can be used as a common helper all of them
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
PR type
PR information
MiniCPM-o image and audio bounds are computed before batch collation.
During batched inference, shorter samples are left-padded by the base
collator, but
image_boundandaudio_boundswere not shifted by thecorresponding padding length. This causes multimodal bounds to become
misaligned for shorter samples when
batch_size > 1.This PR shifts
image_boundandaudio_boundsby each sample's actualleft-padding length after the base collator determines the padded sequence
length.
The adjustment is applied only when padding-free mode is disabled and the
effective padding side is left.
Experiment results