whisper : improve beam search candidate diversity #1947
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
As of ggerganov#1486, whisper.cpp uses a unified KV cache with KQ masking. As a result, depending on their location in the batch, identical sequences in a batch can have slightly different outputs due to floating point rounding errors during reduction. See the discussion in ggerganov#1941 for more details. The beam search code used "has identical sum of log probabilities" as a shorthand for "is an identical token sequence". However, per above, identical tokens do not necessarily result in identical probabilities. Instead, explicitly compare on sequences. This is linear in cost when they are identical, but the lengths are always small and the comparisons are cheap. This increases diversity during beam search. This improves output quality for some short samples I've been working with, at no detectable performance cost. I haven't checked against larger corpuses. Fixes ggerganov#1941
ggerganov
approved these changes
Mar 10, 2024
jiahansu
pushed a commit
to WiseSync/whisper.cpp
that referenced
this pull request
Apr 17, 2024
As of ggerganov#1486, whisper.cpp uses a unified KV cache with KQ masking. As a result, depending on their location in the batch, identical sequences in a batch can have slightly different outputs due to floating point rounding errors during reduction. See the discussion in ggerganov#1941 for more details. The beam search code used "has identical sum of log probabilities" as a shorthand for "is an identical token sequence". However, per above, identical tokens do not necessarily result in identical probabilities. Instead, explicitly compare on sequences. This is linear in cost when they are identical, but the lengths are always small and the comparisons are cheap. This increases diversity during beam search. This improves output quality for some short samples I've been working with, at no detectable performance cost. I haven't checked against larger corpuses. Fixes ggerganov#1941
viktor-silakov
pushed a commit
to viktor-silakov/whisper_node_mic.cpp
that referenced
this pull request
May 11, 2024
As of ggerganov#1486, whisper.cpp uses a unified KV cache with KQ masking. As a result, depending on their location in the batch, identical sequences in a batch can have slightly different outputs due to floating point rounding errors during reduction. See the discussion in ggerganov#1941 for more details. The beam search code used "has identical sum of log probabilities" as a shorthand for "is an identical token sequence". However, per above, identical tokens do not necessarily result in identical probabilities. Instead, explicitly compare on sequences. This is linear in cost when they are identical, but the lengths are always small and the comparisons are cheap. This increases diversity during beam search. This improves output quality for some short samples I've been working with, at no detectable performance cost. I haven't checked against larger corpuses. Fixes ggerganov#1941
iThalay
pushed a commit
to iThalay/whisper.cpp
that referenced
this pull request
Sep 23, 2024
As of ggerganov#1486, whisper.cpp uses a unified KV cache with KQ masking. As a result, depending on their location in the batch, identical sequences in a batch can have slightly different outputs due to floating point rounding errors during reduction. See the discussion in ggerganov#1941 for more details. The beam search code used "has identical sum of log probabilities" as a shorthand for "is an identical token sequence". However, per above, identical tokens do not necessarily result in identical probabilities. Instead, explicitly compare on sequences. This is linear in cost when they are identical, but the lengths are always small and the comparisons are cheap. This increases diversity during beam search. This improves output quality for some short samples I've been working with, at no detectable performance cost. I haven't checked against larger corpuses. Fixes ggerganov#1941
iThalay
pushed a commit
to iThalay/whisper.cpp
that referenced
this pull request
Sep 23, 2024
As of ggerganov#1486, whisper.cpp uses a unified KV cache with KQ masking. As a result, depending on their location in the batch, identical sequences in a batch can have slightly different outputs due to floating point rounding errors during reduction. See the discussion in ggerganov#1941 for more details. The beam search code used "has identical sum of log probabilities" as a shorthand for "is an identical token sequence". However, per above, identical tokens do not necessarily result in identical probabilities. Instead, explicitly compare on sequences. This is linear in cost when they are identical, but the lengths are always small and the comparisons are cheap. This increases diversity during beam search. This improves output quality for some short samples I've been working with, at no detectable performance cost. I haven't checked against larger corpuses. Fixes ggerganov#1941
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.
As of #1486, whisper.cpp uses a unified KV cache with KQ masking.
As a result, depending on their location in the batch,
identical sequences in a batch can have slightly different outputs
due to floating point rounding errors during reduction.
See the discussion in #1941 for more details.
The beam search code used "has identical sum of log probabilities"
as a shorthand for "is an identical token sequence". However, per above,
identical tokens do not necessarily result in identical probabilities.
Instead, explicitly compare on sequences.
This is linear in cost when they are identical,
but the lengths are always small and the comparisons are cheap.
This increases diversity during beam search.
This improves output quality for some short samples I've been working
with, at no detectable performance cost.
I haven't checked against larger corpuses.
Fixes #1941