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Outdated suggestions cannot be applied.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.
Fix the OOM issue mentioned in #23
General Idea
Each time we find a
SpeechEnd
transition, remove the current active speech (and any silence before that) from internal buffer as they won't be used again for any future inferences.Use a
deleted_samples
counter to make adjustment to any index calculations on internal buffer.Also fixed a logic bug in main: When we find a
SpeechEnd
, we should resetself.speech_start_ms
toNone
. This modification is missing in main.Why the modified
*.json
files are correctAll modified part is
current_speech_samples
section. After we find aSpeechEnd
but not yet find anotherSpeechStart
, current speech samples should be 0 to reflect this fact.