Bump rten v0.10.0 #79
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
The changes include significant performance improvements for the GRU operator, which makes recognition of long text lines faster, as well as adapting the thread pool size to the number of physical rather than logical cores, reducing overhead on systems with SMT / hyperthreading.
RTen model execution is multi-threaded by default. To avoid contention between threads in the Rayon global thread pool and RTen's thread pool, run parallel recognition over batches of line images in RTen's thread pool.
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Outdated suggestions cannot be applied.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.
The changes include significant performance improvements for the
GRUoperator, which makes recognition of long text lines faster, as well as adapting the thread pool size to the number of physical rather than logical cores, reducing overhead on systems with SMT / hyperthreading.To adapt to the thread pool changes,
rten::thread_pool().run(...)is now used to run line recognition preprocessing in the same thread pool that is used to execute the recognition model. This avoids contention from having more threads than cores.Benchmarks:
Testing with the
polar-bears.pngexample on an Ice Lake Intel i5:Before: ~1.00s
After: ~570ms
In contrast with the
why-rust.pngexample, which has a larger image but shorter lines, the impact is smaller:Before: ~916ms
After: ~790ms