Optimize find_decision_points in neuromancer agglomeration pipeline.#124
Open
copybara-service[bot] wants to merge 1 commit into
Open
Optimize find_decision_points in neuromancer agglomeration pipeline.#124copybara-service[bot] wants to merge 1 commit into
copybara-service[bot] wants to merge 1 commit into
Conversation
We optimized the decision point identification code to resolve slowness in the pipeline. The main bottlenecks were CPU-bound Python operations on large 3D arrays in a subvolume. Bottlenecks and Fixes: - **Vectorized Relabeling**: Replaced the dict-based list comprehension in `relabel` (connectomics/segmentation/labels.py) with a vectorized implementation using `np.searchsorted`. This reduced the final relabeling time for a 12.5M voxel subvolume from ~16.7s to ~0.2s (78x speedup). - **Slicing Optimization**: Replaced `ndimage.shift` and `np.roll` with NumPy slicing views in the neighbor-checking loop (ffn/utils/decision_point.py), reducing loop time from ~2.1s to ~1.1s and avoiding memory copying. - **DataFrame Aggregation**: Collected NumPy arrays in lists and created a single DataFrame at the end of the loop, reducing pandas overhead. Overall performance for `find_decision_points` on a representative dummy subvolume improved from **26.83s to 7.85s (3.4x speedup)**. PiperOrigin-RevId: 947868587
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Optimize find_decision_points in neuromancer agglomeration pipeline.
We optimized the decision point identification code to resolve slowness in the pipeline. The main bottlenecks were CPU-bound Python operations on large 3D arrays in a subvolume.
Bottlenecks and Fixes:
relabel(connectomics/segmentation/labels.py) with a vectorized implementation usingnp.searchsorted. This reduced the final relabeling time for a 12.5M voxel subvolume from ~16.7s to ~0.2s (78x speedup).ndimage.shiftandnp.rollwith NumPy slicing views in the neighbor-checking loop (ffn/utils/decision_point.py), reducing loop time from ~2.1s to ~1.1s and avoiding memory copying.Overall performance for
find_decision_pointson a representative dummy subvolume improved from 26.83s to 7.85s (3.4x speedup).