Optimize _geometric_adstock for 99% function speedup, reducing total model train time by ~40% - Python version#1217
Merged
alxlyj merged 1 commit intofacebookexperimental:mainfrom Feb 25, 2025
Conversation
…~40% - Replaced loop-based implementation with `scipy.signal.lfilter` - Improved `_geometric_adstock` execution time from **108.9s → 0.495s** (~99% faster) per trial - Reduced overall app runtime from **~277.9s → ~166.7s** (~40% speedup) per trial - Added unit test to ensure numerical correctness (`np.allclose`)
facebook-github-bot
added
the
CLA Signed
benshush
changed the title
alxlyj
approved these changes
Feb 25, 2025
Contributor
alxlyj
left a comment
alxlyj
left a comment
There was a problem hiding this comment.
looks good to me - thanks!
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
3 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Project Robyn
Summary
This PR optimizes
_geometric_adstock, the main performance bottleneck, reducing its execution time by ~99% and improving the total model train time by ~40% on Python version.Fixes
This PR does not address an existing GitHub issue but significantly improves performance.
Type of change
Motivation & Context
_geometric_adstockpreviously accounted for 39% of the total model train time, significantly slowing down model execution.scipy.signal.lfilter, replacing a Python loop with a highly efficient SciPy-based implementation (built on NumPy).Performance Benchmark (tutorial1.ipynb, Trial=1, iterations=2000)
Before Optimization:
_geometric_adstock: 108.9s (39% of total model train time)After Optimization:
_geometric_adstock: 0.495s (99% faster)