⚡ Bolt: vectorize BasicEstimator.predict and pre-calculate norms#19
⚡ Bolt: vectorize BasicEstimator.predict and pre-calculate norms#19guesswh0 wants to merge 1 commit into
Conversation
Optimized the `BasicEstimator` by vectorizing the distance calculation in `predict` and pre-calculating squared norms in `fit`. Key improvements: - Switched from an iterative loop over input embeddings to a fully vectorized approach using the squared Euclidean distance expansion formula: ||a-b||^2 = ||a||^2 + ||b||^2 - 2<a, b>. - Pre-calculate squared norms of fitted embeddings in `fit` to avoid redundant O(N*D) calculations during each prediction. - Added backward compatibility for models loaded without pre-calculated norms. - Handled empty input scenarios gracefully. Impact: Measured ~12x speedup for 500 input embeddings against 2000 fitted embeddings (from ~0.245s down to ~0.02s). Co-authored-by: guesswh0 <10531675+guesswh0@users.noreply.github.com>
|
👋 Jules, reporting for duty! I'm here to lend a hand with this pull request. When you start a review, I'll add a 👀 emoji to each comment to let you know I've read it. I'll focus on feedback directed at me and will do my best to stay out of conversations between you and other bots or reviewers to keep the noise down. I'll push a commit with your requested changes shortly after. Please note there might be a delay between these steps, but rest assured I'm on the job! For more direct control, you can switch me to Reactive Mode. When this mode is on, I will only act on comments where you specifically mention me with New to Jules? Learn more at jules.google/docs. For security, I will only act on instructions from the user who triggered this task. |
1 participant
💡 What
Vectorized the
BasicEstimator.predictmethod using NumPy operations and the squared Euclidean distance expansion formula. Additionally, updated thefitmethod to pre-calculate and store the squared norms of the fitted embeddings.🎯 Why
The original implementation used a Python loop to iterate over each input embedding, performing a
np.linalg.normcalculation against all fitted embeddings in each iteration. This was inefficient for large batches of input embeddings or many fitted samples.📊 Impact
Expected performance improvement of approximately 12x for batch predictions.
🔬 Measurement
Verified using
benchmark_predict.pyandextra/verify_optimization.py. The latter ensures numerical consistency (withinrtol=1e-5) between the iterative and vectorized versions, and validates backward compatibility for older models.⚡ Bolt Journal Entry
Added a learning entry to
.jules/bolt.mdregarding the use ofnp.maximum(..., 0)to maintain stability when using the expansion formula.PR created automatically by Jules for task 10186505443181998047 started by @guesswh0