Optimize token_classification/filter.py and find_label_issues_batched.py for performance #1072
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Outdated suggestions cannot be applied.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.
Summary
This PR partially addresses #862
[ ✏️ Write your summary here. ]
We could skip one list comprehension in find_label_issues. The flatten part for the pred_probs is a lot faster using np.vstack and iterating only on the top level of the list.
In addition, after profiling the function I found that most of the work was done in the label_issues_batched so I made a few changes there also. By batching the label_issues_mask we could reduce the memory usage significantly while slightly improving the runtime at the same time.
For memory I used the memory-profiler library. The code I used for benchmarking is copied below. In addition I sorted the imports in the modified files.
Code Setup
Current version
This PR
Testing
References
Reviewer Notes