scikit-learn-intelex
scikit-learn-intelex copied to clipboard
intel
PCA perf fix and out of preview
The PR aims to improve the performance of oneAPI PCA implementation and refactor the codes to align with sklearnex design principals. Associated oneDAL PR#2601 (should be merged before to ensure the current PR is working properly).
Changes proposed in this pull request:
- Remove mean center from sklearnex (implemented in oneDAL)
- Align dispatcher with existing algorithms
- Refactor code to provide better structure and to remove redundant parts
![azure-pipelines[bot] avatar](https://avatars.githubusercontent.com/in/9426?v=4 "Published by a pub.dev verified publisher"){kind=small}
Perhaps not within scope of this PR, but still facing issues with spmd pca using n_components with float and generally with predict method.
@md-shafiul-alam if there some blocker, can we just leave updated pca in preview and merge it as is? You can probably do required minor fixes as follow up.
I am not sure if there is anything else to be done to move PCA out of preview other than finalizing this PR
I am not sure if there is anything else to be done to move PCA out of preview other than finalizing this PR
Finalizing this PR in case of moving out of the preview taking a long time. I already need this changes on main. Let's leave currently all changes on preview and merge. Afterwards we can do moving out PR
transform method is already wrapped, and fit_transform calls fit and transform. Do we still need to wrap fit_transform? @samir-nasibli
transform method is already wrapped, and fit_transform calls fit and transform. Do we still need to wrap fit_transform? @samir-nasibli
It could probably work, but in this case you will do the same data transfer several times from device to host and back. Better to do it at once.
