Convert scikit-learn models to PyTorch modules

sk2torch converts scikit-learn models into PyTorch modules that can be tuned with backpropagation and even compiled as TorchScript.

Problems solved by this project:

1. scikit-learn cannot perform inference on a GPU. Models like SVMs have a lot to gain from fast GPU primitives, and converting the models to PyTorch gives immediate access to these primitives.
2. While scikit-learn supports serialization through pickle, saved models are not reproducible across versions of the library. On the other hand, TorchScript provides a convenient, safe way to save a model with its corresponding implementation. The resulting models can be loaded anywhere that PyTorch is installed, even without importing sk2torch.
3. While certain models like SVMs and linear classifiers are theoretically end-to-end differentiable, scikit-learn provides no mechanism to compute

    

    

    

   To finish reading, please visit source site