This talk is about the open source package skorch, a wrapper library that allows you to combine the best of sklearn and PyTorch. It covers when it makes sense to use skorch and highlights interesting features.
Scikit-learn is the most widely used general purpose machine learning library in Python. It has a very nice API and many tools are built around it, but its support for neural networks is very limited.
PyTorch is a deep learning platform that makes working with neural networks a blast. However, users often find that they need to write boilerplate and wrapper code to integrate PyTorch into their existing workflows.
A possible solution to those problems is skorch. This open source Python library allows you to easily wrap your PyTorch modules to give them a high level sklearn API. This allows you to:
- tap into the power and flexibility of PyTorch
- get rid of most of the boring boilerplate code
- combine PyTorch nets with the usual sklearn goodies like
Pipelines,GridSearchCV, etc.
Ideally, you already have some basic knowledge of sklearn and PyTorch, or are an advanced user of one of the two libraries who is curious about improving your workflow.
Deep Learning, Data Science, Machine Learning
Domain Expertise:some
Python Skill Level:basic
Abstract as a tweet:Combine the best of sklearn and PyTorch by using skorch. This talk shows you why and how to use skorch and what cool features it has to offer.