Microsoft has open-sourced its debugging and visualization tool, called TensorWatch, for deep learning and AI. TensorWatch is designed to be flexible and extensible so data scientists and build their own custom visualizations, UIs, and dashboards.
“We like to think of TensorWatch as the Swiss Army knife of debugging tools with many advanced capabilities researchers and engineers will find helpful in their work. We presented TensorWatch at the 2019 ACM SIGCHI Symposium on Engineering Interactive Computing Systems,” said Microsoft in a blog post.
To aid customization, TensorWatch uses Jupyter Notebook instead of prepackaged user interfaces. This who don’t know, [Jupyter Notebook] is an open-source web application for creating and sharing documents with contain live code, equations, visualizations and narrative text. It’s tested PyTorch 0.4-1.x., but mostfeatures should also work with TensorFlow eager tensors, said the GitHub page of the project.
TensorWatch has a Lazy Logging Mode that helps users in executing arbitrary queries against live ML training process, return a stream as a result of the query and view this stream using the choice of a visualizer.