TensorBoardX

TensorBoardX is a Python library that enables logging events for TensorBoard visualizations without requiring TensorFlow as a dependency, making it compatible with frameworks like PyTorch, Chainer, MXNet, and NumPy. It supports logging various data types including scalars, images, audio, histograms, text, graphs, and embeddings to help researchers visualize and track machine learning experiment progress. The current version is 2.6.5, released on April 3, 2026, with active development and maintenance.