wtf is netron?

Netron is a viewer for machine learning and deep learning model files. It reads model files in a wide range of formats and renders an interactive graph that shows the model layers, connections between them, and the properties of each operation such as input and output tensor shapes, data types, and operator parameters. The formats supported include ONNX, TensorFlow SavedModel and frozen graph, TensorFlow Lite, PyTorch exports, Core ML, Keras, Caffe, DarkNet, scikit-learn serialized models, and many others. This breadth makes it a useful inspection tool regardless of the framework a model was built in. Netron works in multiple environments. It is available as an Electron-based desktop application for macOS, Windows, and Linux. It also runs in the browser where users can drag and drop a model file directly, and it can be embedded as a web component in other applications. Using it is straightforward: open a model file, and Netron renders the computation graph. Clicking on any node shows its attribute details in a side panel. For large models this provides a practical way to understand the architecture without reading serialized binary formats by hand. When to use it: Netron is most useful for machine learning engineers and researchers who want to inspect a model structure quickly, verifying that an export preserved the expected layers, debugging mismatches between frameworks, or simply understanding an unfamiliar model file received from someone else. It is particularly handy when working with ONNX models during cross-framework conversions. The project is written in JavaScript and distributed as an npm package as well as standalone desktop installers for all major platforms.