OxViz – Onnx Graph Analyzer
Zan Zagar
OxViz is an ONNX model analyzer for iPhone, iPad, and Mac. Import an .onnx file, inspect the graph visually, search operators, trace model flow, review tensor shapes and initializers, and export clean documentation for debugging, sharing, or client work. Import an ONNX model and instantly explore its graph structure, operators, tensors, inputs, outputs, weights, attributes, parameter counts, opsets, and model metadata in a clean native interface. Built for ML engineers, app developers, researchers, and students, OxViz helps you understand what is inside a model without uploading proprietary files to a server. Key features: • Import and inspect ONNX model graphs • View operators, inputs, outputs, tensor shapes, attributes, initializers, and parameter counts • Search nodes and jump through large graphs quickly • Trace upstream or downstream from selected nodes • Review model metadata including IR version, producer info, opsets, inputs, outputs, weights, and operator frequency • Run a Model Health Report to spot common structural issues such as missing shape metadata, disconnected nodes, duplicate names, unresolved graph inputs, and unusual topological flow • Export graphs as PDF, SVG, PNG, or editable .drawio diagrams • Export a professional Model Health Report PDF • Works locally on device, with no account required and no model upload • Native iPhone and iPad experience, with support for compatible Mac use through the iPad app • Automatic graph layout optimized for readability OxViz is designed for real debugging workflows: checking exported PyTorch or TensorFlow models, reviewing model structure before deployment, documenting architectures, comparing what a model contains against what you expected, and preparing ONNX files for Apple-platform work. The app is free to try. Upgrade with a one-time in-app purchase to unlock unlimited imports, exports, and Pro features.