Model Explorer: Simplifying ML models for Edge devices- Google Developers Blog

Introduces Model Explorer, a graph visualization tool for ML models, supporting JAX, PyTorch, TensorFlow, and TensorFlow Lite. It runs locally or in Colab, visualizes large graphs (tens of thousands of nodes) at 60 FPS using WebGL and three.js, and supports metadata overlays. Use cases include understanding model architecture, debugging conversion errors (e.g., PyTorch to TFLite), and identifying performance/numeric bottlenecks with per-node data overlays. It's used internally by teams like Waymo and Google Silicon for optimizing on-device models like Gemini Nano.