TensorSpace

Overview

TensorSpace is an open-source JavaScript framework for rendering 3D visualizations of neural network architectures in the browser. It loads pre-trained models from TensorFlow, Keras, or TensorFlow.js to display layer-by-layer structure and activation flows interactively.

Best for

Best for
Educators and presenters who need browser-based 3D model architecture demos.

Use cases

• Demonstrate deep learning model internals in presentations or documentation
• Debug model architecture by inspecting layer shapes and connections visually
• Build interactive educational demos for neural network concepts

Notes

TensorSpace is an open-source JavaScript framework for rendering 3D visualizations of neural network architectures in the browser. It loads pre-trained models from TensorFlow, Keras, or TensorFlow.js to display layer-by-layer structure and activation flows interactively.

5,175 stars on GitHub. Last updated 2022-12-05. Licensed Apache-2.0.

Use cases

• Demonstrate deep learning model internals in presentations or documentation
• Debug model architecture by inspecting layer shapes and connections visually
• Build interactive educational demos for neural network concepts

Pros

• No backend needed, runs entirely in the browser
• Supports models from major TensorFlow ecosystems
• Visualizes activations and intermediate outputs intuitively

Cons

• Limited to static pre-trained models, no live training or inference
• Unmaintained (last commit 2020) with no recent updates
• Only works with Keras or TensorFlow.js serialized model formats

Indexed from awesome-llmops and enriched against its public facts.