Repolode
Submit
tensorflow

tensorflow/tensorboard

TensorBoard is a web-based visualization toolkit for inspecting TensorFlow training runs, metrics, graphs and embeddings.

Open-source alternative to:Weights & Biases

View on GitHub
Stars
7,241
Forks
1,723
Open issues
758
Last commit
2 days ago

What is it?

TensorBoard is a suite of web applications for inspecting and understanding TensorFlow runs and graphs. It reads event files written by summary writers into a log directory, then displays scalars, histograms, distributions, images, audio, text and computation graphs in the browser.

Runs are loaded automatically from a log directory and its subdirectories, so multiple experiments can be compared side by side. Dashboards organize data by tag, and the scalar charts support zooming, crosshair readouts and custom folders via regular expressions. The Embedding Projector visualizes high-dimensional embeddings from checkpoints, and the Graph Explorer helps inspect model structure using name scopes.

TensorBoard is designed to run fully offline, locally or behind a firewall. It can be used in Chrome or Firefox, and its plugin system allows custom visualizations to be added.

Key features

  • 01Scalar dashboard for tracking metrics such as loss and learning rate over time
  • 02Histogram and distribution views for tensor values across training steps
  • 03Image, audio and text dashboards for summary data
  • 04Graph Explorer for inspecting TensorFlow model structure
  • 05Embedding Projector for visualizing high-dimensional embeddings
  • 06Plugin system for adding custom visualizations

Use cases

  • Monitoring loss and accuracy curves during model training
  • Comparing hyperparameter experiments across multiple runs
  • Inspecting the structure of a TensorFlow computation graph
  • Exploring word or image embeddings from a trained model

Quick start

  1. 1

    Install TensorBoard with pip

    pip install tensorboard
  2. 2

    Point TensorBoard at your log directory and start it

    tensorboard --logdir path/to/logs
  3. 3

    Open http://localhost:6006 in your browser

Things to know

  • Requires summary data written to a log directory by TensorFlow before it shows anything.
  • Officially supported only in Google Chrome or Firefox; other browsers may have bugs.

In collections

Diagrams, charts & animated visualizations

Topics

#tensorflow#machine-learning#visualization#deep-learning#monitoring#embeddings

FAQ

What is tensorboard?
TensorBoard is a web-based visualization toolkit for inspecting TensorFlow training runs, metrics, graphs and embeddings. TensorBoard is a suite of web applications for inspecting and understanding TensorFlow runs and graphs. It reads event files written by summary writers into a log directory, then displays scalars, histograms, distributions, images, audio, text and computation graphs in the browser.
Is tensorboard free to use? What license does it have?
tensorboard is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is tensorboard still maintained?
Actively maintained. The last commit to tensorboard was on Nov 4, 2026.
What can I use tensorboard for?
Monitoring loss and accuracy curves during model training Comparing hyperparameter experiments across multiple runs Inspecting the structure of a TensorFlow computation graph Exploring word or image embeddings from a trained model
How do I install tensorboard?
Install TensorBoard with pip: pip install tensorboard Point TensorBoard at your log directory and start it: tensorboard --logdir path/to/logs Open http://localhost:6006 in your browser
Where can I get tensorboard?
The source code and installation instructions are on GitHub: https://github.com/tensorflow/tensorboard

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects