TensorBoard
An open-source ML visualization tool that runs locally or on your server, works offline, and compares TensorFlow training metrics and model graphs.
TensorBoard is a browser-based toolkit for inspecting TensorFlow experiments on your own machine or server. It's for researchers and ML developers who need to understand training behavior, compare runs, and investigate model performance. It works entirely offline, including behind a corporate firewall or in a datacenter, so experiment data can stay within your own environment.
Its charts track loss, accuracy, learning rate, and other recorded metrics over time. You can compare separate runs to see how a change in hyperparameters affects convergence, rather than assessing each experiment in isolation. TensorBoard organizes recorded data by named tags, which helps keep related measurements together.
Model inspection goes beyond metrics. Graph views show operations and layers, while histograms reveal how weights, biases, and other tensors change during training. Embedding projections reduce dimensionality so you can examine their structure visually. TensorBoard also displays recorded images, text, and audio alongside the experiment's numerical results.
The toolkit includes profiling for TensorFlow programs when you need to investigate execution performance. Its dashboards read experiment records saved to disk and can combine event files into a single run history after a restart. TensorBoard is open source under the Apache 2.0 license.
