TensorFlow code vs TensorFlow GUI by PerceptiLabs - Watch our Live Coding event here !
A visual modeling tool

To build a custom model like simple linear regression, or something more complex like a GAN, just drag and drop components to start.

PerceptiLabs wraps low-level TensorFlow code to create visual components, so you’ll see your model’s architecture as you build.

You’ll build models faster and easier with PerceptiLabs.
Real-time insights

See real-time analytics and granular previews of output from each model component. You can easily track and understand the gradients’ behavior, perform real-time debugging, and see where to optimize your model.

It's never been easier to debug and interpret your model.
Keep track of models and share on GitHub

PerceptiLabs lets you manage multiple models, compare them, and easily share the results back to your team. Export your model as a TensorFlow model or as a Jupyter Notebook.

That’s machine learning at warp speed.
Get started now. Free for developers.


$ pip install perceptilabs
click to copy
$ perceptilabs
click to copy


Check out some of our features
Toggle between GUI and code
Export and share to GitHub and Jupyter
Automatically distributes training on available GPUs
Auto-generated visualizations for every single variable in the model
Easy debugging with warnings and recommendations


v0.11 Release


PerceptiLabs named to the CB Insights AI 100


Getting inside the head of a machine learning scientist