PerceptiLabs is a dataflow driven, visual API for TensorFlow, carefully designed to make machine learning (or deep learning) modeling as intuitive as possible.
The components have settings that you can change and tune
A component automatically generates a visualization of the output
The visualizations update in real-time when you change the component settings
The components auto-generate TensorFlow low-level code, which you can view and edit
The dataflow is defined by connecting the components
The model architecture is visualized as you are building your model output
The shape of the inputs and outputs are automatically calculated, and configs/hyperparameters suggested, to quickly get you up and running
PerceptiLabs shows you tips, warnings and errors while you are building your model
Quickly see where something is wrong and fix it
Debug the model and the code
Get instant feedback to better understand your model
Real-time granular visualizations
Get an overview of the model performance and predictions
Visualize each components variables, such as weights, biases, gradients and outputs
Real-time debugging example:
Understand the gradients if it is vanishing or exploding and where in the model it happens
