What can I inspect inside my models when using the Zetane Insight Engine?
Everything. The model architecture and all the internal tensors, each layer output, feature maps, weights, biases, values and operator attributes.
How easily does Zetane integrate with existing AI frameworks?
After saving your Keras, Pytorch or ONNX model from your own workflow, you can load your model so that you can inspect, debug and validate it in Zetane Viewer. Using the Zetane python package, you can also use the Python-Zetane API to send models, NumPy arrays, text, images, point clouds and meshes to the Zetane Engine.
Can I pass my own inputs through the model for validation?
The Zetane Viewer Pro allows you to load your own input so you can identify, validate and determine how to improve your model.
How do you make all the model information easily understandable?
By carefully encapsulating the data. At the highest level you can navigate your model architecture, zoom on specific nodes. Clicking on the tensor buttons allows you to inspect the tensor and values distributions. Then, each tensor has different types of visualization to choose from.
Can you open MLP, NLP and other types of models?
If the model can be saved you should be able to open it to inspect all its components. Regardless of their purpose, ML models are composed of operation and tensors which we help you access with no effort.