Interpret¶
Model interpretability is often very specific to the model and thus TorchX provides examples of how to create model interpretability apps and components.
We recommend creating a custom application and component built using Captum . Captum provides a number of out of the box analyzers and renderers to understand your model.
See the Model Interpretability App Example and the corresponding Interpret component definition for an example of how to use Captum with TorchX.