Standalone model vs Model pipeline

The AI Verify Toolkit supports two modes of accessing the AI models to be tested.

Modes	Framework Libraries Supported	Dataset Type
Upload AI Model	LightGBM, Scikit-learn, Tensorflow, XGBoost, Keras, PyTorch	Tabular Only
Upload Pipline	Scikit-learn pipeline, Keras, PyTorch	Tabular, image
Connect to AI Model API	Any AI Framework	Tabular Only

The list of datatype formats supported are as follows:

Dataset Type	Formats Supported
Tabular	Pandas, Delimiter-separated Values (comma, tab, semicolon, pipe, space, colon)
Image	.jpeg, .jpg, .png

Upload AI Model

Upload Pipeline

If your dataset requires pre-processing before being fed into the prediction model, you can upload the pre-processing functions together with your model as a pipeline folder.

Currently, the toolkit supports a limited set of models. Check out the full list of framework and algorithm types supported.