openrai 0.0.6

Documentation

Documentation is available online: https://responsibleai.readthedocs.io/en/latest/

To Install

1. The project uses sqlite as storage and requires no additional installation for this.

   pip install openrai

2. If you need to use the dashboard as well you must install the additional requirements

   pip install openrai[dashboard]

3. To start using the library you must configure the folder where the results will be stored and be used by the dashboard

   export DATABASE_FOLDER=YOUR_LOCATION

4. To start the dashboard server use the following command:

   python3 -m RAIDashboard

Demos:

We have added a few demo projects to showcase some of the capabilities of OpenRAI.
to run any of the demos please use 'python demo_filename'. For instance :
python ./demos/adult_demo_grid_search.py

below is a short description of the provided demos:

File: adult_demo_grid_search.py
Description: This demo uses the Adults dataset (https://archive.ics.uci.edu/ml/datasets/adult) to show how OpenRAI can be used in model selection

File: image_class_analysis.py
Description: this demo uses Cifar10 dataset and shows how openRAI can be used to evaluate image classification tasks

File: image_class_training.py
Description: this demo uses Cifar10 dataset and shows how openRAI can be used monitor image processing tasks during training


File: tabular_class_console.py
Description: this demo shows how openRAI can be used without the dashboard to calculate and report on the metrics for a machine learning task

File: text.py
File: text_output.py
Description: these demos show how openRAI and its dashboard can be used for evaluating the natural language modeling tasks

Cisco Research, Emerging Tech and Incubations,

Cisco Systems Inc.