daisybell 0.3.2

Scan AI models for problems

daisybell

A scanner that will scan your AI models for problems. Currently it focuses on bias testing. It is currently pre-alpha.

How to Use

First install it:

pip install daisybell

Run it in this manner (currently supports models from HuggingFace’s repository):

daisybell --huggingface roberta-base --task fill-mask

The scan can output files for further analysis:

daisybell --huggingface roberta-base --task fill-mask --output results/roberta-base

Here is another example with a different bias task.

daisybell --huggingface cross-encoder/nli-distilroberta-base --task zero-shot-classification

That’s it for now. More will come.

Future Work

• More bias tests. More metrics for bias testing based on the research in the field.

• Integration with other types of testing (eg. adversarial robustness)

• More kinds of models besides HuggingFace models. We are especially interested in MLFlow integration.

• Documentation.

Please contribute if you can. Help is always helpful.

License

Apache

Credit

A project of IQT Labs.