💡 Guideline

Set up a clean virtual environnement

Linux setting:

pip install virtualenv
virtualenv myenv
source myenv/bin/activate

Windows setting:

pip install virtual env
virtualenv myenv
.\myenv\Scripts\activate

Install the library

You can install it by directly downloading from PyPi using the command:

pip install dqm-ml

Or you can installing it from the source code by launching the following command:

pip install .

Usage

Each metric is used by importing the corresponding modules and class into your code. For more information about each metric, refer to the specific README.md in dqm/<metric_name> subfolders

Available examples

Many examples of DQM-ML applications are avalaible in the folder /examples

You will find :

2 jupyter_notebooks:

• multiple_metrics_tests.ipynb : A notebook applying completeness, diversity and representativeness metrics on an example dataset.

• domain_gap.ipynb : A notebook demonstrating an example of applying domain_gap metrics to a generated synthetic dataset.

4 python scripts:

Those scripts named main_X.py gives an example of computation of approaches implemented for metrics on samples.

The main_domain_gap.py script must be called with a config file passed as an argument using --cfg.

For example:

python examples/main_domain_gap.py --cfg examples/domain_gap_cfg/cmd/cmd.json

We provide in the folder /examples/domain_gap_cfg a set of config files for each domain_gap approaches`:

For some domain_gap examples, the 200_bird_dataset will be required. It can be downloaded from this link. The zip archive will be extracted into the examples/datasets/ folder.