Welcome to DREAMS_MC’s documentation!
Deep Report for AI Models (DREAMS)
dreams_mc is a Python package designed to simplify the process of creating comprehensive model cards for deep learning models. Model cards(mc) are crucial for documenting the performance, usage, and limitations of AI models, promoting transparency and ethical use of AI.
Features
Generate Comprehensive Model Cards: Easily produce detailed model cards including sections like project overview, dataset description, results with plots, uncertainty estimation, and references.
Configurable: Use a YAML configuration file to customize the model card content, making it adaptable to a wide range of projects and models.
Support for Various Plot Types: Integrate visualizations such as performance graphs and uncertainty estimations directly into your model cards.
Installation
The dreams_mc is hosted on PyPI. You can install the dreams_mc package using pip:
$ pip install dreams_mc
Get started
To generate a model card, ensure you have a configuration file (config.yaml) prepared with the necessary details about your model, dataset, and results. The plots/figures for publishing in the model card is stored in a folder. The folder path should be mentioned in the config file. Here’s a simple example of generating a model card:
from dreams_mc.make_model_card import generate_modelcard
# Path to your configuration file
config_file_path = '/path/to/your/config.yaml'
# Desired output path for the model card HTML file
output_path = '/path/to/output/model_card.html'
# Version number of your model
version_num = '1.0'
# Generate the model card
generate_modelcard(config_file_path, output_path, version_num)
Configuration
The config.yaml file should contain sections for your project overview, dataset description, results, uncertainty estimation, and references. Here’s a brief outline of what each section might include:
Overview: General information about the model and its purpose.
Dataset Description: Details about the dataset used, including preprocessing steps and split ratios.
Results: Model performance metrics, possibly including training/validation losses, accuracy, etc.
Uncertainty Estimation: Information on how uncertainty is quantified in your model’s predictions.
References: Citations or links to relevant papers, datasets, or other resources.
How to cite DREAMS_MC?
Note
For citing us in your publication, Click here for further details.