Operational Procedures
Operational Procedures
- Daily operation guidelines
- User support and query handling
- Feedback collection and management
Data Management procedures
This document outlines the data management practices for training the machine learning models. All of these practices have been taken from the literature and the sources have been cited along with them.
Training set and test set size - Data Science Stack Exchange
How Much Training Data is Required for Machine Learning? - MachineLearningMastery.com
Reporting guidelines:
- Model Cards: https://modelcards.withgoogle.com/about
- Data Cards: https://sites.research.google/datacardsplaybook/
- Healthsheets: https://arxiv.org/abs/2202.13028
- Templates: https://github.com/PAIR-code/datacardsplaybook/tree/main/templates
- Reforms checklist: https://reforms.cs.princeton.edu/
- Data Version Controlling
MLOps softwares we could use:
- MLflow
- Weights and Biases
The MLflow seems to be a better software for ML model management whereas we should be careful about the data management as well which we can maintain using the data version controlling software.