1. To Version control
  2. To Collaborate with peers
  3. To stop reinventing the wheel
  4. To allow rollbacks/rollforwards
  5. To track progress empircally for your model
  6. To enable true A/B tests
  7. To keep track of things
  8. To enable automation (MLOps)
  9. To assess the outputs of the MLModels
  10. To track business value

Alot of these reasons is why Hugging face has been successful (https://huggingface.co/)

Covers: theory of Why register a ML Model?
Estimated time needed to finish: 1 minutes
Questions this item addresses:
  • Why do you need a ML Model Registry?
How to use this item?
  1. Version your model
  2. Share models among the different teams
Author(s) / creator(s) / reference(s)
Denys Linkov
0 comment

Registering ML Models

Total time needed: ~38 minutes
To learn how models can be stored, tracked and collaborated on.
Potential Use Cases
Repeatable and collaborative ML projects
Who is This For ?
INTERMEDIATEMachine learning engineers and data scientists who are looking on how to store and iterate on their models.
Click on each of the following annotated items to see details.
ARTICLE 1. How to build a ML Model Registry
  • Why have a ML Registry?
  • How to build a ML registry yourself?
7 minutes
ARTICLE 2. Storing ML Models in Sagemaker
  • How to register a ML Model in sagemaker?
  • How are models grouped?
  • How to deploy a model?
20 minutes
ARTICLE 3. Storing ML Models in Azure ML
  • How to register a ML Model in Azure ML?
  • How to tag models in Azure ML?
5 minutes
WRITEUP 4. Storing ML Models in ML Flow.
  • How to register a ML Model in ML Flow?
5 minutes
WRITEUP 5. Why do we need a ML Model registry?
  • Why do you need a ML Model Registry?
1 minutes

Concepts Covered

0 comment