Introduction¶
Learn how to serve and deploy the model using BentoML and Docker.
Requirements¶
The following requirements are necessary to follow this part in addition to those described in the first part:
- A GitHub account
- A Google Cloud account
- Docker must be installed to set up and manage the container registry (see the Setup page to install it)
Using different platforms? Read this!
While this guide uses GitHub and Google Cloud for examples, the core MLOps principles and architecture patterns apply to other platforms with some adjustments:
Note
A credit card might be necessary to use cloud services.
Before proceeding with this section, please ensure that you have a valid payment method, as it may be required to utilize cloud services. It is important to note that at the conclusion of this section, you will need to delete the cloud resources that were created to avoid any potential charges.
While the costs associated with this section are expected to be free, it is recommended to review the pricing details of cloud services before initiating this part.
State of the MLOps process¶
With the experiment versioned and reproducible, the model is ready to be served and deployed. In this part, you will address the following issues:
- Model may have required artifacts that are forgotten or omitted in saved/loaded state
- Model cannot be easily used from outside of the experiment context
- Model requires manual publication to the artifact registry
- Model is not accessible on the Internet and cannot be used anywhere
- Model requires manual deployment on the cluster
- Model cannot be trained on hardware other than the local machine
- Model cannot be trained on custom hardware for specific use-cases