Get to know AutoML delivery flow
The chapter describes the end-to-end delivery flow of a machine learning (ML) project.
AutoML complexity analysis
When delivering an ML project, you can face various requirement complexities, leading to different implementation approaches. Different roles can address different use cases, and collaboration between roles can vary.
Not all use cases are available for implementation with out-of-the-box or custom models created by machine learning engineers (MLE). Some require the involvement of data scientists (DA). See the complexity decision diagram below.
Search Engine and training
In WorkFusion, new ML models are trained in two stages:
- Search Engine finds the best parameters or a metamodel for a supplied training set. For more information, read the Compare Search Engines topic.
- Training itself uses the metamodel and optimized parameters to produce the ultimate model.
Model training and tuning
MLEs can train and tune models by coding with AutoML SDK. In addition, both engineers and non-engineer users can leverage the AutoML menu in Control Tower to do the following without coding:
To address more complex use cases or speed up model delivery, MLEs can also use ML Ops Platform, which is a Kubernetes-based optional add-in to the core Product. The add-in is language- and framework-agnostic and allows you to work with complex documents and more powerful models than AutoML SDK ones, including LayoutLM deep learning pipelines. You can work with models trained in ML Ops Platform using the AutoML user interface in Control Tower.