Pipeline artifacts stored in cloud storage – Pipelines using TensorFlow Extended

by Ron Mutschler ASP.NET Certifications, Creation of entity type, Exams of Google, Google Certification Exams

Step 10: Pipeline artifacts stored in cloud storage

Pipeline artifacts are stored in the cloud storage as shown in Figure 8.15:

  • Module folder contains Python files which we had pushed while creating transform and trainer components.
  • Output_model contains trained classification model.
  • Root folder contains artifacts of each of the component of the pipeline:

Figure 8.15: Pipeline artifacts stored in the cloud storage

Deletion of resources

We have utilized workbench, cloud storage to store the data and the artifacts of the pipeline. For deletion of resources, ensure to delete the workbench, clear the data stored in the cloud storage.

Conclusion

We learnt about the TFX, a few of its components and constructed pipeline using some of the standard components. Also, we understood how to use Kubeflow for the orchestration of TFX pipeline on vertex AI.

In the next chapter, we will start understanding and working on feature store of the Vertex AI.

Questions

  1. Which artifacts of the transform component is used in the training component of the pipeline?
  2. What are the different orchestration options TFX supports?
  3. Try using evaluator component between trainer and pusher component and re-construct the pipeline. (Use evaluator component to evaluate the model performance and push it only if the performance is good).