file:///home/mariana.almeida/%C3%81rea%20de%20Trabalho/aits.png
KUBEFLOW PIPELINES X TFX PIPELINES
file:///home/mariana.almeida/%C3%81rea%20de%20Trabalho/KUBE.png
What constitutes an AI Platform instance?
- Containerized implementations of ML tasks
- Containers provide portability, repeatability and encapsulation.
- A task can be single noded or distributed
- A containerized task canc invoke other services
- Specification of the sequence of the steps
- Specified via python SDK
- Input parameters
Current process for building MLOPs pipeline
- Set up a GKE (google kubernetes engine) cluster
- Create a cloud storage bucket for storing data
- Install kubeflow pipelines
- Set up port forwarding
- Create a process to share pipeline with the team
GOOGLE AI PLATFORM PIPELINES DOES ALL THOSE STEPS
file:///home/mariana.almeida/%C3%81rea%20de%20Trabalho/aiplt.png
Overview of MLOPs
file:///home/mariana.almeida/%C3%81rea%20de%20Trabalho/mlopsgpc.png
System Overview
file:///home/mariana.almeida/%C3%81rea%20de%20Trabalho/systemoverview.png