What is Kubeflow
Kubeflow is an open-source platform running on top of Kubernetes and designed to make machine learning (ML) workflow simple, portable and scalable.
Below are some reasons to adopt Kubeflow in an enterprise for their ML journey
- Kubeflow helps in experimentation with training an ML model by providing easy UI also provide subsystems for training models like Jupyter Notebook and few popular ML operators like TensorFlow and PyTorch.
- Kubeflow can handle end to end hybrid and multi-cloud ML workloads
- Kubeflow run on Kubernetes thus kubeflow has all the capability of Kubernetes. Kubeflow can manage the entire AI organization at scale and still be able to maintain the same quality of control
- Kubeflow pipeline can be used to create workflows which can be used for continuous integration and deployment (CI/CD) for ML
- Kubeflow hyperparameter tuner (Katib) can be used to automate tuning of the model
Below are some products similar to kubeflow
- ML Flow – https://mlflow.org/
- TensorFlow Extension – https://www.tensorflow.org/tfx
- AI Platform – https://cloud.google.com/ai-platform
- Current Version – As of March 2021 version 1.2 is GA
- Documentation – Product documentation can be found at https://www.kubeflow.org/docs/
Script to setup kubeflow in Azure environment can be found at https://github.com/volksinfotech/kubeflow