2.3.7. kubeflow

Kubeflow 是由 Google 等公司于 2017 年推出的基于 Kubernetes 的开源项目,它支持常见的机器学习框架。

Kubeflow has three core components:

1. TF Job Operator and Controller:
    Extension to Kubernetes to simplify deployment of distributed TensorFlow workloads.
2. TF Hub:
    Running instances of JupyterHub, enabling you to work with Jupyter Notebooks.
3. Model Server:
    Deploying a trained TensorFlow models for clients to access and use for future predictions.