I wanted to train a model using distributed tensor flow on my kubernetes cluster. Please suggest ways other than kubeflow