Question: What is the recommended way for Data Scientists to run a distributed training job

[mChowdhury-91]

The Data Scientists do not have access to K8s cluster, and hence cannot use commands like kubectl create -f -n kubeflow, in that case what is the recommended way to run distributed training jobs.
Is Kubeflow Pipeline the right approach or Fairing?

[johnugeorge]

You can use python sdk for it . https://github.com/kubeflow/training-operator/tree/master/sdk/python

[mChowdhury-91]

@johnugeorge How do we run a MPIjob using the Python SDK. Is there any api to call the MPIJob yaml file

[mChowdhury-91]

@johnugeorge Can we directly provide the configuration yaml file https://github.com/kubeflow/training-operator/blob/master/examples/tensorflow/dist-mnist/tf_job_mnist.yaml or https://github.com/kubeflow/training-operator/blob/master/examples/pytorch/mnist/v1/pytorch_job_mnist_mpi.yaml to trigger the training via a Kubeflow pipeline?

[johnugeorge]

you can take a look at https://github.com/kubeflow/katib/blob/master/examples/v1beta1/kubeflow-pipelines/kubeflow-e2e-mnist.ipynb