-
Notifications
You must be signed in to change notification settings - Fork 1.2k
Open
Description
Describe the feature you'd like
PyTorch Horovod MPI is supported in sagemaker-pytorch-training-toolkit, however, a Python SDK interface is not provided yet to start distributed training using mpi.
How would this feature be used? Please describe.
With this implementation, user could enable MPI runner like this:
Something similar is done for TensorFlow: https://github.com/aws/sagemaker-python-sdk-staging/blob/master/src/sagemaker/tensorflow/estimator.py#L96
Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.
Additional context
Add any other context or screenshots about the feature request here.