tensorflow/dtensor-gcp-examples

Using DTensor on Google Cloud

DTensor GCP Examples

This project contains examples of using multi-client DTensor on GCP with a cluster of GPUs or TPUs.

Prerequisites

1. gcloud environment on the local console:

gcloud auth login ...
gcloud config set project  ...

2. A GCS bucket that the GCE service account can write into. The bucket is used to demo checkpointing. Set the prefix paths name with

export GCS_BUCKET=<bucket_name>

or edit bootstrap.sh.

For maintainers

Since this requires tf-nightly, periodically update requirements.txt in each directory to known 'good' versions for these examples.

dtensor-gcp-examples's Issues

requirements.txt references deprecated versions of TensorFlow

The requirements.txt file for both gpu/gce-cluster and gpu/gce-node list tf-nightly==2.11.0.dev20220910. This version is no longer supported on the GCP instances that are created.

Which version should be used, or does the code need to be updated for a more recent version of tf?