Batch Job

A Batch Job corresponds to a one-off task that runs to completion and then stops, like training a model or processing data. DGX Cloud Lepton provides a simple way to run and manage batch jobs on the cloud.

Create Batch Job

Navigate to the create job page.

In the resource section, select the node group and your desired resource shape. In this case, we use H100-80GB-HBM3 x 1 from node group h100sxm-0, and leave worker and priority to the default value.

In the container section, we can just use the default image, and in the Run Command field, we can specify the following command to run in the container.

for i in {1..10}; do
  echo "Processing item $i/10..."
  sleep 2
done
echo "Batch job completed successfully!"

By configuring the above settings, we can create a batch job that

• Use one H100 GPU from node group h100sxm-0
• Use default image, run a simple counter that processes 10 items with a 2-second delay between each
• Complete after approximately 20 seconds, and will be archived after 3 days.

View the Batch Job

After the job is created, you can check the status and result of the job in the details page.

Click on the Logs button under Replicas tab, you can see the logs of each replica. Because we have set the worker to 1, so there will be only one replica, and the logs should output the following message.

Processing item 1/10...
Processing item 2/10...
Processing item 3/10...
Processing item 4/10...

Next Steps

For more features about Batch Job, please refer to the following documents:

• Configure your Batch Job
• Job templates
• Predefined environment variables