Google Compute Engine provides a practical way to create a cloud virtual machine for multi-GPU training when local hardware is unavailable. The trade-off is speed and capacity versus hourly cost: shut down environments when they are no longer needed. section 5.1.1
A typical setup involves:
- Creating a Google Cloud account and project, configuring billing, and installing the gcloud CLI. section 5.1.1
- Choosing a GPU-capable machine image and instance configuration. section 5.1.1
- Connecting through SSH, copying project files, and deleting the instance after use. section 5.1.1
$ INSTANCE_NAME="g2-llminprod-example"
$ gcloud compute instances create ${INSTANCE_NAME} --zone=us-west1-a
--machine-type=g2-standard-24 --image-project=deeplearning-platform-release
--image=common-gpu-v20230925-debian-11-py310 --boot-disk-size=200GB --scopes
cloud-platform --metadata=install-unattended-upgrades=False,install-nvidia-
driver=True --maintenance-policy TERMINATE --restart-on-failure
Create a GPU-enabled instance with a selected image, boot disk, cloud-platform scope, NVIDIA driver installation, and restart settings. section 5.1.1
