Selecting GPU Types

Coiled supports running computations with GPU-enabled machines in all the supported cloud providers, please refer to the GPUs documentation for more information.

In both AWS and Azure, the GPU selection is tied to the instance type that you request. For example, in AWS if you select the instance type g4dn.xlarge, this instance type contains an Nvidia Tesla T4 GPU.

The cloud provider that works a bit differently is the Google Cloud Platform (GCP). In GCP, you can attach different GPUs to n1 instances.

The rest of the article will assume that you are using the GCP cloud provider for your account.

You can select the GPU type by using the gpu_type keyword argument in the coiled.Cluster() constructor. For example:

import coiled

cluster = coiled.Cluster(gpu_type="nvidia-tesla-t4")

Note that if you specify a gpu_type you don’t need to pass the worker_gpu keyword argument.

You can also specify both GPU and instance types. For example, if you wish to use a high memory instance.

import coiled

cluster = coiled.Cluster(gpu_type="nvidia-tesla-t4", worker_vm_types=["n1-highmem-2"])

GPU types allowed

Currently we allow only the nvidia-testla-t4 GPU, but you will be able to choose from a wider range of GPU types in the future. You can see a list of allowed GPU types with the command coiled.list_gpu_types().

import coiled

coiled.list_gpu_types()