Creating an Experiment
To create an experiment, you first need to specify few options including cluster, resource, image, start command, volume, environment variables and termination protection.

Runtime (Required)

By now you should have your own custom cluster added to Vessl. Here, you can choose between your custom cluster or Vessl's managed cluster. As noted previously, your custom cluster could be either on-premise or on-cloud. Vessl's managed cluster is on the cloud vendor server.
Vessl's Managed Cluster
Custom Cluster
If you wish to use Vessl's managed cluster, you should specify the type of the resource that the pod will use. Select the Resource under the dropdown menu.
Vessl also supports spot instances:
Check out the full list of resource types and the corresponding prices:
For the Custom Cluster, you should specify the processor type and resource requirements. The experiment job will be automatically assigned to an available node according to the given resource requirements.
Learn more about Clusters:

Image (Required)

You can choose Docker image that the experiment container will use. There are two types of images: the Python image and the public image.
Python Image
Docker Image
Python Images are pre-pulled images provided by Vessl. You can find the available image tags in Vessl's Amazon ECR Public Gallery. These images include the frequently used PyTorch and Tensorflow images. Some of them are pushed by Vessl. You can find detailed information in the Readme file in Docker Hub. Clicks PACKAGES to list and view the installed pip packages.
You can pull images from either Docker Hub or Amazon ECR. Docker Images are not managed by Vessl.

Public Images

To pull images from the public Docker registry, simply pass the image URL. The example below demonstrates pulling the official TensorFlow development GPU image for Docker Hub.

Private Images

To pull images from the private Docker registry, you should first integrate your credentials in organization settings:
Then, check the private image checkbox, fill in the image URL, and select the credentials you have just integrated. The example below demonstrates pulling a private image from AWS ECR.

Start Command (Required)

You must specify the start command in the experiment container. You can put a running script with command-line arguments. You can put multiple commands by using the && command.


You can mount the project, dataset, and files to the experiment container.
Learn more about Volume mount on Vessl:

Environment Variables

Users can set environment variables as key-value pairs. A typical experiment will include variable like LEARNING_RATE and EPOCHS.

Termination Protection

Termination protection allows you to access the experiment container. If you check the checkbox, the experiment will go idle once it finishes running.
Last modified 23d ago