GenSynth Documentation

Creating Your Entities

Before you can start a job, you must have dataset and model entities. For this tutorial you must have the simpnet_tutorial resource package. Your tutorial packages come with pre-made entities you can use (and will use later in this tutorial), but we'll walk you through how to create them , as you will want to be able to make your own in the future.

Creating Dataset Entities

Before you can start a new job, you need to create your dataset, which is an interface to your training, validation, and test data.

Click on the Entities tab. Select Dataset to go to the Dataset Manager. On this page you will create or update the datasets you will use with GenSynth.


For this tutorial, select CIFAR-10 (Tutorial) from the drop-down in the Dataset Selection field. Leave the version as Draft.

The Python interface will appear in the black area and the Dataset Interface Class name will be filled in as Cifar10ShardInterface.


DarwinAI provides samples you can use to start with or to modify to suit your needs. If you alter them, we recommend saving them with a new filename as installing updates to the samples will overwrite your changes.

The Dataset Interface Class field names a Python class in the code; this class must implement the required interface methods. This class name is saved with the Dataset entity.

At this time you may familiarize yourself with the structure of the Python module. Each method of the Cifar10ShardInterface class is documented in the code.

Once you have your dataset in place, click the Validate button to check for errors. When you have validated the dataset, you will be able to use it in the Data section of the New Job tab.

When you are ready to start importing your own datasets for use in GenSynth, refer to Preparing your Data to learn how.

Creating Model Entities

In order to start a new job, you need to have a model entity. For the purposes of this tutorial, we will use a pre-existing model entity. In the Load an existing entity field, choose Simpnet (Use with CIFAR-10).

Click on the Load Tensor Data button to load the data. If there are any errors in the Meta File or Checkpoint, you will receive an error message.

When you are providing your own model, you will have to provide the checkpoint paths and tensor names.


When you want to start importing your own models, see Preparing your Model to learn how.