Publishing models

This feature is not available as a Cypher procedure or function in Aura Graph Analytics.

By default, a trained model is visible only to the user that created it. Making a model accessible to other users can be achieved by publishing it.

Syntax

Publish a model from the catalog:

CALL gds.model.publish(modelName: String)
YIELD
    modelName: String,
    modelType: String,
    modelInfo: Map,
    creationTime: DateTime,
    trainConfig: Map,
    graphSchema: Map,
    loaded: Boolean,
    stored: Boolean,
    published: Boolean

Table 1. Parameters
Name	Type	Default	Optional	Description
modelName	String	`n/a`	no	The name of a model stored in the catalog.

Table 2. Results
Name	Type	Description
modelName	String	Name of the model.
modelType	String	Type of the model. Indicates what training algorithm was used to train the model.
modelInfo	Map	Detailed type-specific information about the trained model.
creationTime	Datetime	Time when the model was created.
trainConfig	Map	Train configuration used for training the model.
graphSchema	Map	Schema of the graph on which the model was trained.
loaded	Boolean	True, if the model is loaded in the in-memory model catalog.
stored	Boolean	True, if the model is stored on disk.
published	Boolean	True, if the model has been published.

Examples

In this section we will illustrate how to publish a model. A pre-requisite for this operation is that a model has already been trained and registered in the model catalog. We will assume here that two models named my-model1 and my-model2 have already been trained and exist in the model catalog. Our initial state can be inspected by listing all models in the catalog:

Listing information about all models:

CALL gds.model.list()
YIELD modelName, modelType, modelInfo, loaded, stored, published

Table 3. Results
modelName	modelType	modelInfo	loaded	stored	published
`"my-model1"`	`"example-model-type"`	`{exampleModelInfo="exampleValue"}`	`true`	`false`	`false`
`"my-model2"`	`"example-model-type"`	`{number=42}`	`true`	`false`	`false`

Publishing a model

All we need to do is to run the procedure and specify the name of the model we want to publish.

Publishing a trained model:

CALL gds.model.publish('my-model1')
YIELD modelName, modelType, modelInfo, loaded, stored, published

Table 4. Results
modelName	modelType	modelInfo	loaded	stored	published
"my-model1_public"	"example-model-type"	{exampleModelInfo="exampleValue"}	true	false	true

We can see that the model name is modified with the _public suffix. When we now list the models in the catalog we can see that the published model has changed.

Listing all models, including the published model:

CALL gds.model.list()
YIELD modelName, modelType, modelInfo, loaded, stored, published
RETURN modelName, modelType, modelInfo, loaded, stored, published
  ORDER BY modelName

Table 5. Results
modelName	modelType	modelInfo	loaded	stored	published
"my-model1_public"	"example-model-type"	{exampleModelInfo="exampleValue"}	true	false	true
"my-model2"	"example-model-type"	{number=42}	true	false	false

The published model is now accessible to all users.