Use a trained model to generate new model artifacts - Amazon Neptune

Use a trained model to generate new model artifacts

Using the Neptune ML model transform command, you can compute model artifacts like node embeddings on processed graph data using pre-trained model parameters.

Model transform for incremental inference

In the incremental model inference workflow, after you have processed the updated graph data that you exported from Neptune you can start a model transform job using a curl (or awscurl) command like the following:

curl \ -X POST https://(your Neptune endpoint)/ml/modeltransform -H 'Content-Type: application/json' \ -d '{ "id" : "(a unique model-training job ID)", "dataProcessingJobId" : "(the data-processing job-id of a completed job)", "mlModelTrainingJobId": "(the ML model training job-id)", "modelTransformOutputS3Location" : "s3://(your Amazon S3 bucket)/neptune-model-transform/" }'

You can then pass the ID of this job to the create-endpoints API call to create a new endpoint or update an existing one with the new model artifacts generated by this job. This allows the new or updated endpoint to provide model predictions for the updated graph data.

Model transform for any training job

You can also supply a trainingJobName parameter to generate model artifacts for any of the SageMaker training jobs launched during Neptune ML model training. Since a Neptune ML model training job can potentially launch many SageMaker training jobs, this gives you the flexibility to create an inference endpoint based on any of those SageMaker training jobs.

For example:

curl \ -X POST https://(your Neptune endpoint)/ml/modeltransform -H 'Content-Type: application/json' \ -d '{ "id" : "(a unique model-training job ID)", "trainingJobName" : "(name a completed SageMaker training job)", "modelTransformOutputS3Location" : "s3://(your Amazon S3 bucket)/neptune-model-transform/" }'
Note

The modeltransform command always runs the model transform on the best SageMaker training job for that training.

See The modeltransform command for more information about model transform jobs.