ML Data Processing to Model Training

ML Data Processing to Model Training

Process graph data, then train a model on the processed output.

1. Create a data processing job that exports and prepares graph data into the processed-data S3 location.
2. Poll the data processing job. Retry while it is still in progress and continue once its status is Completed.
3. Launch a model training job that consumes the completed data processing job's output.
4. Poll the model training job. Retry while it is still in progress and finish once its status is Completed.

What you'll need to have ready

• inputDataS3Location: S3 URI for the input graph data.
• processedDataS3Location: S3 URI where processed output is written.
• trainModelS3Location: S3 location for trained model artifacts.
• modelName: The model type to train (rgcn, transe, distmult, rotate, custom).

For your developer

Arazzo workflow amazon-neptune:ml-dataprocessing-to-training, 4 operations: createDataProcessingJob → getDataProcessingJobStatus → createModelTrainingJob → getModelTrainingJobStatus. Specs on APIs.io.