airflow-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From GitBox <>
Subject [GitHub] troychen728 commented on a change in pull request #3658: [AIRFLOW-2524] Add Amazon SageMaker Training
Date Fri, 03 Aug 2018 17:30:44 GMT
troychen728 commented on a change in pull request #3658: [AIRFLOW-2524] Add Amazon SageMaker

 File path: airflow/contrib/operators/
 @@ -0,0 +1,98 @@
+# -*- coding: utf-8 -*-
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+from airflow.contrib.hooks.sagemaker_hook import SageMakerHook
+from airflow.models import BaseOperator
+from airflow.utils import apply_defaults
+from airflow.exceptions import AirflowException
+class SageMakerCreateTrainingJobOperator(BaseOperator):
+    """
+       Initiate a SageMaker training
+       This operator returns The ARN of the model created in Amazon SageMaker
+       :param training_job_config:
+       The configuration necessary to start a training job (templated)
+       :type training_job_config: dict
+       :param region_name: The AWS region_name
+       :type region_name: string
+       :param sagemaker_conn_id: The SageMaker connection ID to use.
+       :type aws_conn_id: string
 Review comment:
   Thank you very much for your reply. I agree with you that multiple jobs can still run in
parallel, but to my understanding, there's also an option to run in sequential.  More importantly,
I think one thing that is very essential is that, what if there's already a training job running
(whether initiated from Airflow or other means) before a dag is scheduled to run? Since a
training job can takes days to finish, I think it might be a potential use case for Airflow
users. As a result, a DAG can start with a sensor, and have other nodes dependent on the sensor.
It would not be possible to accomplish this if operator, and sensor are coupled together.

This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:

With regards,
Apache Git Services

View raw message