Amazon Forecast uses machine learning (ML) to generate more accurate demand forecasts, without requiring any prior ML experience. Forecast brings the same technology used at Amazon.com to developers as a fully managed service, removing the need to manage resources or rebuild your systems.
To start generating forecasts through Forecast, you can follow three steps of importing your data, training and evaluating a predictor, and then generating forecasts. Starting today, you can now stop an in-progress Forecast resource workflow if you have mistakenly started a job or misconfigured a workflow before starting, giving you more flexibility to manage your Forecast workflows and to experiment.
Previously, because you couldn’t stop APIs in progress, you had to wait for the job to complete and would incur charges for the job. You can now easily stop the following Forecast resource workflows:
In this post, we walk through the steps to stop workflows on the Forecast console. To review the steps through the APIs, refer to the following notebook in our GitHub repo.
Stop a resource job that is importing your datasets
You have two options to stop importing a dataset. One method is via the dataset details page on the Forecast console. On the Datasets page, choose your dataset and in the Dataset imports section, select the import job that you want to stop and choose Stop.
You can also stop a data import job from the job details page. In the Data imports section of your dataset, choose the import job to go to its details page. Then choose Stop.
Stop a resource job that is training a predictor
You have two options to stop a resource job that is training your predictor. One method is on the Predictors page for your dataset group, where you can select a predictor and choose Stop.
Alternatively, you can select the predictor and choose View details. Here you can stop the resource job that is training the predictor by choosing Stop.
Stop a resource job that is exporting backtest forecasts
Backtest forecasts are the forecasted values from the Forecast internal testing method of splitting the data into training and backtest data groups to compare forecasts versus observed data. When training a model, Forecast automatically splits the historical demand datasets into training and backtesting dataset groups. Forecast trains a model on the training dataset and forecasts at different specified stocking levels for the backtesting period, comparing to the observed values in the backtesting dataset group.
To stop a resource that is exporting these backtest results, select a predictor on the Predictors page of your dataset group. In the Predictor backtest exports section, select an export and choose Stop.
Stop a resource job that is generating forecasts
You have two options to stop a resource job that is generating your forecasts. One method is on the Forecasts page of the dataset group, where you can select a forecast and choose Stop.
Alternatively, you can select a forecast and choose View details. You can then stop the resource job that is generating the forecast by choosing Stop.
Stop a resource job that is exporting forecasts
Lastly, you can stop a resource job that is exporting your forecasts. You have two options to do so. One option is to select the forecast export job listed in the Forecast details section and choose Stop.
The second option is to choose the export job to view its details, and then choose Stop.
Stopping a resource halts the resource job workflow but doesn’t delete the resource. All your resources are still retained, and you can continue to call the Describe operation or access them as part of List APIs. After a resource is marked for stopping, it doesn’t count towards your Max Parallel In Progress limits. If you’re already at the limit, it allows you to submit a new job.
We allow only three resources of a given resource type at any time to be in the Stopping state, and you have to wait for one of the resources to go into the STOPPED state before you can stop more resources. After you initiate a stop, you can’t cancel it. You also can’t resume a stopped job. When you stop a predictor training or forecast generation job, you’re billed for the resources used up to the point when the job stopped.
You now have more flexibility in managing your Forecast workflows with the ability to stop in-progress resource workflows that may have been started unintentionally. To get started with this capability, see Stopping Resources and go through the notebook in our GitHub repo that walks you through how to use the Forecast Stop Resource APIs. You can use this capability in all Regions where Forecast is publicly available. For more information about Region availability, see AWS Regional Services.
About the Authors
Namita Das is a Sr. Product Manager for Amazon Forecast. Her current focus is to democratize machine learning by building no-code and low-code ML services. Outside of AWS, she frequently advises startups and is raising a puppy named Imli.
Gunjan Garg is a Sr. Software Development Engineer in the AWS Vertical AI team. In her current role at Amazon Forecast, she focuses on engineering problems and enjoys building scalable systems that provide the most value to end users. In her free time, she enjoys playing Sudoku and Minesweeper.
Punit Jain works as SDE on the Amazon Forecast team. His current work includes building large-scale distributed systems to solve complex machine learning problems with high availability and low latency as a major focus. In his spare time, he enjoys hiking and cycling.
Shannon Killingsworth is a UX Designer for Amazon Forecast and Amazon Personalize. His current work is creating console experiences that are usable by anyone, and integrating new features into the console experience. In his spare time, he is a fitness and automobile enthusiast.