Download Scoring Code from a deployment¶
The behavior of deployments from which you download Scoring Code depends on the MLOps configuration for your organization.
You can download Scoring Code for models as pre-compiled JAR files (with all dependencies included) to be used outside of the DataRobot platform. This topic describes how to download Scoring Code from a deployment. Alternatively, you can download it from the Leaderboard.
For Scoring Code-enabled models deployed to an external prediction server, you can download Scoring Code from a deployment's Actions menu in the Deployments inventory or from a deployment's Predictions > Portable Predictions tab. For Scoring Code-enabled models deployed to a DataRobot prediction environment, you can only download Scoring Code from the Deployments inventory.
Navigate to the Deployments inventory, and then take either of the following steps:
Open a deployment, then navigate to the Predictions > Portable Predictions tab and click Scoring Code.
The Portable Predictions tab is only available for models deployed to an external environment. If you deployed your model to a DataRobot environment, use the Deployments inventory Scoring Code download method.
Open the deployment's Actions menu and click Get Scoring Code.
The Download Scoring Code dialog opens.
Complete the fields described below in the Portable Predictions tab (or the Download Scoring Code dialog).
Element Description Scoring Code Provides a Java package containing your DataRobot model. Under Portable Prediction Method, select Scoring Code. You can alternatively select Portable Prediction Server to set up a REST API-based prediction server. Include Monitoring Agent Downloads the MLOps Agent with your Scoring Code. Include Prediction Explanations Includes code to calculate Prediction Explanations with your Scoring Code. This allows you to get Prediction Explanations from your Scoring Code by adding the command line option:
--with-explanations. See Scoring at the command line for more information.
Include Prediction Intervals (for time series) Includes code to calculate Prediction Intervals with your Scoring Code. This allows you to get Prediction Intervals (from 1 to 99) from your Scoring Code by adding the command line option:
--interval_length=<integer value from 1 to 99>. See Scoring at the command line for more information.
Prepare and download / Prepare and download as source code
- Prepare and download: Downloads the Scoring Code as a Java package. The package contains compiled Java executables, which include all dependencies and can be used to make predictions.
- Prepare and download as source code: Downloads Java source code files. These are a non-obfuscated version of the model; they cannot be used to score the model since they are not compiled and dependency packages are not included. Use the source files to explore the model’s decision-making process. This option is only available if you don't have the monitoring agent and prediction explanations enabled.
Example Provides a code example that calls the Scoring Code from the command line. Copy to clipboard Copies the Scoring Code example to your clipboard so that you can paste it on the command line.
Once the settings are configured, click Prepare and download to download a Java package or Prepare and download as source code to download source code files.
For users not on the 5.0 pricing plan who choose to download Scoring Code, the deployment becomes permanent and cannot be deleted. A warning message prompts you to accept this condition. Use the toggle to indicate your understanding, then click Prepare and download to download a Java package or Prepare and download as source code to download source code files.
When the Scoring Code download completes, use the monitoring snippet to instrument the Scoring Code.
For implementation examples, reference the MLOps agent tarball documentation, which you can download from the Developer Tools page.