r_lang

R Drop-In Template Environment

This template environment can be used to create artifact-only R custom models that use the caret library. Your custom model directory needs only contain your model artifacts if you use the environment correctly.

Supported Libraries

This environment has built for R with support for the caret library of models.

Other dependencies

This environment uses cmrun to run custom model. cmrun uses rpy2 package (by default the latest version is installed) to run R. You may need to adjust rpy2 and pandas versions for compatibility.

Supported Model Types

Due to the time required to install all libraries recommended by caret, only model types that are also package names are installed (ex. brnn, glmnet). If you would like to use models that require additional packages, you will need to make a copy of this environment and modify the Dockerfile to install additional packages. To speed up the build time, you can remove the lines in the # Install caret models section and install only what you need.

To check if your model's method matches its package name, please refer to the official docs

Instructions

1. From the terminal, run tar -czvf r_dropin.tar.gz -C /path/to/public_dropin_enironments/r_lang/ .
2. Using either the API or from the UI create a new Custom Environment with the tarball created in step 1.

NOTE This environment may take more than an hour to build.

Creating models for this environment

To use this environment, your custom model archive must contain a serialized model artifact as an RDS file with the file extension .rds (ex. trained_glm.rds), as well as any other custom code and artifacts needed to use your serialized model.

This environment makes the following assumption about your serialized model:

• The data sent to custom model can be used to make predictions without additional pre-processing
• No additional libraries need to be loaded in order to properly use your model.
• Regression models return a single floating point per row of prediction data
• Binary classification models return two floating point values that sum to 1.0 per row of prediction data
  • The first value is the positive class probability, the second is the negative class probability
• There is a single rds file present

If these assumptions are incorrect for your model, you should make a copy of custom.R, modify it as needed, and include it in your custom model archive.

The structure of your custom model archive should look like:

• custom_model.tar.gz
  • artifact.rds
  • custom.R (if needed)

Please read datarobot-cmrunner documentation on how to assemble custom.R.