Choosing compute resources for running tools in projects
You use compute resources in projects when you run jobs and most tools. Depending on the tool, you might have a choice of compute resources for the runtime for the tool.
Environments
You run assets, create jobs, and launch IDEs like RStudio or JupyterLab in a runtime. The runtime details are specified by an environment template.
Environment templates specify the hardware and software configuration of the runtimes:
- The hardware configuration specifies the amount of processing power and available RAM.
- The software configuration specifies the programming languages, a set of preinstalled libraries, and optional libraries or packages that you can specify.
Included environment templates
You can use the included environment templates to quickly get started, without having to create your own environment templates. The included environment templates are listed on the project's Environments page.
The included environments for notebooks and JupyterLab are added as an affiliate of a runtime release and prefixed with Runtime followed by the release year and release version.
A runtime release specifies a list of key data science libraries and a language version, for example Python 3.11. All environments of a runtime release are built based on the library versions that are defined in the release. This ensures the consistent use of data science libraries across all data science applications.
Runtime releases
IBM udates the library versions of supported runtimes as needed, to address security requirements. The updates change only the <Patch> version of the libraries, not the <Major>.<Minor> versions. Therefore,
your notebook assets will continue to run.
For example: A runtime release supports TensorFlow 2.12. In watsonx 2.2, the runtime release contains TensorFlow 2.14.1. Although TensorFlow might be updated to version 2.14.2 or 2.14.3 in later watsonx 2.2.x releases, it will not be updated to version 2.15.
Library packages that are included in Runtimes
For specific versions of popular data science library packages that are included in Watson Studio runtimes refer to these tables:
| Library | Runtime 25.1 on Python 3.12 |
|---|---|
| Keras | 3.9 |
| Lale | 0.9 |
| LightGBM | 4.5 |
| NumPy | 2.0 |
| ONNX | 1.17 |
| ONNX Runtime | 1.21 |
| OpenCV | 4.10 |
| pandas | 2.2 |
| PyArrow |
|
| PyTorch | 2.6 |
| scikit-learn | 1.6 |
| SciPy | 1.15 |
| SnapML | 1.16 |
| TensorFlow | 2.18 |
| XGBoost | 2.1 |
| Library | Runtime 25.1 on R 4.4 |
|---|---|
| arrow |
|
| car |
|
| caret | 7.0 |
| catools |
|
| forecast | 9.0 |
| ggplot2 | 4.0 |
| glmnet |
|
| hmisc | 5.2 |
| keras | 2.15 |
| lme4 |
|
| mvtnorm | 1.3 |
| pandoc |
|
| psych |
|
| python | 3.12 |
| randomforest |
|
| reticulate | 1.44 |
| sandwich | 3.1 |
| scikit-learn | 1.6 |
| spatial | 7.3 |
| tensorflow | 2.18 |
| tidyr | 1.3 |
| xgboost | 1.7 |
In addition to the libraries that are listed in the tables, runtimes include many other useful libraries. To see the list, select the Manage tab in your project, then click Templates, select the Environments tab, and then click on one of the listed environment templates.
Getting started
Python with GPU and Execution Engine for Apache Hadoop environments are not available by default.
- For Python with GPU environments, the Jupyter Notebooks with Python for GPU service must be installed.
- For Execution Engine for Apache Hadoop environments, the Execution Engine for Apache Hadoop service must be installed.
After these services are installed, you must create your own environment templates to use these environments.
The following table lists environment templates by asset type.
R-based runtimes for notebook editor without Analytics Engine installed do not work on the IBM Power® (ppc64le) platform. R-based runtimes for Data Refinery that are based on R4.3 and above work on the IBM Power® (ppc64le) platform.
| Asset | Programming language | Tool | Environment template type | Available environment templates/compute resources |
|---|---|---|---|---|
| Jupyter notebook | Python | notebook editor | Python virtual environment | Python environments |
| Jupyter notebook | Python | notebook editor | Python virtual environment with GPU | GPU environments |
| Jupyter notebook | Python | notebook editor | Spark | Spark environments |
| Jupyter notebook | Python | notebook editor | Spark | Hadoop cluster |
| Jupyter notebook | R | notebook editor | R environment | R environments |
| Jupyter notebook | R | notebook editor | Spark | Spark environments |
| Jupyter notebook | Python | JupyterLab | Python virtual environment | JupyterLab environments |
| Jupyter notebook | Python | Visual Studio Code editor | Python virtual environment | JupyterLab environments |
| Script | R | RStudio | R environment | RStudio environments |
| Shiny app | R | RStudio | R environment | RStudio environments |
| SPSS Modeler flow | N/A | SPSS Modeler | SPSS Modeler | SPSS Modeler environments |
| Data Refinery flow | R | Data Refinery | Spark | Data Refinery environments |
| Data Refinery flow | R | Data Refinery | Spark | Hadoop cluster |
Changing the environment runtime for a tool
For tools that support multiple runtime environments, you can select a larger environment runtime if you notice that processing is slow.
To change an environment runtime:
- Save any data from your current session before switching to another environment.
- Stop the active runtime under Tool runtimes on the Environments page on the Manage tab of your project.
- Restart the tool and select another environment with the compute power and memory capacity that better meets your requirements.