Machine Learning Associate ยท MLOps, ML runtimes and AutoML
Databricks Runtime for Machine Learning
Advantages of the Databricks Runtime for Machine Learning over the standard runtime: preinstalled libraries, MLflow and GPU support, compatible versions.
Key points
- A runtime built on top of the standard Databricks Runtime that ships with common ML and deep learning libraries already installed (for example MLflow, scikit-learn, XGBoost, LightGBM, Hyperopt, PyTorch, TensorFlow).
- Versions are chosen to work together, and GPU-enabled clusters come with the CUDA libraries.
- You select it by ticking Machine learning when creating compute.
Source: https://docs.databricks.com/aws/en/machine-learning/databricks-runtime-ml
Practice questions
What is the main advantage of Databricks Runtime ML over the standard Databricks Runtime?
- Common ML libraries come preinstalled with compatible versions
- It runs without any cluster
- It replaces Apache Spark with a different engine
- It trains models at no compute cost
Show answer
Answer: A. Common ML libraries come preinstalled with compatible versions
The ML runtime adds preinstalled, version-compatible ML and deep learning libraries on top of the standard runtime. It does not replace Spark, remove the need for clusters, or make training free.
Which libraries come preinstalled in a Databricks Runtime ML release (for example 16.4 LTS ML)?
- MLflow
- Hyperopt
- XGBoost
- A Kubernetes job scheduler
- A hosted Weights & Biases server
Show answer
Answer: A. MLflow, B. Hyperopt, C. XGBoost
MLflow, Hyperopt and XGBoost are included in the ML runtime. Weights & Biases dashboards and a Kubernetes scheduler are not part of it.
True or false: Databricks Runtime ML is a completely separate product that does not include the standard Databricks Runtime components.
Show answer
Answer: False
False. Runtime ML is built on top of the standard Databricks Runtime and adds ML libraries to it.
Practice 8 questions on this topic in Bitzu
Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.
