AI Runtime examples

AI Runtime provides serverless GPU compute for inference, training, and fine-tuning AI and deep learning models. The pages below group notebook tutorials and Databricks CLI recipes by task: classic ML, recommendation systems, computer vision, post-training open-source LLMs, batch inference, and multi-GPU distributed training.

Notebook examples are included in the documentation. Databricks CLI examples link to the AI Runtime Cookbook on GitHub, where you can clone the repository and run its workload configurations and source code.

• • Classic machine learning
  • Examples for traditional machine learning tasks including XGBoost training, zero-shot tabular prediction, and time series forecasting.
• • Recommendation systems
  • Examples for building recommendation systems using modern deep learning approaches like two-tower models.
• • Computer vision
  • Examples for computer vision tasks including object detection and image classification.
• • Post-training OSS models (LLMs)
  • Examples for fine-tuning and post-training open-source large language models, including parameter-efficient methods.
• • Batch inference
  • Examples for large-scale batch inference with Ray Data and vLLM across multiple GPUs.
• • Multi-GPU distributed training
  • Examples for scaling training across multiple GPUs and nodes using the Serverless GPU API.