Databricks Mosaic AI Training Documentation#

Databricks Mosaic AI Training is designed to tackle the challenges of training large AI models.

Getting Started#

Check out our Getting Started tutorial to start training AI models.

Reference Documentation#

Mosaic AI CLI and SDK

Command Line Interface (CLI) and Python SDK

Composer

Open source training framework for large scale Generative AI models

Streaming

Open source library for fast, accurate data streaming from cloud storage

LLMFoundry

Scripts for training, finetuning, and evaluating LLMs with Composer and MCLI

Megablocks

MegaBlocks is a light-weight library for mixture-of-experts (MoE) training. The core of the system is efficient “dropless-MoE” and standard MoE layers.

Additional Resources#