Feature Guide

Compared to NeMo 1.0, NeMo 2.0 provides a more modular and Pythonic approach to configuring experiments. The following guide provides an overview of some of the noteworthy features provided in NeMo 2.0.

• The Link Between Lightning and Megatron Core
• Logging and Checkpointing
• Serialization
• Parameter Efficient Fine-Tuning (PEFT)
• Hugging Face Integration
• Profiling
• Introduction
• Basic Usage
• Pre-training Models
• Fine-tuning Models
• Creating and Running Custom Recipes
• Text Generation
• Advanced Features
• Integration with NeMo-Run
• Learning More About the CLI
• Summary