NVIDIA NeMo

End-to-end framework for building, training, and deploying AI models

View on GitHub

Overview

NVIDIA NeMo is a scalable framework built on top of Megatron-LM for training, fine-tuning, and deploying large AI models including LLMs, speech, and multi-modal models.

Key Features

  • Built on Megatron-LM core
  • Pre-training and fine-tuning workflows
  • PEFT methods (LoRA, P-Tuning, etc.)
  • NeMo Launcher for easy deployment