LoRA Fine-Tuning & Deployment of Nemotron 3 Nano for Text2SQL

End-to-end guide for LoRA fine-tuning NVIDIA Nemotron-3-Nano-30B on a text-to-SQL task (BIRD SQL) and deploying with NVIDIA NIM or vLLM. Two notebook options:

• AutoModel – Uses the NeMo AutoModel container for training and checkpointing.

• Megatron Bridge – Uses Megatron Bridge for training.

What’s Inside

File	Description
finetuning_deployment_guide_automodel.ipynb	Full walkthrough (NeMo AutoModel)
finetuning_deployment_guide_mbridge.ipynb	Full walkthrough (Megatron Bridge)
bird_sql/	Dataset preparation helpers (BIRD SQL → chat-templated JSONL)

Quick Start

Open either notebook to get started. A Deploy on Brev one-click option is also available for the Megatron Bridge example: