How-toUpdated 2026-06-27 · 16 min read · by RTXsparks Lab

Fine-tuning LLMs on RTX Spark: LoRA, QLoRA & Full-Rank

How to fine-tune 8B–70B models on RTX Spark using LoRA, QLoRA, and NVIDIA NeMo. Includes recipes, VRAM budgets, and reproducible configs.

Recipes

Axolotl configs provided.

Frequently asked questions

Multi-node fine-tuning?

Yes with NeMo + PP=2 on a 2-node Spark cluster.

Related guides