Skip to main content
Free AI courses. Free certificates of completion.Learn how
NeuroLearn AIFREE AI LEARNING
Start learning
Back to All Courses
AdvancedFine-Tuning & Model Customisation Free Verifiable Certificate

Fine-Tuning Open LLMs: LoRA, QLoRA and Deployment

Decide when to fine-tune, build clean training data, train efficiently with LoRA and QLoRA, evaluate for forgetting, and serve the result.

1h 5m of material
3 Modules (6 Lessons)
Free · no account needed to read
Access Fee
FREE100% Free Forever
Full Interactive Curriculum Access
In-Browser Prompt Studio & Code
Official Certification Exam
Verifiable PDF/PNG Certificate
Start Learning Now

The assessment unlocks once you have worked through the lessons.

Independent provider. Not an accredited awarding body. Not a regulated qualification.

What You Will Master

  • Choose between prompting, retrieval and fine-tuning by naming the failure and pricing each option at real volume
  • Build training data with the correct chat template, deduplication and a leakage-free split
  • Estimate memory for full fine-tuning, LoRA and QLoRA, and choose rank, alpha and target modules deliberately
  • Read training and validation loss together and keep the checkpoint that generalises best
  • Evaluate a fine-tuned model for target gains, retained capabilities, safety and catastrophic forgetting
  • Size GPU memory for weights and the KV cache and plan quantised, batched serving

Skills Acquired

LLM Fine-TuningLoRA & QLoRATraining Data CurationOverfitting & Early StoppingForgetting & Regression EvaluationModel Serving & Quantisation

Prerequisites

  • Comfortable writing Python and running scripts from a terminal
  • A working understanding of transformers; the free Deep Learning & Transformers course covers it
  • Familiarity with evaluation basics; the free LLMOps course is a useful companion

Curriculum Syllabus

3 modules · 6 lessons · 1h 5m · one assessment

Module 1

Module 1: Deciding and Preparing

Choose the right technique for the failure, then build training data worth imitating.

1.1When Fine-Tuning Is the Right Tool
11 min
1.2Building a Training Dataset That Teaches the Right Thing
11 min
Module 2

Module 2: Training Efficiently

Fine-tune on modest hardware with LoRA and QLoRA, and stop at the checkpoint that generalises.

2.1LoRA and QLoRA: Fine-Tuning on a Single GPU
11 min
2.2Running Training: Hyperparameters, Loss Curves and Overfitting
11 min
Module 3

Module 3: Evaluating and Serving

Prove gains without hidden regressions, then serve the model within memory and latency budgets.

3.1Evaluating a Fine-Tuned Model: Gains, Regressions and Forgetting
10 min
3.2Serving Fine-Tuned Models: Quantisation, Memory and Throughput
11 min
Course author

Md. Rajib Hawlader

Principal AI Research Scientist, NeuroLearn AI Labs

Writes and maintains the NeuroLearn AI curriculum.

Who am I

Ready to Begin Fine-Tuning Open LLMs: LoRA, QLoRA and Deployment?

Self-paced curriculum with free instant certificate issuance upon passing the evaluation.

Start First Lesson