Skip to main content

Explore Problems

Certifications Quiz Pricing

⌘K

Fine-Tuning & Model Customization

Course Contents

Prompting vs RAG vs Fine-Tuning — The Decision

Good Reasons to Fine-Tune

Bad Reasons to Fine-Tune

Fine-Tuning + RAG Together

Estimating Cost & Effort

How Training Actually Works (Backprop Recap)

The Memory Bottleneck & Memory Math

Numerical Precision & Quantization (GGUF/llama.cpp, GPTQ, AWQ)

Full Fine-Tuning vs PEFT

LoRA & QLoRA Explained

Fine-Tuning Hyperparameters & Tactics

Data Quality, Coverage & Quantity

Sourcing & Annotating Data

Synthetic Data Generation

Model Distillation

Cleaning, Dedup & Formatting

Hands-On: Build a Fine-Tuning Dataset

The Post-Training Pipeline: SFT → DPO → RL

Direct Preference Optimization (DPO)

Reasoning Fine-Tuning with GRPO

Model Merging (SLERP, TIES, DARE)

Evaluating a Fine-Tuned Model

Hands-On: Fine-Tune a Small Model with LoRA

Reasoning Fine-Tuning with GRPO

Teach a model to reason with GRPO — sampling a group and rewarding what works, the group as its own baseline, verifiable rewards, reward design and the aha moment, failure modes, and GRPO with TRL.

Loading...

Loading content...