Tracks Courses Pricing

© 2026 Capstok · learn by doing

Mission About Pricing MCP Contact Feedback Terms Privacy

Tracks Courses Pricing

machine-learning/Advanced AI Engineering

The SFT pipeline — dataset format, training config, evaluation

medium

Learn with your AI

Open this lesson in your favourite AI. It'll walk you through the why, explain the demo, and quiz you on the try-it list.

Open in Claude Open in ChatGPT

Module 1 · Task 3 of 10

Fine-Tuning Fundamentals

01LoRA explained — low-rank decomposition and why it works
02Full fine-tune vs LoRA vs QLoRA — memory, time, quality tradeoffs
03The SFT pipeline — dataset format, training config, evaluation
04DPO — direct preference optimization without a reward model
05Quantization — int8, int4, GPTQ, AWQ, bitsandbytes
06Serving fine-tuned models — vLLM, TGI, Ollama comparison
07Evaluating fine-tuned models — MT-Bench, Alpaca Eval, custom evals
08RLHF overview — reward model, PPO loop, why it's hard
09Project · Fine-tune and deploy a domain-adapted model
10Quiz me — Fine-Tuning Fundamentals

© 2026 Capstok · learn by doing

Mission About Pricing MCP Contact Feedback Terms Privacy