Goal path for ml-curious developers

Train and Fine-Tune Models

The maths intuition, neural networks and transformers, then datasets, LoRA fine-tuning, preference tuning and shipping your model.

Steps
9
Lessons
45
Time
10 h 50 min
Done
0 / 45

Already know some of this? Start at step 1 (beginner), step 4 (intermediate), step 6 (advanced).

  1. Maths Intuition

    Track 11: ML and Deep Learning Foundations for AI Engineers, 4 lessons

    1. 1. Vectors, Similarity and Probability, Visually Beginner, 11 min
      Coming soon
    2. 2. Matrices: What a Layer Really Does Beginner, 11 min
      Coming soon
    3. 3. Gradients, Loss Functions and Gradient Descent Intermediate, 12 min
      Coming soon
    4. 4. Optimisers: SGD, Adam, Learning-Rate Schedules Advanced, 11 min
      Coming soon

  2. Neural Networks

    Track 11: ML and Deep Learning Foundations for AI Engineers, 5 lessons

    1. 5. From One Neuron to a Network Beginner, 11 min
      Coming soon
    2. 6. Backpropagation, Step by Step Intermediate, 13 min
      Coming soon
    3. 7. PyTorch Basics Intermediate, 13 min
      Coming soon
    4. 8. Project: Digit Recogniser Intermediate, Project, 45 min
      Coming soon
    5. 9. Before and Around Transformers: CNNs, RNNs, Regularisation Advanced, 11 min
      Coming soon

  3. Transformers and Attention

    Track 11: ML and Deep Learning Foundations for AI Engineers, 7 lessons

    1. 10. Embeddings: Meaning as Numbers Beginner, 11 min
      Coming soon
    2. 11. The Transformer in One Picture Beginner, 12 min
      Coming soon
    3. 12. Self-Attention, Multi-Head and Positions, Step by Step Intermediate, 13 min
      Coming soon
    4. 13. Code a Transformer Block Advanced, 15 min
      Coming soon
    5. 14. KV Cache and Why Long Context Is Expensive Advanced, 11 min
      Coming soon
    6. 15. Mixture of Experts, Beyond Transformers, and Looking Inside Models Advanced, 11 min
      Coming soon
    7. 16. Interpretability Basics: Look Inside a Model Advanced, 18 min
      Coming soon

  4. How Training Works

    Track 10: Training LLMs, 5 lessons

    1. 17. Fine-Tune vs Prompt vs RAG Beginner, 15 min
      Coming soon
    2. 18. How LLM Training Works: Tokens, Loss, Pre-training vs Fine-tuning Beginner, 18 min
      Coming soon
    3. 19. GPUs for Training: Colab, Kaggle, Cloud GPUs Safely Intermediate, 12 min
      Coming soon
    4. 20. Pre-training at Scale, Explained Intermediate, 13 min
      Coming soon
    5. 21. Scaling Laws and Compute-Optimal Training Advanced, 12 min
      Coming soon

  5. Training Data

    Track 10: Training LLMs, 4 lessons

    1. 22. Chat and Instruction Data Formats Beginner, 14 min
      Coming soon
    2. 23. Build a 300-1,000 Example Dataset: Clean, Split, Licence-Check Intermediate, 22 min
      Coming soon
    3. 24. Synthetic Data With Guardrails Intermediate, 13 min
      Coming soon
    4. 25. Deduplication and Decontamination Advanced, 12 min
      Coming soon

  6. Fine-Tuning Methods

    Track 10: Training LLMs, 5 lessons

    1. 26. Full Fine-Tuning vs Parameter-Efficient (PEFT) Beginner, 12 min
      Coming soon
    2. 27. LoRA Explained and Applied Intermediate, 13 min
      Coming soon
    3. 28. QLoRA: Bigger Models on Small GPUs Intermediate, 12 min
      Coming soon
    4. 29. Your First Fine-Tune on Colab (Instruction Tuning) Intermediate, 25 min
      Coming soon
    5. 30. Make It Better: Data, Hyperparameters, Continued Pre-training, Multi-GPU Advanced, 18 min
      Coming soon

  7. Preference Tuning and Alignment

    Track 10: Training LLMs, 3 lessons

    1. 31. RLHF and Preference Tuning, Explained Beginner, 12 min
      Coming soon
    2. 32. DPO Hands-On (and ORPO, KTO, Reward Models) Intermediate, 14 min
      Coming soon
    3. 33. GRPO and RL for Reasoning Advanced, 14 min
      Coming soon

  8. Training Tools (One Lesson Each)

    Track 10: Training LLMs, 5 lessons

    1. 34. Hugging Face Transformers for Training Beginner, 12 min
      Coming soon
    2. 35. Hugging Face TRL and PEFT Intermediate, 12 min
      Coming soon
    3. 36. Unsloth (and LLaMA-Factory) Intermediate, 12 min
      Coming soon
    4. 37. Axolotl Advanced, 12 min
      Coming soon
    5. 38. torchtune Advanced, 12 min
      Coming soon

  9. Evaluate, Compress and Ship

    Track 10: Training LLMs, 7 lessons

    1. 39. Benchmarks Explained, and lm-evaluation-harness Beginner, 11 min
      Coming soon
    2. 40. Evaluate Your Fine-Tune Properly Intermediate, 22 min
      Coming soon
    3. 41. Quantization: GGUF, GPTQ, AWQ, bitsandbytes Intermediate, 13 min
      Coming soon
    4. 42. Merge, Export, Run Your Model (and Model Merging) Intermediate, 18 min
      Coming soon
    5. 43. Distillation: Teach a Small Model From a Big One Advanced, 13 min
      Coming soon
    6. 44. Capstone: Fine-Tune for a Real Work Task Advanced, Project, 15 min
      Coming soon
    7. 45. Model Merging With mergekit: SLERP, TIES, DARE Advanced, 18 min
      Coming soon

Ticks and “you are here” are saved only in this browser.

Other paths

Changing goals is fine. Lessons you finish count on every path that includes them.