1. Introduction
3. Building the Model
4. 1. Prerequisites
   1. 1.1. Software
   2. 1.2. The Training Data
   3. 1.3. The Pipeline
5. 2. The Tokenizer
   1. 2.1. Building the Vocabulary
   2. 2.2. Encoding and Decoding
   3. 2.3. Why Word-Level Tokens?
   4. 2.4. Complete Code
6. 3. The Autograd Engine
   1. 3.1. The Math You Need
   2. 3.2. The Value Class
   3. 3.3. Derived Operations
   4. 3.4. The Computation Graph
   5. 3.5. Backward Pass
   6. 3.6. Complete Code
7. 4. Neural Network Primitives
   1. 4.1. Linear
   2. 4.2. Softmax
   3. 4.3. RMSNorm
   4. 4.4. Complete Code
8. 5. The Model
   1. 5.1. Configuration
   2. 5.2. Parameters
   3. 5.3. Embeddings
   4. 5.4. Attention
   5. 5.5. MLP
   6. 5.6. Residual Connections
   7. 5.7. The KV Cache
   8. 5.8. Creating the Model
   9. 5.9. Running the Model
   10. 5.10. Complete Code
10. Training and Inference
11. 6. Training
    1. 6.1. The Training Configuration
    2. 6.2. The Training Loop
    3. 6.3. Watching It Learn
    4. 6.4. Complete Code
12. 7. Saving the Model
    1. 7.1. What Gets Saved
    2. 7.2. Loading the Model
13. 8. Generation
    1. 8.1. The Generation Loop
    2. 8.2. Sampling Strategies
    3. 8.3. The KV Cache
    4. 8.4. Example Output
    5. 8.5. Complete Code
15. Putting It to Work
16. 9. Smoke Test
    1. 9.1. Train the Model
    2. 9.2. Generate Sentences
    3. 9.3. What You Have Built
    4. 9.4. Complete Code
17. 10. Fine-Tuning
    1. 10.1. The Question Dataset
    2. 10.2. Run the Fine-Tuning
    3. 10.3. Generate Questions
    4. 10.4. Catastrophic Forgetting
    5. 10.5. Complete Code