Oreilly – Build a Large Language Model from Scratch (early access), Video Edition 2024-7

Description

Build a Large Language Model from Scratch course (early access), Video Edition. This video tutorial guides you to create, train, and configure a large language model (LLM) from scratch. In this informative book, bestselling author Sebastian Raschka guides you step-by-step through creating your LLM, explaining each step with clear text, diagrams, and examples. You’ll go from initial design and creation to pre-training on a general set and then fine-tuning for specific tasks. Large language models (LLMs) powering advanced AI tools like ChatGPT, Bard, and Copilot seem like a miracle, but they’re not magic. This book makes LLMs pointless by helping you build your own LLM from scratch. You will gain a unique and valuable insight into how LLMs work, learn how to assess their quality and pick up specific techniques to fine-tune and improve them.

What you will learn

• Designing and coding all parts of an LLM
• Preparation of a suitable dataset for LLM training
• Fine-tune LLMs for text classification with your own data
• Apply instruction fine-tuning techniques to ensure your LLM follows instructions
• Loading preset weights into an LLM

This course is suitable for people who

• Looking to learn how to build and train large language models.
• Interested in better understanding how LLMs work.
• They want to improve their skills in the field of machine learning and artificial intelligence.
• Looking for a practical guide to building your LLM.

Course specifications Build a Large Language Model from Scratch (early access), Video Edition

• Publisher:  Oreilly
• Lecturer: Sebastian Raschka
• Training level: beginner to advanced
• Training duration: 8 hours and 12 minutes

Course headings

• Chapter 1. Understanding Large Language Models
• Chapter 1. Applications of LLMs
• Chapter 1. Stages of building and using LLMs
• Chapter 1. Introducing the transformer architecture
• Chapter 1. Utilizing large datasets
• Chapter 1. A closer look at the GPT architecture
• Chapter 1. Building a large language model
• Chapter 1. Summary
• Chapter 2. Working with Text Data
• Chapter 2. Tokenizing text
• Chapter 2. Converting tokens into token IDs
• Chapter 2. Adding special context tokens
• Chapter 2. Byte pair encoding
• Chapter 2. Data sampling with a sliding window
• Chapter 2. Creating token embeddings
• Chapter 2. Encoding word positions
• Chapter 2. Summary
• Chapter 3. Coding Attention Mechanisms
• Chapter 3. Capturing data dependencies with attention mechanisms
• Chapter 3. Attending to different parts of the input with self-attention
• Chapter 3. Implementing self-attention with trainable weights
• Chapter 3. Hiding future words with causal attention
• Chapter 3. Extending single-head attention to multi-head attention
• Chapter 3. Summary
• Chapter 4. Implementing a GPT model from scratch to generate text
• Chapter 4. Normalizing activations with layer normalization
• Chapter 4. Implementing a feed forward network with GELU activations
• Chapter 4. Adding shortcut connections
• Chapter 4. Connecting attention and linear layers in a transformer block
• Chapter 4. Coding the GPT model
• Chapter 4. Generating text
• Chapter 4. Summary
• Chapter 5. Pretraining on Unlabeled Data
• Chapter 5. Training an LLM
• Chapter 5. Decoding strategies to control randomness
• Chapter 5. Loading and saving model weights in PyTorch
• Chapter 5. Loading pretrained weights from OpenAI
• Chapter 5. Summary
• Chapter 6. Finetuning for Classification
• Chapter 6. Preparing the dataset
• Chapter 6. Creating data loaders
• Chapter 6. Initializing a model with pretrained weights
• Chapter 6. Adding a classification head
• Chapter 6. Calculating the classification loss and accuracy
• Chapter 6. Finetuning the model on supervised data
• Chapter 6. Using the LLM as a spam classifier
• Chapter 6. Summary
• Chapter 7. Finetuning to Follow Instructions
• Chapter 7. Preparing a dataset for supervised instruction fine tuning
• Chapter 7. Organizing data into training batches
• Chapter 7. Creating data loaders for an instruction dataset
• Chapter 7. Loading a pretrained LLM
• Chapter 7. Finetuning the LLM on instruction data
• Chapter 7. Extracting and saving responses
• Chapter 7. Evaluating the fine-tuned LLM
• Chapter 7. Conclusions
• Chapter 7. Summary
• Appendix A. Introduction to PyTorch
• Appendix A. Understanding tensors
• Appendix A. Seeing models as computation graphs
• Appendix A. Automatic differentiation made easy
• Appendix A. Implementing multilayer neural networks
• Appendix A. Setting up efficient data loaders
• Appendix A. A typical training loop
• Appendix A. Saving and loading models
• Appendix A. Optimizing training performance with GPUs
• Appendix A. Summary
• Appendix A. Further reading
• Appendix A. Exercise answers
• Appendix D. Adding Bells and Whistles to the Training Loop
• Appendix D. Cosine decay
• Appendix D. Gradient clipping
• Appendix D. The modified training function
• Appendix E. Parameter-efficient finetuning with LoRA
• Appendix E. Preparing the dataset
• Appendix E. Initializing the model
• Appendix E. Parameter-efficient finetuning with LoRA

Course images

Sample video of the course

download link

Download part 1 – 1 GB

Download part 2 – 6.3 MB

File(s) password: www.downloadly.ir

File size

1 GB