Introduction to large language models (Lesson 01) #33
+331
−3
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Collaborator
Author
|
Converted back to draft, this is way too wordy so I'm going to trim it down a bit before subjecting someone to reviewing this. 😄 |
Collaborator
roshansuresh
left a comment
roshansuresh
left a comment
There was a problem hiding this comment.
Overall the intro lesson reads really well! The intro table of contents shouldn't include Abstraction layers there I think. I've left some small comments in the chapters.
| - **1. Natural language processing**: an overview of the general field of natural language processing | ||
| - **2. Introduction to LLMs**: what are large language models, and how do they work? | ||
| - **3. LLM architecture**: a deeper dive into what makes LLMs tick. | ||
| - **4. Demo: Visualizing embeddings** Visualizing how meanings are represented in LLMs. |
Collaborator
There was a problem hiding this comment.
For clarity, I would have it say "Demo - Visualizing embeddings:"
|
|
||
| Ultimately, NLP aims to make human-computer interaction as intuitive as human-to-human exchanges, so it can be used in fields as diverse as healthcare diagnostics, explaining complex legal documents, and personalized education. | ||
|
|
||
| ### NLP Methods |
Collaborator
There was a problem hiding this comment.
Perhaps we can call this "NLP Methods - A History" or something?
|
|
||
| There are really *two* different learning modes for LLMs. First, by training on huge bodies of text in the next-word-prediction task, we end up with what are often called *foundational*, *pretrained*, or *base* models. These are general purpose models that embody information from extremely broad sources. | ||
|
|
||
| However, these foundational models don't work well in special-purpose jobs like personal assistants, chatbots, etc. A *second* second training step fine-tunes these models on smaller, labeled datasets for specific applications. |
Collaborator
There was a problem hiding this comment.
This is actually a really important point. I think we should reiterate this in the chatbots lesson
Collaborator
There was a problem hiding this comment.
Also, a typo in second second
|
|
||
| If you have ever been using ChatGPT and it has asked you to rank two responses, this is OpenAI collecting data for future rounds of RLHF. | ||
|
|
||
| The result with fine-tuning is the production of specialized models built on top of the same foundation. While in this course we will not go through the process of building your own LLM, the excellent book [Build a Large Language Model from Scratch](https://www.manning.com/books/build-a-large-language-model-from-scratch) by Sebastian Raschka, walks you through this in detailk using PyTorch if you are interested. The above picture is adapted from his book. |
|
|
||
| The result with fine-tuning is the production of specialized models built on top of the same foundation. While in this course we will not go through the process of building your own LLM, the excellent book [Build a Large Language Model from Scratch](https://www.manning.com/books/build-a-large-language-model-from-scratch) by Sebastian Raschka, walks you through this in detailk using PyTorch if you are interested. The above picture is adapted from his book. | ||
|
|
||
| In the next section we will dig into the details about how LLMS actually work. As we said, it isn't just that they are *large*, but their *architecture*, that makes them so powerful. |
Collaborator
There was a problem hiding this comment.
typo in "LLMS" (maybe fine but its probably better to have it be LLMs)
| For a quick overview of LLM function, [check out this video](https://www.youtube.com/watch?v=5sLYAQS9sWQ). | ||
|
|
||
| ## 4. Demo: Visualizing embeddings | ||
| In the following demonstration we will visualize text embeddings based on their semantic similarity. Instead of similareity of single words like `apple` and `phone`, we will look at similarities among entire *sentences*. |
Collaborator
There was a problem hiding this comment.
typo in "instead of similareity"
| ``` | ||
| We first need to load our OpenAI API key so you can use the embedding model that is part of their suite of models. We will discuss this more in the lesson on chat completions | ||
|
|
||
| > TODO: handle this discussion better -- put in README for this week and discuss handling of api keys, where to put it, that it will recursively seach parent dsirectories up to root). Please be sure not to share your API key! |
Collaborator
There was a problem hiding this comment.
You may want to remove this
| plt.show() | ||
| ``` | ||
|
|
||
| This mapping shows that similar movies tend to cluster together in the embedding space. E.g., Christmas movies in one region of the space, romcons in another (note Love Actually is a Christmas Romcom). Feel free to drop new summaries in to see where they fall in this movie map. |
Collaborator
There was a problem hiding this comment.
Do we show the plot here or just ask them to run this code themselves?
|
|
||
| > To fill in later: brief motivational preview here. Briefly explain why this lesson matters, what students will be able to do by the end, and what topics will be covered. Keep it tight and motivating. | ||
| > For an introduction to the course, and a discussion of how to set up your environment, please see the [Welcome](../README.md) page. |
Collaborator
There was a problem hiding this comment.
We can remove the prior statement
| 2. [OpenAI Chat Completions API](02_open_ai_api.md) | ||
| Intro and overview of openai api chat completions endpoint. Go over required params (messages/model), but also the important optional params (max_tokens, temperature, top_p etc). Mention responses endpoint (more friendly to tools/agents). Discuss and demonstrate use of moderations endpoint. | ||
|
|
||
| 3. [Abstraction layers](03_abstractions.md) |
Collaborator
There was a problem hiding this comment.
We don't have a separate Abstraction layers chapter as far as I know
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Overview of natural language processing and LLMs. Closes #27
Will mainly contribute
python-200/lessons/05_AI_intro/01_intro_nlp_llms.md