LLM Broad Knowledge Flashcards
(9 cards)
What is a Large Language Model (LLM)?
An LLM is a type of AI that can generate and understand human-like language by learning patterns from massive amounts of text. It predicts the next word (or token) in a sequence based on context.
How does an LLM generate text?
It takes in a sequence of words, converts them into numbers, processes them through a neural network (specifically a transformer), and outputs a probability distribution over possible next words. It selects the most likely next word and repeats.
What is the architecture used in most LLMs today?
Transformers. Specifically, transformer decoders, which are neural networks that use self-attention to weigh the importance of different words in a sentence when generating the next word.
What is a token in an LLM?
A token is a small piece of text—usually a word fragment, full word, or punctuation—that the model processes. Text is split into tokens before being fed into the model.
What is a Transformer in machine learning, and how does attention work within it?
A transformer is a powerful sequence model that uses attention to understand which parts of the input matter most when predicting the next word.
This is the heart of the transformer.
Self-attention allows each word (token) to “look at” all other words in the sentence and decide which ones are most relevant to understanding its own meaning
What is Generative AI
What is Generative AI it is a subset of Artificial Infelligence (AI) it focuses on generating new content such as images text audio and video
What are the two most common different types of LLMs you can use
- Propriety Models (LLMs as a service) Pros are speed of development and quality the cons are Cost data privacy and vendor lock in
2.Open Source Models
Pros are Task Taikoring Inference Cost and Control
Some cons are Upfront time investment Data requirement and Skill sets
How can you train LLMs Models
You can use one or the other or birth you can use Pre-training where you initially train your LLC on a large set of data
The second thing is Find tuned Models it is the process of further training a pre-trained model
What is RAG
RAG (Retrieval-Augmented Generation) is an architecture that combines a language model (like GPT) with an external knowledge retriever to generate more accurate and up-to-date answers by retrieving relevant documents first, then generating text based on those documents.
