Teach on Udemy

Turn what you know into an opportunity and reach millions around the world.

Learn More

Your cart is empty.

Keep shopping

Generative AI for Complete Beginners

Name: Generative AI for Complete Beginners
Rating: 4.4 (1192 reviews)

Learn GenAI, LLM, Langchain, Prompt Engineering in a project based approach

Created bySourabh Sinha

Last updated 7/2024

English

English [Auto],Hindi [Auto],

What you'll learn

Fine tuning an LLM
Prompt Engineering
Gen AI Basics
OpenAI API Usage
Langchain Basics

Coding Exercises

This course includes our updated coding exercises so you can practice your skills as you learn.

Course content

4 sections • 27 lectures • 2h 13m total length

What is Generative AI?3:22
Generative AI refers to models that create new data from prompts. It can generate text, images, audio, video, and code, enabling intuitive content creation.
Resources0:01
Ven Diagram of AI x GenAI1:56
Explore the Venn diagram of AI, machine learning, deep learning, and generative AI. Learn how these domains use data to identify patterns and power creative content.
Roadmap to learn Generative AI10:21
Navigate the roadmap to generative ai by mastering Python fundamentals, ml and dl concepts, nlp, llms, and deployment, plus vector databases, apis, and cloud ecosystems for practical product-ready ai.
Preparing our machine for GenAI Experiments7:55
Install Python and Jupyter Notebook, create a virtual environment, and explore local LM tools such as Jan dot AI, LM Studio, and Ulama to prepare your machine for GenAI experiments.
Using OpenAI API Programmatically9:39
Learn to use the OpenAI API programmatically and via GUI with Postman, including creating an API key, setting authorization headers, and building chat completions with GPT-3.5-turbo.
LLM Toolbax : Temperature2:03
Explore how temperature in large language models shapes creativity and output style. Learn why a temperature around one balances factual accuracy and creative responses.
LLM Toolbox: Max Token1:11
Explore max tokens as a tool to restrict generation length for token saving or cost control, demonstrated by setting max tokens to five and noting the finish reason becomes length.
Understanding Tokenization9:18
Explore tokenization basics, from corpus and vocabulary to character, word, and subword methods. Learn how token types impact semantics, model performance, and pricing, with practical insights from the OpenAI tokenizer.
Practical Implementaion: Character Tokenization1:20
Learn to implement character tokenization by converting a string into a list of characters, producing tokens as text, and test the simple implementation.
Practical Implementation: Word Tokenization1:29
Learn to implement word tokenization in Python by splitting text on spaces to produce individual words, and understand its advantages over character or subword tokenization and challenges with out-of-vocabulary words.
Practical Implementation: Sub Word Tokenization2:09
Explore subword tokenization that breaks rare words into meaningful subword units such as character engrams and byte pair encodings, improving handling of misspellings and out-of-vocabulary words.
Word Vectors and Embeddings3:35
Explore how tokenization, vectorization, and embeddings transform raw text into meaningful word vectors, and learn how cosine similarity reveals relationships among words like dog, wolf, and fish.
Practical: Word Vectorization and Embeddings4:56
Demonstrate tokenizing sentences with auto tokenizer, converting tokens to IDs, and creating word embeddings using Bert base uncased in PyTorch, revealing a 1-by-768 embedding and its cosine similarity.
Practical: Cosine Similarity3:12
Explore cosine similarity to compare word embeddings built from token ids and PyTorch tensors, using dog, wolf, and fish to show higher similarity for related words like happy and cheerful.
Basics of Generative AI
Experimentation with OpenAI API

Introduction to LLM Training2:32
Explore how large language models use transformers, with encoders, decoders, and attention, to generate and translate text. Learn data collection, cleaning, bias considerations, and supervised training with backpropagation.
Metrics and Benchmarks3:26
Evaluate language models using a metrics toolbox: coherence, relevance, factual accuracy, and fairness, and compare architectures and benchmarks to select the right LM for a given use case.
Introduction to Fine Tuning2:40
Fine tuning a pre-trained large language model tailors it to a specific domain by training on a curated dataset, reducing bias, and shaping outputs into concise lists or JSON.
Why Fine Tuning is needed?3:04
Explore why fine tuning matters for text moderation models, showing how tailored datasets and prompts improve multilingual toxicity detection, indecent content labeling, and cost efficiency.
Data Collection for Fine Tuning10:10
Explore fine tuning on OpenAI: prepare JSON training and validation data with a system, user, and assistant prompt structure, convert from CSV to JSON, and use Dspam to score toxicity.
Estimating the cost of Fine Tuning5:15
Estimate fine-tuning costs by calculating total training and validation tokens with the Tiktokin library and GPT 3.5 turbo, then apply OpenAI's 8 USD per million tokens to derive the cost.
Formatting Data for Fine Tuning9:41
Format data for OpenAI fine-tuning by converting data frames to jsonl, stringifying content fields, and creating a scores column with stringified JSON to support system prompts and training data.
Fine tuning in Action7:36
Install openai packages, upload training and validation jsonl files, and start a fine tuning job with GPT-3.5-turbo. Monitor training and validation losses for early stopping and learning rate tweaks.
Using custom fine-tuned model3:58
Learn how to deploy a custom fine-tuned model via the OpenAI API, swap in your model name, and run chat completions with the OpenAI SDK, including training data considerations.
Serving fine tuned model as an API7:54
Learn how to serve a fine-tuned model via an api using fast api and uvicorn, with environment-managed keys, a post endpoint for chat completions, and openai integration.
Fine Tuning Quiz
Hands on Fine Tuning Excercise

Requirements

Python Basics e.g. OOP
Knowledge of RestAPI would be useful

Description

This course is a practical guide to learning Generative AI concepts:

- Large Language Models

- Tokenization, Word Vectors and Embeddings

- Fine tuning LLMs

- Langchain

- Prompt Enginnering concepts

- Using OpenAI API

The field of artificial intelligence has seen incredible advances in recent years, with one area gaining significant traction - generative AI. This cutting-edge technology is poised to revolutionize how we create and interact with all kinds of digital content.

So, What exactly is generative AI? At its core, it refers to AI models that can generate new data, rather than just analyzing existing data. This could include generating text, images, audio, video, computer code, and more - often starting from just a basic prompt or input from a user. What makes this technology so powerful is how user-friendly it is becoming. You can simply describe a scene or concept, and the AI model generates high-quality digital content in response - almost like magic!

1. Artificial Intelligence:

This prime spot is reserved for just plain AI. It's a broad term, the overarching goal: machines that mimic human intelligence. That includes everything from playing chess to diagnosing diseases, from composing music to writing this blog.

2. Machine Learning:

Machine Learning (ML), a subset of AI. It's where the magic of learning from data happens. ML algorithms don't need explicit programming – they gobble up data, identify patterns, and improve their performance over time.

3. Deeper and Deeper: When ML Gets Fancy - Deep Learning

This is ML on steroids, using complex artificial neural networks loosely inspired by the human brain. Deep learning is the secret sauce behind many of AI's recent breakthroughs, allowing for crazy-powerful stuff like image and speech recognition.

4. Generative AI

This is the elephant baby of the AI world or the rebellious teenager with a paintbrush. It uses machine learning to create entirely new content, from composing electronic dance music symphonies to generating hyperrealistic images of, well, anything you can imagine (including, unfortunately, deepfakes so convincing they'd make our grandma believe the orange cats can dance).

Who this course is for:

Beginners willing to understand gen ai
Beginner willing to learn LLM
Students who want to build projects on Generative AI
Students willing to learn Langchain

Generative AI for Complete Beginners

What you'll learn

Explore related topics

Coding Exercises

Course content

Introduction15 lectures • 1hr 2min

LLM Fine Tuning Arc10 lectures • 56min

Prompt Engineering1 lecture • 6min

Major Project: Building a PDF Chatbot1 lecture • 9min

Requirements

Description

Who this course is for: