Introduction to Generative AI With Python and Open AI API

Explore few-shot prompting, using a few examples to set context for a model, contrast with zero-shot prompting, and discuss limitations and the upcoming chain-of-thought prompting.

Learn zero-shot prompting with the OpenAI API by building a simple chat flow using system and user prompts, GPT-3.5, without examples, and tuning temperature and max tokens.

Explore chain of thought prompting to teach models how to reach answers, contrasting zero-shot, few-shot, and automatic approaches, and discuss limitations and practical tips.

Learn how automatic chain of thought prompting overcomes the limits of manual prompts by clustering questions with embeddings, selecting a representative question from each cluster, and generating zero-shot reasoning.

Explore prompt engineering with the OpenAI API to generate synthetic housing data as a CSV. Learn setup, prompts, and a Python workflow to create ten rows.

Learn to generate a Python program with OpenAI API that creates a csv using pandas, producing 100 rows with house size, price, location, and bedrooms.

Create a multi-table csv or three pandas dataframes in python using OpenAI prompts to define housing, location, and house type schemas with foreign keys and coherent data relationships.

Generate data for fine tuning OpenAI GPT models by prompting product name and category to produce descriptions, format input-output pairs, and expand samples with a for loop and regex parsing.

Explore fine tuning OpenAI's LLMs using the official docs, understand when to use fine tuning versus retrieval augmented generation, and learn training data formats, token limits, and cost considerations.

Fine tune OpenAI models via the playground dashboard and console, prepare JSON training data with ten examples, configure epochs, monitor training loss and validation, and test the fine tuned model.

Fine tune a language model with Python code using the OpenAI API, including setup, data preparation, train/validation split, uploading files, launching the fine-tuning job, and inference.

Explore batch API theory for OpenAI, learning asynchronous, 50% lower cost batch processing with large datasets and embeddings, with non-instant responses and compatibility with fine-tuned models.

Explore using the OpenAI batch API in the Playground to process JSON data with chat models, build system and user prompts, prepare input files, run batches, and read structured outputs.

Learn to categorize movies from the IMDb top 1000 with Python and the OpenAI batch API. Prepare data with pandas and JSON, upload a CSV, and retrieve results.

Apply batch processing to generate image captions using the OpenAI API, composing prompts with image URLs and titles, and handling batch jobs from data import to results retrieval.

Explore image generation with DALL-E 2 and DALL-E 3, including editing, variations, prompts, and safety rewriting, through theory before coding in Python with the OpenAI API.

Generate images with the OpenAI API in Python by selecting dall-e 3 or dall-e 2 models, providing prompts, choosing HD or standard size, and requesting variations and edits with masks.

Explore OpenAI's text to speech API with six built-in voices, streaming capability, two models: tts and tts one hd, and language limits with no custom voice training.

Explore OpenAI's whisper speech-to-text API, covering transcription and translation, input limits and formats, time-stamped outputs, chunking, and prompting to improve transcripts.

Learn to build text-to-speech and speech-to-text pipelines in Python using the OpenAI API, including streaming output to audio files. Practice transcription and translation with whisper models and multilingual prompts.

Explore the theory of embeddings, where tokens map to vectors of floating point numbers, capture meaning, measure relatedness by cosine similarity, and enable uses like search, clustering, and recommendations.

Explore how to generate embeddings with OpenAI API, selecting models, handling 1536-dimension outputs, and applying L2 normalization in practical Python examples.

Practice generating embeddings in Python with OpenAI API, focusing on token encoding and cost. Load review data, merge title and text, filter long samples, and create embeddings with a utility.

Build a QnA system with embeddings and a rag workflow, leveraging a vector store, cosine similarity, and GPT models to answer questions based on retrieved documents while managing token budgets.

Explore semantic search with embeddings and cosine similarity to match product descriptions to relevant reviews, using a consistent embedding model and simple retrieval of top results.

Build a code search tool with embeddings and cosine similarity by parsing code, extracting functions, creating embeddings, and performing semantic search across a repository.

Explore embeddings through 2D visualization with t-SNE, deepen learning with random forest regression and classification, apply zero-shot classification, perform k-means clustering, and summarize clusters with GPT-4.

Discover the assistant API to build your own AI assistant in your application using OpenAI models, with Python functions and tools like code interpreter, file search, and function calling.

Start the assistant API in the OpenAI dashboard, test with the playground, and set up a math tutor assistant with a selected model and code interpreter before coding in Python.

Build and test an OpenAI assistant by creating the assistant and thread, sending messages, running the process, and deleting the assistant and thread.

Explore the assistant api deep dive, configuring instruction, tools, and tool resources, and learn about threads, runs, and code interpreter for data visualization.

Learn how OpenAI's built-in vector store enables file search by chunking documents, embedding them, and performing keyword and semantic queries, with practical limits and setup tips.

Explore using OpenAI file search to index documents, upload to a vector store, attach stores to an assistant or thread, and query for contextual results.

Explore the theory of function calling with the OpenAI API, connecting models to external tools via metadata, arguments, and structured outputs.

Learn to implement function calling with the OpenAI API using chat completion and assistant workflows. Define function metadata, trigger tool calls, and execute results to produce weather insights.