AI: Data Science, ML, GenAI in Python + ChatGPT

Explore unsupervised, supervised, and reinforcement learning with K-means, PCA, KNN, regression, and Q-learning for classification, forecasting, and agent-based decision making.

Survey deep learning models from rnn, lstm, and gru to cnn, autoencoders, and gan, with notes on time series, image classification, and generative training.

Set up your Python development environment with Visual Studio Code, install Python and essential extensions, then create and run your first Python file using the terminal.

Learn to create and run your first Python program, print hello world, install libraries with pip, import pandas, and execute scripts locally via command line and Jupyter environments.

Build a calculator class with a constructor initializing the value to zero, then implement add, subtract, multiply, divide, and reset functions to manipulate it.

Learn how casting transforms data types in data science and coding, converting between integers, strings, floats, booleans, and collections like lists and tuples for tasks such as JSON conversion.

Learn to reshape arrays in numpy using the reshape function, converting one-dimensional arrays into two- or three-dimensional forms and flattening to one dimension for neural nets and machine learning.

learn how to use and convert Python data types, from integers and floats to strings, lists, tuples, sets, dictionaries, booleans, and memoryview, with practical data extraction and processing examples.

Master Python error handling using try and except blocks to catch zero division, value, type, and runtime errors, ensuring the program proceeds to the next workflow step and reports issues.

Apply exploratory data analysis to raw data, clean and describe features, engineer moving average and monthly return, and visualize with box plots and histograms to reveal insights before modeling.

Part 2 of EDA applies detailed exploratory data analysis to Tesla stock data, covering cleaning, descriptive statistics, rolling volatility, correlations, heatmaps, and time-series decomposition.

Build a from-scratch large language model using transformers with multi-head attention and encoder-decoder blocks, trained on a Wizard of Oz dataset to demonstrate end-to-end large language model setup.

Learn to fine-tune Llama 2 with a JSON dataset using supervised fine-tuning and four-bit quantization, then build a chat model you can test and deploy on Hugging Face.

Fine-tune llama 3 with your own dataset using LoRA adapters, guided by Onslaught notebooks and alpaca cleaned data, then save and share the fine-tuned model on Hugging Face.

Use hugging face to authenticate with an API key, choose Gemma 7B or Mistral 7B, and generate prompts like 'write me a poem about machine learning'.

Explore hugging face to access the latest language models and datasets, set up with transformers and torch, and run or fine-tune Llama 3 in Colab or Jupyter notebook.

Train a GPT-style decoder on your own data by implementing the transformer blocks, preparing a tiny Shakespeare dataset, and tuning hyperparameters in PyTorch to generate Shakespeare-like text.

Explore how LangChain enables building apps with large language models, integrating OpenAI, HuggingFace, and custom LLMs, with prompts, streaming, document loaders, and Streamlit.

Learn to use the Falcon 7B Instruct model for chat with LangChain, transformers, and Hugging Face. Load options, tokenizers, pipelines, prompts, and temperature tuning for interactive LLM applications.

Explore sentiment analysis with a pre-trained, fine-tuned Distilbert SST-2 model from Hugging Face via FastAPI REST API, showcasing predicted label (positive/negative) and confidence scores, hosted locally or via ngrok.

Explore how various large language models and platforms—ChatGPT, Claude, Mistral, Gemini, Huggingface—help analyze the Superstore data with quick insights, Python code, and visualizations for ecommerce KPIs.

Use AI chatbots locally when internet access is limited by running models like GPT4All, llama variants, falcon, mistral, and orca in desktop or offline clients.

Explore deep learning fundamentals by overviewing subfields and models, including RNNs, LSTMs, GRUs, CNNs, autoencoders, GANs, and deep reinforcement learning, with a focus on sequence data applications.

Train a long short-term memory autoencoder on stock price time series to detect anomalies using moving averages, standard deviation thresholds, and a 30-day prediction window.

Build and train a convolutional neural network to classify handwritten digits using the mnist dataset. Use conv2d, relu, max pooling, dropout, flatten, dense, and softmax, with data preprocessing and evaluation.

Explore etl workflows and data exploration with metal sheet in JupyterLab, generating code from visual filters for shipping date, city, profit, and category to shape a ready data frame.