Visão Geral

Ao longo do Curso Open  Source Generative AI, você aprenderá sobre arquiteturas baseadas em transformadores de IA, os fundamentos da programação Python para implantações de IA e a implantação de modelos de Transformer de código aberto. Você também explorará a importância dos requisitos de hardware no desempenho da IA, comparando diferentes arquiteturas de GPU e entendendo como combinar os requisitos de IA com hardware adequado. O curso se aprofunda em técnicas de treinamento, incluindo retropropagação, descida de gradiente e várias tarefas de IA, como classificação, regressão e agrupamento

Objetivo

Após participar com êxito do Curso Open  Source Generative AI você será capaz de:

• A mecânica do aprendizado profundo
• Arquitetura do Transformador
• Requisitos de hardware
• Usando Llama para realizar tarefas de processamento de linguagem natural
• Implantar um modelo de linguagem natural
• Ajuste fino do modelo

Materiais

Conteúdo Programatico

The Mechanics of Deep Learning

1. Gain Experience using the tools and practices to successfully train deep neural networks
2. Learn the mechanics of deep learning
3. Explore prompt learning methods for controlling and influencing model outputs
4. Learn the process of training popular industry models
5. Learn how to apply various transformations or modifications to existing dataset to create additional training examples.
6. Use CNNs (convolutional Neural Networks) to analyze images and perform image augmentation
7. Learn how CNNs improve image prediction accuracy
8. Hands-on exercises to perform fine-tuning and prompt learning with llama-based transformer models

Pre-trained model activities

1. Learn how to select and use pre-trained models to produce immediate results
2. Learn how to train models using RNNs (Recurrent Neural Networks) that will train models in a sequential and time-dependent manner
3. Gain proficiency with Transfer learning techniques to reduce training time and improve generalization, that is, have the AI perform well on unseen or previously unseen data
4. Write an AI data generator that will produce synthetic data to train your model. This is critical when actual data is not available, too costly, or restricted.
5. Learn how to train a model using feature extraction, by writing a program, that will derive a set of informative features from classroom provided original data. You will then use the data as input for machine training.

Transformer architecture

1. Build a transformer architecture
2. Learn why the self-attention matrix is a key component of a transformer model
3. Learn how to compute the self-attention matrix based on:
4. Input Representation: (Query, Key, and Value)
5. Computing Similarities
6. Attention Weights
7. Weighted Sum
8. Final Output
9. Use a pretrained model to translate English to Spanish

Use llama to perform Natural Language Processing tasks

1. Rewrite a classic poem of a well-known author in the tone of another well know author
2. Use Named Entity Recognition (NER) to identify Cajun food in recipes
3. Use domain specific models to improve project accuracy

Deploy a natural language model

1. Download, install and implement a trained NLP model
2. Deploy your NLP to perform the following tasks:
3. text classification
4. sentiment analysis
5. machine translation
6. chatbots
7. question answering