[PyTorch] PyTorch Tutorial Projects

Getting Started with aitextgen

Perform text-based AI training and generation using OpenAI's GPT-2 and EleutherAI's GPT Neo/GPT-3 architecture.

Go Project 

aitextgen is a text-generation library created by Max Woolf that uses GPT-2 and GPT Neo/GPT-3 architecture in combination with PyTorch, Hugging Face Transformers, and python-lightning.

This ML Showcase entry is a fork of the notebooks in the aitextgen repo. The contents are as follows:

• Train a Custom GPT-2 Model + Tokenizer w/ GPU
• Train a GPT-2 (or GPT Neo) Text-Generating Model w/ GPU
• Generation Hello World
• Training Hello World
• Hacker News Demo
• Reddit Demo

 

Object Detection With Detectron2

Go Project

This tutorial implements the new Detectron2 Library by facebook, and shows how to train on your own custom objects for object detection.

We use a public blood cell detection dataset, which is open source and free to use. You can also use this notebook on your own data.

To train our detector we take the following steps:

• Install Detectron2 dependencies
• Download custom Detectron2 object detection data
• Visualize Detectron2 training data
• Write our Detectron2 training configuration
• Run Detectron2 training
• Evaluate Detectron2 performance
• Run inference on test images

 

PyTorch Tutorial: Building a Neural Network

Go Project

PyTorch has quickly become today's most popular deep learning platform. In this tutorial by Soumith Chintala, one of the creators of PyTorch, you'll learn how to construct neural networks in PyTorch using the torch.nn package. Topics covered include:

• Defining a neural network
• Processing inputs
• Computing the loss
• Backpropagating the error
• Updating the network weights

 

PyTorch Tutorial: Training a Classifier

Go Project

PyTorch has quickly become today's most popular deep learning platform. In this tutorial by Soumith Chintala, one of the creators of PyTorch, you'll learn how to train an image classifier on the CIFAR10 dataset using the torchvision package. Specifically, we'll see how to:

• Load and normalize the CIFAR10 training and test datasets using torchvision
• Define a Convolutional Neural Network
• Define a loss function
• Train the network on the training data
• Test the network on the test data

 

PyTorch Tutorial: Working with Tensors

Go Project

PyTorch has quickly become today's most popular deep learning platform. This tutorial by Soumith Chintala, one of the creators of PyTorch, explains how to use the library's core functions for working with tensors, performing basic operations, converting tensors to NumPy arrays, and using CUDA.

 

Play Super Mario Bros with a Double Deep Q-Network

Go Project 

By the end of this tutorial, you will have a working PyTorch reinforcement learning agent that can make it through the first level of Super Mario Bros (NES).

For a complete breakdown of the code and the theory behind it, check out Play Super Mario Bros with a Double Deep Q-Network on the blog.

 

PyTorch Tutorial