Examples

This gallery of examples uses luz to train and validate a range of common deep learning architectures. The gallery also demonstrates basic and advanced usage of luz.

CharGPT

intermediate

Train a character-level GPT-2 on Shakespeare texts.

See code

Binary classification

basic

Demonstrates using pre-trained models to build a binary classification model.

See code

Autoencoder

basic

Builds an autoencoder for the MNIST dataset. Demonstrates overwriting the predict method

See code

Virtual batch size

advanced

Showcases how to create a custom fully customized training step

See code

Simple CNN

basic

Trains an AlexNet-like CNN to classify digits of the MNIST dataset

See code

DCGAN

intermediate

Deep Convolutional Generative Adversarial Network to generate images.

See code

MixUp augmentation

intermediate

Demonstrates using the MixUp callback

See code

Triplet loss

intermediate

Train an embedding model to minimized the triplet loss.

See code

UNET implementation

intermediate

Implements a UNET model to separate the background of images of cats and dogs.

See code

Text classification from scratch

basic

Implements text classification from scratch.

See code

Training a causal language model from scratch

advanced

Implements datasets and trains a causal language model from scratch using R source code.

See code