classification and regression

Titanic : simple data for classification according to categories and numbers

Iris : simple data for classification according to categories and numbers

Boston Housing : simple data for regression according to categories and numbers

Image classification and generation

Mnist: 60,000 28x28 images (digits) for simple tests (generation or classification)

FashionMnist: 10,000 28x28 images (clothes) for simple tests (generation or classification)

CelebA: 60,000+ larger and color images (generation)

Cifar-10: small color images in 10 categories (classification)

Flickr8k: small color images associated with text (text 2 image, image 2 text)

Text classification and generation

Children's Stories: 250 children's stories in English to begin with text generation

French cooking recipes

English poems (generation)

French news (generation)

Movie Reviews (in English): Classification

English news with 4 categories (classification)

classification and regression

classification model trained with the Titanic dataset

regression model trained with the Boston Housing dataset

Image classification and generation

image classification model trained with the fashion mnist dataset

auto encoder model trained with the fashion mnist dataset

auto encoder model trained with the celebA dataset

variational auto encoder (VAE) model trained with the fashion mnist dataset

variational auto encoder (VAE) trained with the celebA dataset

decoder (VAE) trained with the fashion mnist dataset

decoder (VAE) trained with the celebA dataset

diffusion model trained with the fashion mnist dataset

diffusion model trained with the celebA dataset (with EMA)

Text classification and generation

text classification model trained with the English news dataset and the word tokenizer limited to 10,000 tokens lowercase

text generation model using LSTM trained with the Children's Stories dataset and the word tokenizer lowercase (6170 tokens)

text generation model transformer (2 blocs of 4 attention heads) trained with the Children's Stories dataset and the word tokenizer lowercase (6170 tokens)