Examples

Interactive deep learning demos built with torch.js. Train models in your browser with WebGPU acceleration.

Train a neural network to recognize handwritten digits. Draw your own and watch the model predict in real-time.

Watch a linear model learn to fit a random polynomial function. Visualize the learning process in real-time.

Explore latent space representations. Generate new digits by sampling from the learned distribution.

LSTM network learns to predict sine waves. Watch future values being forecasted in real-time.

Upscale low-resolution images using a convolutional neural network with sub-pixel convolutions.

Character-level language model. Train a tiny GPT to generate Shakespeare-style text.

Train without backpropagation using Hinton's Forward-Forward algorithm. Each layer learns independently.

Deep Convolutional GAN learns to generate realistic MNIST digits from random noise.

RNN/LSTM learns to generate Shakespeare-style text character by character.

Learn image similarity by comparing pairs. Determines if two digits are the same or different.

Want to build your own? Check out the documentation.