Learn

Hands-on tools I’ve built for my courses. Everything runs in the browser, with nothing to install.

Tool · Diagram ⇄ PyTorch

Torch Canvas

A visual editor for neural networks that writes PyTorch as you draw, and turns PyTorch modules back into diagrams. It runs in the browser with nothing to install.

Open Torch Canvas All 22 example models →

GPT-2 (124M architecture) 124.44M Uses GPT2Block Open in Torch Canvas → Llama 2 7B 6.74B Uses LlamaDecoderLayer Open in Torch Canvas → BERT-base (masked LM) 109.51M Uses BertLayer Open in Torch Canvas → Qwen2.5-VL 3B 3.75B Uses Qwen2_5_VLVisionBlock, Qwen2_5_VLVisionStage, Qwen2_5_VLPatchMerger, Qwen2_5_VLDecoderLayer Open in Torch Canvas →

Interactive demos

• Single Neuron Explorer

  A neuron with two inputs draws a line (hyperplane) through input space. Drag the weights and watch the decision boundary move.

  CSC4220 perceptrondecision boundary

• 3-Input Neuron Explorer (3D)

  The same idea in three dimensions, where the boundary becomes a plane.

  CSC4220 perceptrondecision boundary

• ROC Curves and Data Separability

  Move the threshold and the class overlap to see how the ROC curve and model quality respond.

  CSC4220 evaluationROC

• The Optimization Lens: Group Handouts

  In-class exercise framing different learning algorithms as optimization problems.

  CSC4220 optimization

• Operator Norm as a Matrix Equation

  A 2×2 matrix warps the unit circle; find the input direction it stretches most.

  linear algebraoperator norm

• Operator Norm via Power Iteration (3D)

  A 3×3 matrix turns a sphere into an ellipsoid while power iteration homes in on the maximum stretch.

  linear algebraoperator norm

Course pages, with slides, are under Teaching.