Training loop playground

Every model in this book is trained by the loop you can watch here: predict, measure the loss, nudge the parameters, repeat. Everything runs in your browser; nothing you do here leaves the page.

Data
Model
Step 0. Loss: —

What to notice: the line model always finds the same two numbers; the polynomial has more freedom and can overshoot at a high learning rate; the tiny network needs many more steps to bend into shape.