poloclub/transformer-explainer — reverse-engineered prompt

Build me an interactive web app that explains how Transformer models like GPT work visually. I want someone with no deep learning background to be able to type their own text into the page, run a small GPT 2 style model in the browser, and see how the model predicts the next words step by step.

Make the experience feel like an educational playground, with clear visual sections for tokens, embeddings, attention, layers, probabilities, and next token prediction. As the user edits the text, the visuals should update so they can connect what they typed to what the model is doing internally. Add simple explanations beside each part so it teaches while people explore, not just charts without context.

Please make it polished, responsive, and easy to run locally in a browser. Use the existing modern JavaScript app setup, and look up current docs online if you need to.