Transformer Math Explorer [P]

This is an interactive math reference for transformer models, presented via dataflow graphs, all the way down to elementary math. Covers models from GPT-2 to Qwen 3.6, with MLA, MoE, RoPE, MTP, hybrid attention, and other variants toggleable. Originally made this for myself to keep track of all the variations. If you find errors or find something unintuitive or misleading let me know!