⬅️ The Backward Pass

Watch a red gradient glow flow right to left through the graph — local derivative × upstream gradient, at every node.

📋 How To Use This Lab
  1. Set x and y with the sliders, then press Trigger backward() ▶.
  2. Press Next Step ▶ to reveal one gradient edge at a time, right to left.
  3. Turn on Show Math Formula to see the symbolic ∂ notation, not just numbers.
  4. Watch x especially — its gradient is the SUM of two separate backward paths.
Delta
🤖 DELTASet x and y, then trigger backward() — I'll walk the graph in reverse and show you every gradient multiplication.
🧑‍🏫 PROF. TORCHbackward() starts with df/df = 1 at the output, then multiplies local derivatives backward through the chain rule.

🎛️ Input Values