Your question is Debugging Non-Convergent Models. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
What strategies do you use to debug models that are failing to converge?
Describe a practical, ordered debugging process rather than listing isolated fixes. Cover data validation, preprocessing, loss and gradient inspection, optimization settings, model capacity, numerical stability, and how you distinguish underfitting from an optimization failure. Include the metrics or plots you would monitor and a small implementation demonstrating your approach.