nan values in regression model

[snash4]

Hi,
I get nan value for loss when I use the quadratic model
Y_predicted = X * X * w + X * u + b
What is the possible reason for this?