Decision boundary with a circle shape

I have a problem trying to create a decision boundary a circle instead of a line
All I have to do is to change the model to w1x1^2+w2x2^2+b and instead of multiplying the loss by x1 or x2 in the symultanious update I multiply it by the features squared right ???
Because when I did that the learning started taking very long even though I changed the rate to a lot of different values and the circle model is not converging.
Can anyone help???

can you share your code?

Here is the link to the drive

I don’t think changing the model is required - but you need to add new features using quadratic terms, and you need a training set that populates a 2D range of examples.

It’s covered in MLS Course 1 Week 3.