Deterministic Solution for Excel Solver LM and Neural Net

I’m trying to replicate @jeremy 's excel notebook example but I’m coming up with a different average losses using Excel’s solver. Is this process of solving for the linear model parameters non-deterministic? I tried inputting the values that Jeremy used for starting params (obviously without knowing the full un-rounded number so small differences) and came out of the at solver with a an average loss of .134, not the .08 that Jeremy got.

I downloaded the notebook directly from the course site on GH so data is the same.

What am I missing here?

before solver

after solver