Gradient function not able to find optimal theta but normal equation does

Question

I tried implementing my own linear regression model in octave with some sample data but the theta does not seem to be correct and does not match the on