Firstly, you should try to set the exposure times so that the exposure level of the source and the sky is below 10,000 ADU. If this is the case, then you do not need to worry about non-linearity. If this is not the case, then you should think about how non-linearity may effect your data and if you really think that you should apply a non-linearity correction, then you are on your own at this point, because we have not developed any software to do this. The reason for this are twofold. Firstly, different parts of the array are read out at different times, so the non-linearity correction at row 512 is always going to be larger than the correction at row 1. Secondly, since the bias is a function of the flux, it is difficult to know the precise value of the signal to correct. This second obstacle is not as severe as you may at first think as the error you would introduce by ignoring the zero level offset will in most case be much smaller than the correction itself. If this does not deter you and you develop a way of correcting the non-linearity, then we'd be happy to hear from you.
The non linearity of the array can be fitted with the following function.
Since we have not developed software to correct for non-linearity, this formula has not been tested.