How do I extend the prediction window for Cox-Time model? about pycox HOT 1 CLOSED

The Cox-methods' predictions are dependent on the baseline_hazards_ which are computed non-parametrically from the training set (similarly to the Kaplan-Meier estimator). The methods can therefore only compute predictions at the time-points defined by the baseline_hazards_.
So when you are only able to obtain predictions up to time 1012, this is likely because 1012 is the highest duration in your training set. The argument max_duration can only be used to decrease the maximum prediction duration (to decrease computations), and cannot be used to increase the maximum duration.

If you want to interpolate between the time-points of your predictions, or if you want to extrapolate beyond them, you need to code this on your own, as this is not supported.
I would also advice you to be careful when prediction for higher durations than what you observe in your training set. As you have no data at such high times, your estimates will probably not be very good.

from pycox.