Google TFT prediction on training / validation

Hello all, I got some questions about Google TFT implementation on pytorch

1 - If I understood well, the predict method of tft model, returns prediction of the last max_prediction_length values. how to predict on whole training / validation set using tft model? is there any built-in function for that ?
2 - The plots I get have time index on xaxis, is there any way to show date (which is a feature on my dataset) on xaxis ?
3 - Since the model takes min, max encoder / prediction length which one does it use on training / prediction ? max ?

Thanks for the help :slight_smile: