What if my own dataset have missing values at some time steps for TFT model?

I want to apply TFT model to my own dataset, but there are missing values at some time steps. My question is whether the model can handle missing values? Do I need to preprocess my data to keep the time series continuous?

No file, test, or entry point is named. Start by locating the TFT example notebook and its dataset-preparation steps, then determine whether missing time steps and missing values are supported; done means documenting the required preprocessing and continuity behavior for users.

Written by the indexing model from the issue text.