Why does making larger n_forecasts change outcome for small yhats

[JCCen]

Hello everyone,

First off, I'd like to express my appreciation for the fantastic work on the Neural Prophet package.

I've observed an unusual behavior. When I construct a model with n_forecasts=2, it performs well. However, if I build another model with identical parameters but change n_forecasts to 150, the outcomes differ significantly.

I anticipated that the predictions from the second model at a 2 day time horizon (yhat2) would be nearly the same as the first. Yet, there's a noticeable drop in the performance of the second model for 2D horizon forecast. Could anyone shed some light on why this might be happening? Additionally, what steps could I take to mitigate this issue?

Thank you for your guidance and support.