About the length of data

[kelen1]

Hi! Thanks for your wonderful work first! When I try to reproduce your code, I find that in the preprocessed data, x takes the data of the first 50 time steps, and y is after 50, but many of the original data are 0-49 in length, so y is empty, does the true value of y still exist?

[jchengai]

Hi, @kelen1, the regression mask is used to prevent the problem you mentioned during the training

https://github.com/jchengai/forecast-mae/blob/cb86ea92601d23a8af7713389e9fb78d7e546a65/src/model/trainer_forecast.py#L89-L91

[kelen1]

Thanks!