I have a question about how training has been done for the models. The ERA5 dataset has weather data for each hour from 1979 to 2018, where you used 1979 to 2016 for training, and evaluation/forecasting is done from 2017 to 2018 for 3 days/5 days. Did you use all the data from 1979 to 2016, like hourly data to train the model or daily data or 3-day data for training as well, as it is a little bit confusing to see in the paper and repo?
Hi,
I have a question about how training has been done for the models. The ERA5 dataset has weather data for each hour from 1979 to 2018, where you used 1979 to 2016 for training, and evaluation/forecasting is done from 2017 to 2018 for 3 days/5 days. Did you use all the data from 1979 to 2016, like hourly data to train the model or daily data or 3-day data for training as well, as it is a little bit confusing to see in the paper and repo?