Could you please share the checkpoints of the downscaling results in paper?

[Teenye]

I would like to know if the downscale experiment only used data from three variables (z500,t850,t2m with a time interval of 6 hours) for training from 1979 to 2015, validation period in 2016, and testing periods in 2017 and 2018? Can you provide checkpoints for the method results in the paper。

[jasonjewik]

Hey @Teenye, thank you for your interest in our work. Since I am left @aditya-grover's lab a year ago, I do not have access to the machines which contained the checkpoints. @tung-nd do you know if the checkpoints are still available?

[Teenye]

Thank you for your response. I tried running the code for climatelearn to do a downscaling experiment, but the results were still far inferior to those in the paper. I would like to confirm if the dataset configuration for the experiment I am referring to is correct, with three variables (z500, t850,t2m with a time interval of 1 hour) for training from 1979 to 2015, validation period in 2016, and testing periods in 2017 and 2018? Is the time interval 1 hour or 6 hours?

[Teenye]

Hello, may I ask for your advice again. Because I saw in the code that t2m, z500, and t850 were used as input variables, and the output was only one variable (selected from three), is the result of the article actually training a model for each of these three variables separately?

[jasonjewik]

@Teenye as far as I recall... this is correct.