You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
earth2studio.data.utils.prep_data_array assumes that "time", "lead_time", and "variable" dimensions occur before other dimensions (e.g., "lat", "lon", "ensemble"). This leads to inconsistent tensors/coords. Most importantly, this prevents the "ensemble" dimension from being moved around, e.g., for broadcasting.
Example:
from earth2studio.data import NCAR_ERA5, prep_data_array
from earth2studio.utils.time import to_time_array
times = to_time_array(["2023-03-24 12:00:00"])
variables = ["t2m", "tcwv"]
ncar_era5 = NCAR_ERA5()
da = ncar_era5(times, variables)
x, coords = prep_data_array(da.transpose("lat", "lon", ...))
Version
main
On which installation method(s) does this occur?
No response
Describe the issue
earth2studio.data.utils.prep_data_array
assumes that"time"
,"lead_time"
, and"variable"
dimensions occur before other dimensions (e.g.,"lat"
,"lon"
,"ensemble"
). This leads to inconsistent tensors/coords. Most importantly, this prevents the"ensemble"
dimension from being moved around, e.g., for broadcasting.Example:
Result:
x.shape
-torch.Size([721, 1440, 1, 2])
list(coords)
-['time', 'variable', 'lat', 'lon']
The text was updated successfully, but these errors were encountered: