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
When doing inference with LAM, the model will output an interior prediction + the boundary conditions.
During rollout, when we feed the output back to the input, we need a way to replace the boundary conditions with new initial condition at the next timestep (e.g.: ERA5).
We could treat this boundary layer as a forcing: input fields that get updated at every tilmestep from disk. Using the cutout as a mask.
The text was updated successfully, but these errors were encountered:
When doing inference with LAM, the model will output an interior prediction + the boundary conditions.
During rollout, when we feed the output back to the input, we need a way to replace the boundary conditions with new initial condition at the next timestep (e.g.: ERA5).
We could treat this boundary layer as a forcing: input fields that get updated at every tilmestep from disk. Using the cutout as a mask.
The text was updated successfully, but these errors were encountered: