Reimplement model #27
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The parametrized source is nowhere near smooth enough. On the mean, this leads to large bias. And the bias is not smooth at all. This could possibly result from under training.
call_with_xr wraps a call to a PyTorch Module which takes in and spits out dictionaries.
Having trained a momentum parametrization, I now need to plug it into the module. The path to the trained momentum variable should be loaded in UWNET_MOMENTUM_MODEL environmental variable.
This allows SAM to use the velocity field adjusted by the neural network.
SAM + NN with khyp=1e16 goes unstable after 300 time steps (10 hrs) of model time. The previous results I had were with khyp=5e16.
This run blew up for some reason. This notebooks shows that the problem is isolated to the boundaries.
Because the momentum source appears unstable near the boundaries, I will mask the neural network there.
I used masked boundary conditions. Unforunately, this did not improve the longer-term biases in the simulation
CAM radiation expects 4 bytes reals, but the zenith angle was computed with type "real". Then SAM is compiled in double precision this caused the model to crash after 1 time step. This bug is now fixed
This commit makes it much easier to pass input to the torch model. Now it can accept inputs with a size (*, z, y, x) whereas before it needed (*,z, x*y). Moving the reshaping into the bottom level method MLP.rhs makes it much easier to deal with.
nbren12
force-pushed
the
reimplement-model
branch
from
20b7a53
to
e413019
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.