Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Add ML coupling #14

Draft
wants to merge 22 commits into
base: datawave_ml
Choose a base branch
from
Draft

Add ML coupling #14

wants to merge 22 commits into from

Conversation

jatkinson1000
Copy link
Member

@jatkinson1000 jatkinson1000 commented Aug 6, 2024 •
edited
Loading

This PR adds a coupling routine that calls the neural net to close #13

  • Write an interface function to be called instead of the existing scheme that will implement the NN
    • Concatenate the data as required into the expected input format - requires help from @surbhigoel77
    • Transform this reshaped/concatenated data into a torch tensor
    • Run the NN
    • Deconcatenate the returned data into physical variables
    • Use the variables as appropriate in the ptend updates

It is currently WIP.

@jatkinson1000 jatkinson1000 added the Fortran This is an issue on the Fortran/CAM side of the coupling project label Aug 6, 2024
@jatkinson1000 jatkinson1000 self-assigned this Aug 6, 2024
@jatkinson1000 jatkinson1000 mentioned this pull request Aug 19, 2024
Open
4 tasks
@jatkinson1000 jatkinson1000 force-pushed the coupling branch 2 times, most recently from aa106e6 to 8cb0e6c Compare August 19, 2024 10:37
@jatkinson1000
Add variable for the convect_dp_gw net filepath to be read from namel…
fd6e669 
…ist.
@jatkinson1000
WIP Adapt gw_drag.F90 to import ftorch and read in a net from file. T…
bed301e 
…ODO: Move reading net to init routine.
@jatkinson1000
Move reading of NN to initialisation routine for gw_drag.
7968fa9
@jatkinson1000
Add a gw_final() subroutine to destroy NN instance.
b3ac3e9
@jatkinson1000
Remove spurious test code.
16b53ac
@jatkinson1000
Update source to use latest version of FTorch following API Changes.
dfd451d
@jatkinson1000
WIP add gw_ml file with starting point for a routine to run the ML pa…
0d334db 
…rameterisation via FTorch.

* Implements reshaping of the data to the form expected by the net as a starting point.
  * Consider using torch layout argument to avoid transposition at a
    later date.

* Does not implement any normalisation of features as we need to discuss
  how to do this on a chunked grid when data is provided for the global
  grid.

* Not currently called from anywhere.
@jatkinson1000
Add normalisation file to namelist variables, read, and distribute.
01950bc
@jatkinson1000
Create gw_ml initialisation routine to read in net and normalisation …
fa35aed 
…variables from file.
@jatkinson1000
Add finalisation routine to gw_ml to destroy the net.
ca19081
@jatkinson1000 jatkinson1000 mentioned this pull request Sep 9, 2024
Open
@jatkinson1000 jatkinson1000 mentioned this pull request Sep 27, 2024
Open
@jatkinson1000 jatkinson1000 mentioned this pull request Nov 4, 2024
Merged
@jatkinson1000
Copy link
Member Author

Before merging need to:

  • Update docs to point at correct branch (datawave_ml, not coupling).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@surbhigoel77 surbhigoel77 Awaiting requested review from surbhigoel77

@yqsun91 yqsun91 Awaiting requested review from yqsun91

Assignees

@jatkinson1000 jatkinson1000

Labels
Fortran This is an issue on the Fortran/CAM side of the coupling project
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

Couple CAM to loaded NN
1 participant
@jatkinson1000