m2lines / CAM-ML

Community Atmosphere Model
0 stars 0 forks source link

Add yog code #9

Closed jatkinson1000 closed 1 week ago

jatkinson1000 commented 3 months ago

This adds calls to the YOG routine in physpkg in CAM, whilst also running ZM. Should close #4

Next steps:

jatkinson1000 commented 3 months ago

As discussed with @paogorman we would prefer to call the YOG scheme after deep convection. This is applied in https://github.com/m2lines/CAM-ML/pull/9/commits/5df40aa7ba38e3b3ddc6514f204668a705d7e1e6

MarionBWeinzierl commented 2 months ago

This adds calls to the YOG routine in physpkg in CAM, whilst also running ZM. Should close #4

* A call to initialise the net is made in `phys_init` - [19886bf](https://github.com/m2lines/CAM-ML/commit/19886bf8a8f824fdb2752bb17f19a590b5489cc6)

* A call to the net is made ~before~ after `convect_deep_tend` in `tphysbc` - [dc0a005](https://github.com/m2lines/CAM-ML/commit/dc0a005b229075deb3cb3139a721e9a5e35197d2) and [5df40aa](https://github.com/m2lines/CAM-ML/commit/5df40aa7ba38e3b3ddc6514f204668a705d7e1e6)

  * This also includes timing of the YOG scheme using the inbuilt CESM timing infrastructure

* Cleanup of the net and weights is made in [ee3005b](https://github.com/m2lines/CAM-ML/commit/ee3005be9331e9705422db3e86c359fe8a62cdc2)

Next steps:

* [ ]  Take the multiple call blocks into a more contained routines:

  * Net init and addfld into a yog_init routine
  * ptend_init, net call, outfld into a yog_main routine

* [ ]  Add the NN code currently held in SourceMods.

* [ ]  Update the xml as discussed with @paogorman to have a separate yog parameter, rather than assigning as an option to `deep_scheme`. This will be done in a refactor of [Xml updates for YOG #2](https://github.com/m2lines/CAM-ML/pull/2) which this should then be rebased off of.

Just to clarify, are the "next steps" something you are doing in this PR, or a new one? As it's a checklist for this one (which is not checked). If it's done in a different PR, it would be good to move the list there (or into the associated issue, that's a matter of preference), and link that from here.

jatkinson1000 commented 1 week ago

As commented on in #14 this PR was was superseded by that PR so I am closing this now.