Hi,
just heard the great talk by @liellnima at the ML4ESM workshop, and was super excited to see that xMIP is used!
I wanted to reach out and see if there is any way I can help integrate xMIP better with ClimateSet.
I particular I noted this in the paper
The structure processor was implemented with xmip (Busecke and Abernathey, 2020)
and follows their guidelines. Note that it is significantly slower than the CDO-implemented modules.
I am very curious to learn what causes the performance difference. Most of the corrections of xmip are applied lazily and should be quite cheap! If you have a minimal example that we could work through, it would be great to post an issue on the xmip issue tracker.
A small comment on the citation: It would be tremendous if you could update the citation to the actual software package doi:10.5281/zenodo.7519179 in the published paper. That would help me a lot in tracking useage and justify working on this software.
Again many thanks for using xMIP and please feel free to reach out with any questions/comments.
Hi,
just heard the great talk by @liellnima at the ML4ESM workshop, and was super excited to see that xMIP is used!
I wanted to reach out and see if there is any way I can help integrate xMIP better with ClimateSet.
I particular I noted this in the paper
I am very curious to learn what causes the performance difference. Most of the corrections of xmip are applied lazily and should be quite cheap! If you have a minimal example that we could work through, it would be great to post an issue on the xmip issue tracker.
A small comment on the citation: It would be tremendous if you could update the citation to the actual software package doi:10.5281/zenodo.7519179 in the published paper. That would help me a lot in tracking useage and justify working on this software.
Again many thanks for using xMIP and please feel free to reach out with any questions/comments.