Changes to inference_medium_range.py

[pgarg7]

Earth-2 MIP Pull Request

Description

I have changed the run_forecast function within the inference_medium_range.py module in order to make sure that deterministic scores are only calculated for variables which are same in model's input and output channels. I added some code in which it finds the indexes of model's in_channels in model's out_channels and then calls the metric on only those indexes. In this way, it's not raising an error related to input/output shape mismatch specially for Graphcast_Operational model. -- This PR "closes #156"

Checklist

• [x] I am familiar with the Contributing Guidelines.
• [ ] New or existing tests cover these changes.
• [ ] The documentation is up to date with these changes.
• [ ] The CHANGELOG.md is up to date with these changes.
• [ ] An issue is linked to this pull request.

Dependencies

[NickGeneva]

/blossom-ci

[NickGeneva]

Hi @pgarg7

Appreciate the bug report and the PR. I'll take an additional look at this today and hopefully run some tests. Left some initial comments but no rush.

[nbren12]

I think some of this is fixed in #154.

[nbren12]

154 fixed this bug, so this PR is no longer needed. Graphcast has an input tp06 that is not available in most data sources. There is an experimental PR (#94) which adds it cds.DataSource, but I didn't think that implementation was robust enough to merge yet.