"Development model output" mode?

[k-a-webb]

It would be great to run REF (locally) for development model runs, for which the output may not (yet) be fully compliant with the CMIP6 CMOR standards, or include all the necessary metadata.

I attempted to ingest CMIP7 CMORized output for a development model run, but ran into an issue with the mandatory requirement for licence_id in the cmip7_parsers (https://github.com/Climate-REF/climate-ref/blob/main/packages/climate-ref/src/climate_ref/datasets/cmip7_parsers.py#L165).
While I could get around this by removing the mandatory requirement, I wonder if there's been a "development model output" mode of the REF that I should be using?
I expect I'll also run into issues with the constraints of the diagnostics (e.g., timeranges).

I'll try to add to this issue other issues I run into as I attempt to work with development models.

[lewisjared]

We implemented the CMIP7 support to follow the CMIP7 global attributes as closely as we could. We haven't implemented a looser set of requirement for semi-CMORised outputs to date.

I'm happy to take a PR which to add support to opt out of the mandatory attributes with sane defaults for some attributes. This should be gated behind an environment variable or config.

What time-period do you have currently available?

[k-a-webb]

The first iteration of model runs are single-year. I would expect this could lead to issues with the diagnostic configurations as well, if there are hardcoded timeranges. At this stage, being able to run any diagnostics (e.g., timeseries) would be helpful. Closer to the CMIP7 deadline we would want to test all diagnostics.