Add CML GitHub Action

[gchhablani]

This PR is a work-in-progress. Current action items include:

• Use the container from moja global's repo directly instead of pulling and creating a container.
• Look into self-hosted runners on CML. Current CML run takes 2+ hrs.
• Run post-process scripts and get results in the report.

[aornugent]

@HarshCasper - I think we can add the FLINTcloud containers here. Docs on Running Jobs in a Container

-runs-on: [ubuntu-latest]
+container: ghcr.io/moja-global/rest_api_gcbm:master

[HarshCasper]

Agreed.

[gchhablani]

Yes, I will add the container for the workflow.

[aornugent]

I've been tinkering with the post-processing tools to get some nice summaries. We can add this step to our CML Action:

# from the gcbm_project/ folder
python3 ../tools/CompileGCBMResults/compileresults.py sqlite:///output/gcbm_output.db --output_db sqlite:///output/compiled_gcbm_output.db

Then we want to pull out annual summaries. Here's an example that returns the total forest biomass in megatonnes:

import sqlite3
import pandas

conn = sqlite3.connect("output/compiled_gcbm_output.db")

query = f"""
        SELECT years.year, COALESCE(SUM(i.pool_tc), 0) / 1e6 AS total_biomass_mt
        FROM (SELECT DISTINCT year FROM v_age_indicators ORDER BY year) AS years
        LEFT JOIN v_pool_indicators i
            ON years.year = i.year
        WHERE i.indicator = 'Total Biomass'
            AND (years.year BETWEEN 1920 AND 2020)
        GROUP BY years.year
        ORDER BY years.year
        """

df = pd.read_sql_query(query, conn)
ax = df.plot.line("year") # x-axis
ax.figure.savefig('output/total_biomass_mt.png', dpi=300)

@yodavo might have some good suggestions for relevant results summaries!