airmosaic-ai

Using an external CHEER GAMS release

English 中文

The CHEER downloads page offers CHEER demo in Python and GAMS and CHEER-P demo in GAMS. Its GAMS versions require a GAMS license. The model archive, GDX databases, GAMS installation and license remain external to this repository. The Python CGE tutorial in this skill is an independent synthetic model; it does not execute CHEER.

Inspect and reproduce an obtained release

  1. Record the official download page, archive filename, download date, release/base-year statements and archive hash. Keep an untouched copy and work from a separate external copy. Inspect files before running them.
  2. Identify the actual entry point; run.gms was the entry in the previously inspected demos, but verify the obtained archive. Read enabled $include, $batinclude, $gdxin and model/solve statements. GAMS include resolution depends on the working directory and InputDir. Check file case on case-sensitive hosts and locate every referenced GDX file before execution.
  3. Check gams -v, an available model-compatible solver and the GAMS license. If any are missing, report source preflight only. Do not call a parsed source tree a numerical reproduction.
  4. Run the unchanged benchmark first from the verified model work directory. If its confirmed entry is run.gms, gams run.gms lo=2 is an example. Record the command, listing (.lst), solver/model status and output files. Compare the solved benchmark with the supplied SAM and published reference totals before editing scenarios.
  5. Edit only the external working copy. For a new shock, locate the parameter definition and active scenario assignment; change the intended value or include rather than a report variable. Keep benchmark and scenario copies separate, use the same closure and numeraire, and document affected region, sector, year and units. For an MPSGE release, its official syntax guide explains $SECTORS, $COMMODITIES, $CONSUMERS, $PROD, $DEMAND and model generation. Changes to structural equations require recalibration and benchmark checks.
  6. Re-run, inspect listing and numeric residuals, and compare paired benchmark/scenario prices, quantities, income and environmental outputs. A successful solver status without benchmark replication is insufficient. Keep results, logs and private data outside Git.

In the previously reviewed CHEER-P demo, sets.gms defined indices, cal.gms handled calibration, elasticities.gms held assumed elasticities, emission.gms contained environmental quantities, core.gms contained equilibrium logic, and run.gms orchestrated the run. Treat these names as search starting points: trace the enabled include graph in the downloaded release before editing. For a sector expansion, change sets, SAM mapping, parameter dimensions and every affected equation together; for a parameter scenario, change the active assignment after baseline replication; for an environmental satellite, verify whether it enters an equation or only a report.

The previously inspected CHEER archives had some include-path case mismatches. This is a release-specific preflight observation, not a claim about all versions or a license to alter source files automatically.

This site is open source. Improve this page.