GenomeDK standalone validation
Historical v18 installation
The following historical record applies to the archive and greml command named below. The current standalone command is gsuite. Renaming the frontend does not requalify or modify that archive; a new Linux bundle must run its installer checks under the new command name. No legacy command alias is provided.
On 18 October 2026, the operator reported successful binary production and fresh-runtime installer validation on GenomeDK. The producer completed with accepted: true, exit code zero, and native_sources_included: false. An independent sha256sum -c invocation in the supplied terminal output reported OK for the resulting archive.
| Item | Recorded value |
|---|---|
| Binary archive | greml-genomedk-20261018-104107.tar.gz |
| Archive SHA-256 | b69e70295d51610c1a23d149b93d9f6eab02de518d61a6a64baad3351bd971a5 |
| Tools kit | greml-binary-tools-20261018-v3.tar.gz |
| Tools-kit SHA-256 | 95c253639625600cf2e6523362b692f80852c999dfc532bc6046bc229ae79012 |
| Native source bundle | single-step-yaml-genomedk-20261017-v18 |
| YAML frontend revision | 6ac66ac9daacaefed1f6b09ce80000663af4103d |
| gsolve revision | 79c57755b2dd7ba5acc21c371924b703b05c8a2d |
| greml revision | 84c9b0cd3f2142b9c60222520d5741260e18265e |
| gbits revision | 5778df80a26123afe594818cf8913a1aec14dfb2 |
| gmat revision | 5042ffc505b5a77fdf92677f197217d64ca3cb8d |
| Producer fixes retained in gsuite | 4cfae42e7fb6d0bc8b29b4a4eec4aef119cf2e55 |
The matched v18 source manifest identifies the numerical/frontend revisions. The later producer fixes normalize checksum line endings and separate the external-data production recipe from the tiny example catalog. They leave numerical equations, YAML syntax and the v18 native installation unchanged.
Historical installer checks
The v18 producer admitted its archive only after its installer created a fresh minimal Conda runtime and completed:
- Host, payload-integrity, private-library resolution and BLAS/linking checks.
- The configured campaign of 59 small reference checks and 38 model examples.
- An installed
greml runtest on the small CSV/PLINK example. - A small CHOLMOD/PARDISO coefficient-agreement comparison.
- Result collection and final payload verification.
The public acceptance record is based on the operator-supplied terminal output. Detailed per-case counts, residuals, backend differences, package URLs and resource reports remain in the installation receipts on GenomeDK; those files have not been supplied for this documentation update. No per-case numerical values or timings are inferred from the final acceptance message.
Keep the following files with the installation:
installation.json
runtime-explicit.txt
blas-abi.json
native-linking.json
validation/exact/validation.json
validation/exact/summary.csv
validation/runner/execution.json
validation/backends/agreement.jsonScope and evidence
This establishes a successful GenomeDK binary installation with a fresh runtime and the configured small model checks. It does not establish universal Linux portability, industrial throughput, statistical calibration, or PEV precision. The installer checks architecture and the recorded glibc requirement on each recipient host. Numerical validation continues to run on installation.
In v18, the production recipe for the external 1M dataset was retained under recipes/. It was not an installation test, and the dataset was not included. The 42 tiny YAML files were retained in models/; the model-only end-to-end campaign ran 38 and excluded four uncertainty examples. No numerical reference was skipped to address the missing external dataset.
Binary artifacts remain private validation artifacts under the existing distribution policy. This documentation publishes instructions and aggregate acceptance/provenance only; it does not publish binaries, private paths, source archives, input records or detailed HPC logs. Website publication is separate from software release approval.
See the standalone user guide for installation and model use, and the general qualification limits.