The Dataset Viewer has been disabled on this dataset.
PerfAgent test_db
Per-instance stable-test-suite artifacts for the PerfAgent harness. The harness reads these files to help decide which tests are used to validate the agent's submitted patch on the GSO and SWE-fficiency-Lite benchmarks.
Layout
gso/<instance_id>/stable_suite.tar.gz # 102 instances
swefficiency/<instance_id>/stable_suite.tar.gz # 100 instances
swefficiency/<instance_id>/.testmondata # 100 instances, pytest-testmon seed database
Each stable_suite.tar.gz unpacks to stable_suite/ and contains:
| Member | Read by the harness | Purpose |
|---|---|---|
excluded_collectors.txt |
yes, on the host | Test files that fail to collect. Turned into --ignore= pytest arguments. |
excluded_nodeids.txt |
yes, in the container | Flaky or unstable test node IDs. Deselected through EXCLUDED_NODEIDS_FILE. |
included_nodeids.txt |
no | The stable node IDs that survived calibration, kept for reference. |
manifest.json, run_summaries.json |
no | Calibration metadata: pytest args, pass counts per run. |
.testmondata is a pytest-testmon SQLite database generated by pytest-testmon. The harness copies it to /.testmondata in the docker container so that selective test runs
only execute tests that are related to the agent's patch. GSO does not need a seed as it rebuilds the testmon database in the container.
Sizes (download):
| Benchmark | Instances | stable_suite.tar.gz |
.testmondata |
Total |
|---|---|---|---|---|
| GSO | 102 | 29 MB | not needed | 29 MB |
| SWE-fficiency-Lite | 100 | 54 MB | 14.5 GB (largest 304 MB) | 14.5 GB |
Fetch only one benchmark with --include "gso/*" or --include "swefficiency/*".
- Downloads last month
- 22