Fix the pre-existing astropy/pandas/shapely benchmark workflows - #7

Merged

atharvas merged 17 commits into

mainfrom

fix-benchmark-workflows

May 18, 2026

Commits on May 15, 2026

Commits on May 16, 2026

cap per-benchmark timeout at 5s for the large astropy/pandas shards
atharvas
committed
93af712

Commits on May 17, 2026

extract value-serialization helpers to module level in storage
atharvas
committed
838e415
add opt-in --parallel process-pool execution for capture/verify/baseline
atharvas
committed
cd2a291
run astropy/pandas roundtrip shards with --parallel
atharvas
committed
f056d40
ignore .claude/ and astropy .fits benchmark artifacts
atharvas
committed
6b50f32
recover serially when a parallel worker dies instead of failing the batch
atharvas
committed
f6ac822
tolerate unpicklable benchmark params instead of aborting capture
atharvas
committed
c3ef253
preserve param_names list type in unpicklable-param fallback
atharvas
committed
27a0569
compare pandas DataFrame/Series with tolerance instead of failing
atharvas
committed
b5d4928
make baseline-to-verify gate robust to non-deterministic benchmarks
atharvas
committed
f0eed63
report regressed benchmark names in verify summary
atharvas
committed
35e5a5e

Commits on May 18, 2026

drop stability-aware baseline; it doubled baseline time on large shards
atharvas
committed
aa49abd