Baseline lock and regression net¶
Captured on 2026-06-14 with the unmodified master tree.
- Test target:
ctest -R golden_baseline --output-on-failure - Full-suite target:
docker compose run --rm dev ctest --output-on-failure - Benchmark snapshot will be recorded here once a golden benchmark manifest exists.
Captured test list (U0)¶
| CTest Name | Status |
|---|---|
sanity.* |
passed |
elm_cpu.* |
passed |
os_elm.* |
passed |
os_celm.* |
passed |
h_os_elm.* |
passed |
rbf_features.* |
passed |
demo_api.* |
passed |
cuda_infra.* |
skipped (CPU-only) |
elm_gpu.* |
skipped (CPU-only) |
Golden baseline expectations¶
BatchElmwith fixed weights/biases and deterministic input targets reproduces a bounded prediction MSE on training data.OsElmafter initialization reproduces bounded prediction MSE on the same data.- Online
OsElm::updateimproves fit on held-out data within the same bounds. - The golden test file is the regression anchor for U1 onward. Any later step must keep these three cases green.
Regression net¶
Use this document to record when a refactor changes numeric behavior. Updates should include the step that triggered the change, a brief justification, and a new threshold if the old one is no longer appropriate.