Benchmark profiles, CLI entrypoints, gaming detectors, and the benchmark runner surface.
Package source: tabnetics.benchmarks
Package overview
Benchmark runner and profile surface backing tabnetics-benchmark / python -m tabnetics.benchmarks.cli, exposing the profile registry for systematic paired comparisons and the runner that enforces the validation-catalog data policy: evidence-bearing runs use the HuggingFace mirror of public upstream sources and do not silently fall back to synthetic proxies.
Stable exports
FS_METHOD_SETS(constant) - Source. Module-level constant exported by the package surface.
Related modules
tabnetics.benchmarks.cli- Source. Installed benchmark CLI entrypoint backingtabnetics-benchmark/python -m tabnetics.benchmarks.cli; the packaged surface defaults todf_stage_position="after_fs"and enforces the validation-catalog data policy: the HuggingFace bundle is the operational mirror of public upstream sources for evidence-bearing runs, and synthetic fallback is forbidden there.tabnetics.benchmarks.gaming_detectors- Source. Read-only anti-gaming diagnostics for benchmark result analysis.tabnetics.benchmarks.profiles- Source. Benchmark method-set profiles.tabnetics.benchmarks.runner- Source. Run integrated DF+FS benchmarking with SOTA comparison and ablations.
Module details
tabnetics.benchmarks.__init__
Benchmark runner and profile surface backing tabnetics-benchmark / python -m tabnetics.benchmarks.cli, exposing the profile registry for systematic paired comparisons and the runner that enforces the validation-catalog data policy: evidence-bearing runs use the HuggingFace mirror of public upstream sources and do not silently fall back to synthetic proxies.
No top-level public symbols are exported directly from this module.
Documentation and webpages on this site are generated from authoritative internal sources using a combination of deterministic rules and generative AI. Errors are possible. Please report issues via GitHub Discussions or email [email protected].