Module lab 

Inference superiority lab — instruments for evidence-based methodology.

Plan: docs/plans/inference-superiority-lab-and-toolset-plan.md

Native-only. No Ollama product dependency; optional external A/B via CLI script.

Re-exports

pub use ablate::run_ablation_matrix;

pub use ablate::AblationRow;

pub use audit_path::audit_hot_path;

pub use audit_path::HotPathAudit;

pub use auto_improve::default_search_space;

pub use auto_improve::format_lockin_summary;

pub use auto_improve::run_auto_improve;

pub use auto_improve::AutoImproveConfig;

pub use auto_improve::LabConfig;

pub use auto_improve::LockInPackage;

pub use auto_improve::TrialResult;

pub use device_roof::calibrate_device_roof;

pub use device_roof::DeviceRoof;

pub use experiment_log::append_run_csv;

pub use experiment_log::ExperimentRun;

pub use experiment_log::CSV_HEADER;

pub use micro::run_q4k_soa_microbench;

pub use micro::MicrobenchResult;

pub use timeline::run_decode_timeline;

pub use timeline::DecodeTimeline;

pub use campaign::default_toggle_space;

pub use campaign::run_optimization_campaign;

pub use campaign::save_campaign_report;

pub use campaign::CampaignConfig;

pub use campaign::CampaignReport;

pub use config_space::Configuration;

pub use config_space::ConfigurationSpace;

pub use config_space::ParameterDef;

pub use config_space::ParameterValue;

pub use experiment::append_experiment_jsonl;

pub use experiment::load_experiment_log;

pub use experiment::run_experiment;

pub use experiment::run_experiment_with_quality;

pub use experiment::BenchResultSerde;

pub use experiment::ExperimentConfig;

pub use experiment::ExperimentResult;

pub use experiment::PhaseSnapshotSerde;

pub use experiment::QualityScore;

pub use experiment::ThermalSnapshot;

pub use hypothesis::evaluate_verdict;

pub use hypothesis::BeliefGraph;

pub use hypothesis::ExperimentVerdict;

pub use hypothesis::Hypothesis;

pub use pareto::ApplicationProfileWeight;

pub use pareto::ParetoFrontier;

pub use pareto::ParetoPoint;

pub use search::expected_improvement;

pub use search::KnnSurrogate;

pub use search::SearchEngine;

pub use search::SobolSequence;

pub use search::TrackAndStopBandit;

Modules

ablate

Ablation matrix: toggle axes × decode-proxy tok/s → experiment CSV.

audit_path

Hot-path wiring auditor — makes unfinished integration visible.

auto_improve

Autonomous multi-hour lab loop: measure → search configs → re-measure → lock-in.

campaign

End-to-end optimization campaign orchestration.

config_space

Typed configuration space for the optimization lab.

device_roof

Device peak calibration — real roofline anchors (plan L1.2).

experiment

Experiment execution engine — wraps the existing run_bench harness with quality verification, thermal sampling, and structured result collection.

experiment_log

Append-only experiment CSV (plan §5.1).

hypothesis

Hypothesis backlog and typed belief graph with confidence cascade.

micro

Isolated Q4_K SoA GEMV microbench (CPU oracle + optional CUDA).

pareto

6-dimensional Pareto frontier computation for experiment scoring.

search

Search and optimization engine: Sobol quasi-random initial design, a k-NN surrogate with Expected Improvement acquisition (Bayesian-style), and a Track-and-Stop multi-armed bandit for A/B testing.

timeline

Decode timeline: host phase counters + optional GPU profile + tok/s.