amm-optimizer 0.2.0__tar.gz

amm_optimizer-0.2.0/.gitignore

@@ -0,0 +1,173 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/latest/usage/project/#working-with-version-control
+.pdm.toml
+.pdm-python
+.pdm-build/
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# COCO raw output and post-processing reports (regenerate with benchmarks/*.py)
+exdata/
+ppdata/
+
+# stray file written by COCO when problems are created
+._bbob_problem_best_parameter.txt
+
+# console logs of arm runs (run_arms.py, quick_scan): keep them out of commits
+*.stdout
+run.stdout

amm_optimizer-0.2.0/CHANGELOG.md

@@ -0,0 +1,117 @@
+# Changelog
+
+## Unreleased
+
+## 0.2.0 (release of the freeze)
+
+- Public release as `amm-optimizer` (import `amm`; the implementation keeps its code name `atm`). Packaging only: the code is that of tag `freeze-2026-10-22` (c46, config hash `2153948ebed51150`); added the `amm` alias package, `CITATION.cff`, and an AMM README. No algorithmic change.
+
+## Earlier (pre-release notes)
+
+- 2026-10-22, cycle 71 (all opt-in; `c46` unchanged):
+  - `Variant.stage_extend_max_per_dim`: `stage_extend_restart` only at budgets up to that many x D. Batch 89 rejected the uncapped option (losses at 1000 x D), Batch 92 accepted the option capped at 300 x D, and its confirmation on fresh ids 730-769 did not confirm (f8 + f9 +.007 [-.002, +.017]), so it is not in a preset.
+  - Batch 91 (`atm20s` on `c46`, mass check against four A/A draws) was rejected under the letter (a named depth exceeded by .001; lesson 54's addendum).
+  - `c46` is the configuration frozen for the paper (`docs/FREEZE_2026-10-22.md`, tag `freeze-2026-10-22`).
+- 2026-10-22, cycle 70 (the gaps4 toy, `docs/research/toys/gaps4.md`; all options opt-in, `c45` unchanged):
+  - `Variant.prefix_low_beta` / `prefix_low_beta_min_dim`: in the ES prefix, a finite rugged curvature reading below the threshold keeps the (1+1)-CMA-ES where the rugged kind would be lq2 populations (Batch 88).
+  - Not in a preset: `centre_point` / `centre_point_min_dim` (a centre-point evaluation; not screened, because its f19 gain scores bbob's construction of f19), `smooth_bfgs_second` (rejected in the toy).
+  - `atm.presets`: `c46` = `c45` + `prefix_low_beta=0.15` + `prefix_low_beta_min_dim=10`, confirmed on fresh ids 690-729 (f23 at D 10-40, 100 x D +.046 [+.040, +.051]); `c46` is the best composite and the configuration frozen for the paper. Batch 89 (`stage_extend_restart` on `c45`) was rejected.
+- 2026-10-22, cycles 67-69 (all options opt-in; `c43` and `c44` unchanged):
+  - `Variant.rugged_try_rest` (`"cutkink"`), `rugged_try_cap_per_dim`, `rugged_try_max_per_dim`: the rugged route's capped BFGS try continues to the end of the budget where its best point reads a kink (the f13 toy, `docs/research/toys/f13_cells.md`; Batch 79).
+  - `Variant.prefix_ridge_after` / `prefix_ridge_radius` / `prefix_ridge_max_per_dim`: a ridge reading of the gated prefix's SLSQP descent hands the rest of the prefix to restarted COBYQA (f13's ridge; Batch 79), with `prefix_nan_reading="plateau"` (Batch 76b).
+  - `Variant.prefix_descent_retries`: the one-descent prefixes retry a descent that did not move (the parked2 toy's `retry5`; Batch 83). Also from that toy, opt-in and not in a preset: `prefix_nan_min_per_dim`, `prefix_nan_clip`, `prefix_nan_reprobe`, `prefix_plateau_ties`, `prefix_fractal_min_dim` / `prefix_fractal_min_per_dim`, `stage_extend_restart`.
+  - `Variant.rugged_direct_beta` / `rugged_direct_min_beta`: on a rugged hand-over reading in that band, the finisher's populations start at once (the gaps3 toy; Batch 84). Also from gaps3, not in a preset: `sep_route_low` / `sep_route_low_dim` (Batches 85 and 85b rejected).
+  - `Variant.box_normalize`: `minimize` maps any box onto [-5, 5]^D and runs there (the novel-ideas toy, `docs/research/toys/novel_ideas.md`; bit-identical on bbob).
+  - From the 132-ideas stream, not in a preset: `sampled_polish_per_dim` (Batch 81 rejected) and `cma_handoff_cap_low_per_dim` / `cma_handoff_cap_low_max_per_dim` (Batches 81 and 81b rejected under the letter of their mass checks; parked).
+  - `atm.presets`: `c45` = `c44` + `rugged_try_rest="cutkink"` + `prefix_nan_reading="plateau"` + `prefix_ridge_after=7` + `prefix_ridge_max_per_dim=30` + `box_normalize` + `prefix_descent_retries=5` + `rugged_direct_beta=1.0` + `rugged_direct_min_beta=0.5`, confirmed on fresh ids 650-689 (f13 at D 5, 300 x D +.075; f13 at D 20-40, 30 x D +.095; f12-f14 at D 20-40, 100 x D +.045; f10-f12 at D 2-5, 30 x D +.047; all functions at D 5-10, 300 x D +.006); `c45` was the best composite until `c46`.
+- 2026-10-21, cycles 63-66:
+  - `Variant.small_switch` (and `small_switch_*`): at budgets up to 10 x D the prefix's first SLSQP descent is read at two checkpoints and hands the rest to restarted COBYQA (`docs/research/toys/small10.md`, Batch 77).
+  - `Variant.prefix_nan_reading`: where an undefined prefix curvature reading goes (opt-in; Batch 76 rejected, rescreen 76b).
+  - `atm.presets`: `c44` = `c43` + `prefix_plateau_pop` from D 5 (Batch 75) + `small_switch` (Batch 77), confirmed on fresh ids 610-649 (f7 at D 5-40, 100 x D +.079; all functions at D 5-20, 10 x D +.012); `c44` is the best composite.
+- 2026-10-21, cycle 61:
+  - `atm.cobyqa_phase.run_cobyqa_restarts`: restarted SciPy COBYQA in the box (the small-budget engines toy, `docs/research/toys/small_budget.md`).
+  - `Variant.small_engine` / `small_engine_max_dim` / `small_engine_max_per_dim` / `small_engine_radius`: the COBYQA route for the smallest budgets.
+  - `atm.presets`: `c43` = `c42` + `small_engine="cobyqa"` + `small_engine_max_dim=3`, confirmed on fresh ids 366-399 (D 2, 10 x D +.026; D 3 +.017); `c43` is the best composite.
+- 2026-10-21, cycle 60:
+  - `Variant.basin_descents` / `basin_descents_max_per_dim` / `basin_keep_atm`: on `basin_route`'s smooth, not-multi-basin reading at D 2-3, SLSQP descents from the design, then the usual path from the design's best point with ATM's hand-over share counted after them.
+  - `atm.presets`: `c42` = `c41` + `basin_descents=1` + `basin_keep_atm`, confirmed on the D 2 ids 400-449 (f6 / f8 / f9 / f12 / f13 at D 2-3, 100-300 x D +.103; all functions there +.016).
+- 2026-10-21, cycle 59:
+  - `qn_phase.run_scaled_bfgs(box_qp=)`: SLSQP's bound handling (the quadratic model's minimiser over the box, `_box_qp`) when -H g leaves the box; it repairs f10 (the BFGS engine toy).
+  - `QnRoute.smooth_bfgs_box_qp` / `smooth_bfgs_descents`; routes `qn43` (`qn40` + `box_qp`) and `qn44` (one `box_qp` descent, then the staged route).
+  - `atm.presets`: `c41` = `c40` + `cma_local_nonsep="qn44"`, confirmed on fresh ids 570-609 (f6 / f12 / f13 / f14 at D 5-40, 300 x D +.163; all functions +.033).
+- 2026-10-21, cycles 56-58:
+  - `Variant.des_max_per_dim` (and the research switch `des_skip_step`): the one-descent prefix of `es_kind="slsqplq10hd"` below that many x dim (default 30, `q7ed`'s envelope).
+  - `docs/research/parked/`: the parked-arms study (catalog of gains and losses, envelopes, mitigations).
+  - `atm.presets`: `c40` = `c39` + `es_kind="slsqplq10hd"` + `prefix_full_min_per_dim=0` + `des_max_per_dim=31`, confirmed on fresh ids 530-569 (D 5-10, 10 x D +.024; D 3-5, 30 x D +.050).
+- 2026-10-21, cycles 52-55:
+  - `atm.basin_phase`: a Sobol design's rank-quadratic fit and a curvature probe route smooth multi-basin readings (Gallagher) at D 2-3 to MLSL (`Variant.basin_route`, `basin_max_dim`, `basin_min_per_dim`, control `basin_control`).
+  - `Variant.sep_stall_min_per_dim`: the stall separability check from a budget on.
+  - `Variant.fork_seed` (research): independent continuations from one hand-over, for the learned-router dataset (`docs/research/router/`).
+  - `atm.presets`: `c39` = `c38` + `basin_route` + `sep_check="stall"` from 1000 x D, confirmed on fresh ids 490-529 (f21 + f22 at D 2-3 +.107; D 5, 3000 x D +.040).
+- 2026-10-20, cycles 50-51:
+  - `atm.chain_phase`: a pair test of additivity (3 x dim), a band check of the coupling's orientation and linearity (6 x dim), and a coordinate search along the directions that undo a triangular coupling (the f20 toy study).
+  - `Variant.chain_route` / `sep_route` (with `chain_min_per_dim`, `sep_route_min_per_dim`, `additive_min_dim` and the control `additive_control`).
+  - `step_phase.run_step_phase` keeps its samples in arrays (bit-exact, much faster at large budgets) and takes `check_tol`.
+  - `atm.presets`: `c38` = `c37` + `chain_route` + `sep_route`, confirmed on fresh ids 326-365 (f20 at 300-3000 x D +.773; f3 at 100 x D +.711).
+- 2026-10-20, cycles 41-48:
+  - `qn_phase.run_scaled_bfgs`: a restarted nonsmooth BFGS on forward differences along its inverse Hessian's eigenvectors, with Armijo and weak-Wolfe line searches (the local-engine toy study, idea #58).
+  - `cma_local_nonsep` routes `qn30`-`qn39`, with `QnRoute.kink_engine`, `rugged_try_*` and `rugged_ipop_low_*`. `qn38` = `qn29` + the BFGS kink route from D 5 + the capped rugged try at D 5-9 up to 500 x D + rugged IPOP from D 5 with 2000 x D left.
+  - Prefix options `Variant.prefix_plateau_pop`, `prefix_plateau_rugged` and `prefix_plateau_min_dim` (opt-in; claim P did not confirm), and the D 2 `travel_*` options (rejected, opt-in).
+  - `atm.presets`: `c37` = `c36` + `qn38`, confirmed on fresh ids 286-325 (f13 at D 10-40, 1000 x D +.304; f13 at D 5, 300 x D +.377; f15 at D 5-10, 3000 x D +.106).
+- 2026-10-20, cycles 39-40:
+  - `atm.presets`: `c36` = `c35` + `qn29` (`gal3c` + `fract2`) + `prefix_sep_probe` / `prefix_full_per_dim=100` / `prefix_full_min_per_dim=10`, confirmed on fresh ids 246-285 (all functions at 30-100 x D +.016; f23 at D 20-40 +.044).
+  - The kink route (`qn23` / `qn24`) stays opt-in after three unconfirmed fresh-id estimates.
+  - New opt-in options: `Variant.separable_local` (cycle 40, under screen) and the research switch `local_force`; `qn28` / `qn29` routes.
+- 2026-10-20, research cycles 35-38 (`RESULTS.md`, Batches 59-62 and the cycle 36 confirmation):
+  - `cma_local_nonsep` routes:
+    - `qn20` (`gal3c`: `run_sampled_restarts`, Gallagher restarts from the best unclaimed sample point with a gap-scaled stop);
+    - `qn21` / `qn26` (`fract` / `fract2`: `micro_slopes`, a fractal reading leaves the rugged route for restarted SLSQP; `qn26` centres the probe off the bounds);
+    - `qn23` / `qn24` (`ridge`: on a kink the cap gate refused, `kink_check` and SLSQP at a 1e-11 step, `qn24` capped at 500 x D);
+    - `qn25` (`qn24` + `gal3c`);
+    - `qn27` (`lam16`: the rugged route starts at 16x lambda; `pop_cma` rule `ipop16`).
+  - Prefix switches (cycle 38): `Variant.prefix_sep_probe`, `prefix_full_per_dim`, `prefix_full_min_per_dim` and `cheap_probe_per_dim`.
+  - `atm.presets`: `c35` = `c34` + `qn20`, confirmed on fresh ids 211-245 (f21 + f22 +.191); `c35` is the best composite. `fract` was withheld, and `ridge`'s fresh claim did not confirm (+.028, p .084).
+- 2026-10-19, research cycles 33-34 (`RESULTS.md`, Batch 58 and the `c34` confirmation):
+  - `cma_local_nonsep` routes `qn18` (`QnRoute.cap_gate`: at the stage cap the `StallMonitor` is read once; a progressing first descent continues to SciPy's own exit) and `qn19` (`qn18` plus `second_descent_min_left_per_dim`: one more descent from a uniform point after a gated converged one with at least 400 x D left). The rejected Batch 58 arms stay opt-in: `es_kind="slsqplq10hd"` (`q7ed`) and `nm_stage="qnprefix30gd"` (`leangd`).
+  - `atm.presets`: `c34` = `c33` + `qn19`, confirmed on fresh ids 196-210 (f8 + f9 at D 20-40 +.172; every cell +.015) and characterised; `c34` is the best composite.
+- 2026-10-19, research cycles 27-31 (`docs/research/audit_2026-10-17/RESULTS.md`, Batches 54-57):
+  - `Variant.qn_fallback="lbfgsb"` (`atm.qn_phase.run_lbfgsb_restarts`, `failure_majority`): an SLSQP call whose failed descents outnumber its converged ones hands its remaining evaluations to restarted L-BFGS-B (ftol = gtol = 0, 1e-9 step); a local stage that ends that way gives L-BFGS-B the rest of the budget. Only f12 shows the signature in `c31`'s own runs.
+  - `cma_local_nonsep` routes `qn15` (rugged readings with >= 640 x D left run unprotected IPOP from a uniform point; `pop_cma` rule `ipop8`) and `qn17` (the same from D 20 only, `QnRoute.rugged_ipop_min_dim`); `sep_check="stall"` (STEP's two-axis check once its best has not improved for two sweeps). `qn15` and `sep_check="stall"` were rejected as screened (Batch 55) and stay opt-in.
+  - `atm.presets`: `c32` = `c31` + `qn_fallback="lbfgsb"` (confirmed on fresh ids 176-185: f12 +.296, +.208 over pycma-IPOP) and `c33` = `c32` + `qn17` (confirmed on fresh ids 186-195: f17-f18 at D 20-40 +.171); `c33` is the best composite. Both are characterised bit for bit (`tests/characterisation.json`), with the two opt-in paths `qn15` and `sep_stall`.
+  - **f_opt repair (harness):** `quick_scan.raw_fopt_table` built `f_opt` as `f(0) - (f(0) - f_opt)` from a COCO data line with 10 significant digits, which put 291 f12 optima (D 2-40, mostly D 10-40, where f(0) is 1e8-1.6e9) on the wrong hundredth (by .01-.36; checked against f at COCO's x_opt). It now reads the exact value from the data header (`header_fopt`). `tools/repair_fopt_cache.py` repaired the cache and wrote `benchmarks/results/fopt_corrections_2026-10-19.json`; `load_run` re-expresses older files exactly and flags runs that stopped on a false final-target hit (`fopt_false_hit`), which `tools/rerun_false_hits.py` re-runs (deterministic per problem and seed) and rewrites on the new basis (`fopt_basis`). Every f12 score before this date is affected.
+  - `scipy>=1.16` is now a declared dependency (the research presets need SciPy).
+  - Research tooling: rule v5 (`screen_v5.py`, the `eb2` shrinkage gate of `cycle27_gates.py`), the posterior goal ledger (`goal_ledger.py`), `fresh_claims.py`, and the consolidated PDF builder (`docs/research/build_consolidated_pdf.py`).
+- 2026-10-19 integration of the research line (`atm-audit-experiments`, composites `c23`-`c31`), the code-review branch and `refactor` (`docs/MERGE_2026-10-19.md`). Every characterised run of the research tip reproduces bit for bit, except 14 round-1 arms that use a removed option.
+  - `refactor`'s `Variant.nm_max_dim` (the simplex limit of `local_method="nelder_mead"`) is now `simplex_max_dim`; `nm_max_dim` keeps the research meaning (SciPy Nelder-Mead in `atm.nm_phase`).
+  - The research line's `ATMConfig.improved()` (the Wave 5 preset) is now `ATMConfig.improved_wave5()`; `improved()` is `refactor`'s Tier 5 hybrid.
+  - `atm.presets` holds `c23`-`c31` (`c31` is the confirmed best composite).
+  - `quick_scan` has both `--year` and `--instance-ids`, and resumes journals across the merge's added and removed fields.
+  - `refine_fraction` and `cma_handoff` exclude each other.
+- `atm.presets`: the research composites `c23`-`c27` as named configurations (`get`, `names`, `with_changes`), equal to `benchmarks/arms/cNN_config.txt`; `quick_scan --config preset:NAME,...`.
+- Bit-exact characterisation of `c23`-`c27`, every `cma_local_nonsep` route and all recorded arm configs on bbob (`tests/test_characterisation.py`, `tools/record_characterisation.py`).
+- Phase routing refactored without behaviour change: `atm.qn_phase.QN_ROUTES`, `atm.api._prefix_kind`, and hand-over functions instead of the closure in `minimize` (`docs/CODE_REVIEW_2026-10-19.md`).
+- Faster full-quadratic model steps (cached upper-triangle indices, no `n x D x D` intermediate); bit-identical results; `c26` CPU time -9 to -24% at D 2-10.
+- Harness: `quick_scan --resume` survives new default `Variant` fields; atomic f_opt cache; `run_arms --retries`.
+- `ATMConfig.variant` (`atm.config.Variant`): opt-in algorithm variants E01-E09 of `docs/EXPERIMENTS.md`, every default equal to the original (bit-identical, tested); `ATMConfig.to_dict/from_dict/config_hash`; `atm.plan.resolve_config`; `minimize(bounds=...)`; `CountedObjective.evaluate_columns(known=)`. Listed in `docs/DEVIATIONS.md`.
+- `ATMConfig.improved()`: the best configuration of `docs/IMPROVEMENTS.md` (needs `bounds=`); `ATMConfig()` stays the original.
+- Wave 7 (multimodal regression found by the Tier 4 holdout): `Variant.es_small_dim` / `es_small_max_budget_per_dim` (the local prefix only up to 30 x D in D <= 5), `refine_max_dim` (no tail above 10 dimensions) and `es_init_share`; `ATMConfig.improved()` now uses the first two, `ATMConfig.improved_wave6()` is the previous preset (commit da1ef9a). Development result: no group loses against the plain composite at 30-300 x D, multimodal f15-19 at 100 x D +.006 / +.012 (Tier 1 / 2) instead of -.048 / -.035; `docs/EXPERIMENTS.md` section 9.7. Config hashes (the schema gained fields, so every hash changed, the original's too, with identical behaviour): `ATMConfig()` 59c51e668986d345, `improved()` a2f27a8ba7e2ed96, `improved_wave6()` 0188cb38b8d456bf (was a859d9d5071ea792 before the new fields).
+- Variants E12 `refine_fraction` (+ `refine_min_budget_per_dim`) and E13 `local_method="nelder_mead"` (+ `nm_max_dim`, `nm_refine_max_dim`): a late refinement of ATM's best point and a restarted bounded Nelder-Mead simplex search (`atm.simplex`, no SciPy) for the ES phase and the tail; `ATMConfig.improved()` uses both (Tier 1 primary +.243, was +.163; see `docs/EXPERIMENTS.md` section 9.6).
+- `benchmarks/breakdown.py` (group and dimension tables), Wave 6 results under `benchmarks/results/scan/wave6/`.
+- Variant E15 `covariance="rank_mu"` (+ `cov_rate`, `cov_select`, `cov_cap`, `cov_max_dim`): CMA-style full covariance for the random search, D <= 10; with the other variants it closes half of the gap to CMA-ES at 1000 x D.
+- Variant E14 `scale_learning="diagonal"`: per-coordinate widths learned from accepted moves (f2 +.21 at 1000 x D, opt-in).
+- Variant E21 `es_phase="prefix"` (+ `es_fraction`, `es_init_factor`, `es_max_budget_per_dim`): a (1+1)-ES for the first part of small budgets, then ATM; with the Wave 1 variants it reaches the (1+1)-ES at <= 100 x D and +.17 over the original at 1000 x D.
+- Variants E18 `allocation="credit"`, E19 `early_accept` and `es_init_share` were added and then removed again (no aggregate gain; see `docs/DEVIATIONS.md` and `docs/EXPERIMENTS.md` section 9.2).
+- `tests/golden/presets.json` pins the cost traces of the default, composite, Wave 6 and v2 presets.
+- Scoring fix: exact bbob optima (the cached values floored converged runs at 2e-8); old results are rescored on loading; raw COCO evaluation (`--penalty raw`) and the 2019 start protocol (`--start coco2019`); `benchmarks/bbob_compare.py` (cocopp plots), `legacy_vs_reference.py`, `docs/IMPROVEMENTS.md`, `docs/research/gecco2019/`.
+- `benchmarks/quick_scan.py` v2 (stable seeds, provenance, process pool, resume, cluster bootstrap), `run_arms.py`, `summarize_arms.py`.
+- CI workflow (3 OSes, 3 Python versions, oldest NumPy, COCO parity smoke job).
+- `py.typed` marker.
+- COCO/bbob benchmark harness (`benchmarks/run_bbob.py`, `compare_bbob.py`, `postprocess_bbob.py`), see `docs/BENCHMARKS.md`.
+- Faster parent selection (`best_eighth` sorts once, only the parent kind that is read is selected); bit-identical results.
+- RNG draw ledger test and `atm.legacy_stream` (all draws kept only for stream alignment in one module).
+- Block profiling (`atm.profiling`), see `docs/PERF.md`.
+
+## 0.2.0.dev0
+
+- Clean re-implementation of ATM as the `atm` package; bit-identical to `BjBodner/atm@69828c6` with `ATMConfig(strict_budget=False)`.
+- Strict budget by default, NaN/inf handling, conductor robustness, opt-in correct second best. All differences are listed in `docs/DEVIATIONS.md`.

amm_optimizer-0.2.0/CITATION.cff

@@ -0,0 +1,15 @@
+cff-version: 1.2.0
+message: "If you use this software, please cite it as below (the paper's reference will be added when it is public)."
+title: "AMM: budget-aware routing of classical solvers by online probes"
+version: 0.2.0
+license: Apache-2.0
+type: software
+repository-code: "https://github.com/BjBodner/amm-optimizer"
+authors:
+  - family-names: Bodner
+    given-names: Benjy
+keywords:
+  - black-box optimization
+  - derivative-free optimization
+  - COCO
+  - BBOB

amm_optimizer-0.2.0/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.