test: direct unit suites for opt Hparam and BaseSearcher (78 tests)

[arham766]

What does this PR do?

Type of change: new tests

Part of the unit-coverage initiative in #1902. Direct suites for the two search-framework primitives (no direct coverage existed; the duplication audit against test_dynamic.py, tests/unit/torch/nas, and test_autoquant.py is in the review notes):

• test_hparam.py (44 tests) — construction/dedup/original folding, configurability incl. the _force_configurable context manager, active-value validation with exact error messages, active_slice/enforce_order LongTensor semantics, choices-setter subset rules, importance registration (exact max-normalization values, multi-estimator summation, unsortable rejection), iand merging, and pickle round-trip (required for Megatron distributed broadcast per the source NOTE).
• test_searcher.py (34 tests) — default-config contract, sanitize merge rules, the search() hook order and training-mode restoration, deepcopied default state, checkpoint path-resolution table (file/dir/extensionless/pathlib), strict vs non-strict state loading, end-to-end resume across two search() calls, scoring, and forward-loop construction with exact iteration counts.

Usage

N/A — tests only.

Testing

78 tests, <1s, hermetic (CPU-only, tmp_path, deterministic). Adversarially reviewed: 12 expectations independently re-derived from source, 6/6 seeded mutations killed, no vacuous tests, Windows/Linux path-agnostic. Full tests/unit/torch/opt dir passes alongside (excluding the pre-existing test_lr_config.py env failures).

Before your PR is "Ready for review"

• Is this change backward compatible?: ✅
• If you copied code from any other sources or added a new PIP dependency, did you follow guidance in CONTRIBUTING.md: N/A
• Did you write any new necessary tests?: ✅
• Did you update Changelog?: N/A
• Did you get Claude approval on this PR?: N/A (external contributor)

Additional Information

Issue: #1902

Summary by CodeRabbit

• Tests
  • Added comprehensive unit coverage for Hparam, including defaults, configurability, active/choices updates, ordering, importance handling, merging, and pickle roundtrips.
  • Added extensive BaseSearcher tests covering search flow, checkpoint save/load, resume behavior, scoring, and forward-loop/data-loader execution rules.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: f6bc7aee-5ec8-429e-b73f-78a8d29ceaba

📥 Commits

Reviewing files that changed from the base of the PR and between b96a785 and 0935013.

📒 Files selected for processing (2)

• tests/unit/torch/opt/test_hparam.py
• tests/unit/torch/opt/test_searcher.py

---

📝 Walkthrough

Walkthrough

Adds two new unit test modules: one exhaustively testing the Hparam primitive (construction, configurability, active/choices setters, order enforcement, importance/sortability, __iand__ merging, pickle roundtrip), and another testing BaseSearcher (config sanitization, search workflow, checkpointing, scoring, forward-loop construction). No public entities are modified.

Changes

Hparam Unit Tests

Layer / File(s)	Summary
Construction, properties, and configurability tests/unit/torch/opt/test_hparam.py	Tests cover module setup, construction semantics (original/active defaults, choices folding/sorting, min/max, iteration, repr), and configurability rules including manual override and the _force_configurable() context manager.
Active setter, active_slice, and choices setter tests/unit/torch/opt/test_hparam.py	Tests validate the active setter, active_slice/enforce_order behavior (permutation conversion/validation, reset), and the choices setter (subset shrinking, required membership, rejection of new values).
Importance/sortability and __iand__ merging tests/unit/torch/opt/test_hparam.py	Tests cover importance registration/normalization, order mode, sortability requirements, __iand__ merge semantics, and a pickle roundtrip preserving full state.

BaseSearcher Unit Tests

Layer / File(s)	Summary
Test fixtures, config sanitization, and search() workflow tests/unit/torch/opt/test_searcher.py	Adds concrete BaseSearcher subclasses/helpers and tests config defaulting/rejection, hook execution order, training-mode restoration, state reset/deep-copy, and state_dict() reflection.
Checkpoint path resolution and persistence tests/unit/torch/opt/test_searcher.py	Tests checkpoint path resolution, save/load roundtrips, directory creation, strict vs non-strict key handling, and resuming a search from a saved checkpoint.
Scoring and forward-loop construction tests/unit/torch/opt/test_searcher.py	Tests eval_score() requirements/behavior with the silent flag, and construct_forward_loop() rules including exclusivity constraints and max_iter_data_loader handling.

Estimated code review effort: 2 (Simple) | ~15 minutes

Related Issues: None found.

Related PRs: None found.

Suggested labels: tests

Suggested reviewers: None specified.

🚥 Pre-merge checks | ✅ 6

✅ Passed checks (6 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly describes the main change: adding direct unit test suites for Hparam and BaseSearcher.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Security Anti-Patterns	✅ Passed	Only two test files changed; no new modelopt/examples code or listed security anti-patterns (# nosec, unsafe loads, trust_remote_code=True, eval/exec) were found.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}

[arham766]

Consolidated into #1927 per maintainer feedback in #1902 — the suite was trimmed to only the lines codecov reports uncovered, with parametrization clusters deduplicated. Closing in favor of that PR.