test(rewards): extract reward-agnostic contract tests and scope fixtures

[DorisMai]

The older PR #272 is broken into 3 smaller PRs for (1) improve synthetic data generation, (2) refactor reward tests, and (3) add structure factor reward. This PR addresses (2).

Changes relative to main branch

1. Extract reward-agnostic contract tests into a shared test_reward_function_contract.py from test_real_space_density_reward.py.
2. Move reward fixtures from tests/conftest.py into a session scoped tests/rewards/conftest.py.
3. Add tests/rewards/reward_input_helpers.py for shared reward-input construction.

Changes relative to the previous PR #272

No changes were requested or comments made for these files in the previous PR, except for the suggestion on reward fixture scope.

Summary by CodeRabbit

• Tests
  • Added GPU-marked, reward-agnostic contract tests covering protocol conformance, determinism, output correctness, ranking behavior, gradient flow, batching, loss optimization improvement, and numerical stability.
  • Refocused real-space density reward tests to edge cases and input-to-reward-input validation, removing substantial prior coverage.
  • Improved reward test infrastructure with suite-scoped fixtures for parsing the CIF, building the reward function, and extracting coordinates, plus shared helpers for scattering indices and reward inputs.
  • Removed overlapping session-scoped fixtures from the broader test setup.

[coderabbitai]

Warning

Review limit reached

@DorisMai, you've reached your PR review limit, so we couldn't start this review.

Next review available in: 54 minutes

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: c4793376-6c11-4d68-9c23-7e72cc9b69da

📥 Commits

Reviewing files that changed from the base of the PR and between f399ffe and 4ac7581.

📒 Files selected for processing (1)

• tests/rewards/test_real_space_density_reward.py

📝 Walkthrough

Walkthrough

Changes

Reward fixtures and input construction are scoped to tests/rewards/. Generic reward protocol, correlation, gradient, batching, and stability tests are centralized, while test_real_space_density_reward.py retains real-space-specific coverage.

Reward test organization

Layer / File(s)	Summary
Reward fixtures and input construction tests/conftest.py, tests/rewards/conftest.py, tests/rewards/reward_input_helpers.py	Moves 1vme reward fixtures into the rewards test scope and adds helpers for scattering indices, B factors, and occupancies.
Shared reward contract coverage tests/rewards/test_reward_function_contract.py	Adds parameterized contract tests covering protocol behavior, loss ordering, gradients, optimization, batching, and numerical stability.
Real-space-specific test coverage tests/rewards/test_real_space_density_reward.py	Removes duplicated generic tests and retains real-space edge cases plus structure-to-reward input validation.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related PRs

• diff-use/sampleworks#333: Updates the same 1vme structure-density fixture path while this PR relocates its reward fixtures into tests/rewards/.

Suggested reviewers: xraymemory, marcuscollins, abdelsalam-abbas, k-chrispens

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately summarizes the main refactor: extracting contract tests and moving fixtures into a scoped rewards conftest.
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.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch dm/refactor-reward-tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[k-chrispens]

Generally looks good to me, I think we should have it somewhere on the roadmap to add some better tests here, but not quite sure where that fits in these days with all the loss function talk we had back in June!