fix: address third-party deprecation warnings

[zywind]

Summary

• Replace GLiNER.batch_predict_entities (deprecated) with .inference
• Rename warmup_ratio → warmup_steps in TrainingHyperparams to match the transformers API; warmup_ratio is kept as a deprecated alias that emits a DeprecationWarning and copies its value to warmup_steps; floats in (0, 1) are still interpreted as ratios by TrainingArguments, so 0.05 default is unchanged in behaviour
• Suppress huggingface_hub resume_download UserWarning at the GLiNER load site (upstream gliner passes the deprecated arg)
• Suppress scipy ks_2samp asymptotic fallback RuntimeWarning in the tutorial notebook via a scoped with warnings.catch_warnings() block around builder.run()
• Pre-compile torchao and range_regex in the brev setup script so their one-time SyntaxWarnings go to the setup log instead of notebook output
• Pin BNB_CUDA_VERSION=128 in the brev kernelspec — bitsandbytes skips odd CUDA releases so 12.9 has no native binary and always falls back to 12.8; update to 130 when upgrading CUDA_EXTRA to cu130+

Test plan

• Unit tests pass (mise run test)
• Re-run the tutorial notebook and confirm no red warning highlights in JupyterLab

Summary by CodeRabbit

• New Features

  • Training configuration now supports warmup_steps as a fixed count or fraction of total training steps, while maintaining compatibility with the deprecated setting.
  • GLiNER inference now supports updated APIs with fallback handling for older versions.

• Bug Fixes

  • Reduced expected runtime and deprecation warnings during training, model loading, and tutorial execution.
  • Improved CUDA 12.9 compatibility and setup reliability.

• Documentation

  • Updated the Safe Synthesizer tutorial to run without expected large-sample warnings.

[coderabbitai]

Note

Reviews paused

It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.

Use the following commands to manage reviews:

• @coderabbitai resume to resume automatic reviews.
• @coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

• ▶️ Resume reviews
• 🔍 Trigger review

Walkthrough

The pull request renames the training warmup setting with backward compatibility, updates GLiNER API handling, adjusts CUDA and package setup, and suppresses expected warnings in the tutorial notebook.

Changes

Compatibility and runtime updates

Layer / File(s)	Summary
Training warmup configuration src/nemo_safe_synthesizer/config/training.py, src/nemo_safe_synthesizer/training/huggingface_backend.py, tests/conftest.py, tests/training/test_huggingface_backend.py	warmup_steps accepts integer steps or fractional ratios. The deprecated warmup_ratio field maps to warmup_steps and emits a deprecation warning. The backend and fixtures use warmup_steps.
GLiNER loading and prediction compatibility src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py, tests/pii_replacer/test_detect.py	GLiNER prediction prefers inference and falls back to predict_entities. Model loading suppresses the deprecated warning. Tests target the updated API.
Runtime provisioning and tutorial warning handling script/brev/setup.sh, docs/tutorials/safe-synthesizer-101.ipynb	The generated kernelspec sets BNB_CUDA_VERSION=128. Setup pre-compiles torchao and range_regex while suppressing syntax warnings and ignoring compilation failures. The tutorial suppresses the expected ks_2samp warning during builder.run().

Estimated code review effort: 3 (Moderate) | ~25 minutes

Possibly related PRs

• NVIDIA-NeMo/Safe-Synthesizer#672: Related to the GLiNER detection API changes.
• NVIDIA-NeMo/Safe-Synthesizer#687: Related to provisioning changes in script/brev/setup.sh.

Suggested labels: chore

Suggested reviewers: kendrickb-nvidia

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 25.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
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.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the pull request's primary changes to address third-party deprecation warnings.

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch yunfeng/fix-deprecation-warnings

---

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

[greptile-apps]

Greptile Summary

The PR updates third-party integrations and configuration naming to remove expected deprecation warnings while preserving legacy warmup configuration compatibility.

• Replaces deprecated GLiNER batch prediction calls with inference and preserves per-record fallback behavior.
• Introduces validated warmup_steps configuration with migration from the deprecated warmup_ratio.
• Scopes expected warning suppression around GLiNER loading and SciPy KS evaluation.
• Updates Brev provisioning and documentation for CUDA index discovery, welcome-file handling, and notebook setup.

Confidence Score: 5/5

The PR appears safe to merge because no blocking failure remains.

No blocking failure remains.

Important Files Changed

Filename	Overview
src/nemo_safe_synthesizer/config/training.py	Adds the canonical warmup_steps field and preserves legacy warmup_ratio inputs through validated migration with canonical-field precedence.
src/nemo_safe_synthesizer/training/huggingface_backend.py	Passes the canonical warmup setting into Hugging Face TrainingArguments.
src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py	Moves GLiNER batch prediction to the current inference API and scopes suppression of the upstream download warning.
src/nemo_safe_synthesizer/evaluation/components/text_semantic_similarity.py	Suppresses only SciPy’s expected exact-to-asymptotic KS fallback warning around the affected calls.
script/brev/setup.sh	Updates Brev provisioning for source-provided welcome content, CUDA source-index discovery, and package precompilation.

_{Reviews (11): Last reviewed commit: "fix(brev): align setup with launchable r..." | Re-trigger Greptile}

[codecov]

Codecov Report

❌ Patch coverage is 95.55556% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
.../evaluation/components/text_semantic_similarity.py	63.63%	4 Missing ⚠️

📢 Thoughts on this report? Let us know!

[coderabbitai]

Actionable comments posted: 4

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 26bfb997-28a4-4382-b896-0f29d0f8d29b

📥 Commits

Reviewing files that changed from the base of the PR and between d806aaf and 22e36e2.

📒 Files selected for processing (8)

• docs/tutorials/safe-synthesizer-101.ipynb
• script/brev/setup.sh
• src/nemo_safe_synthesizer/config/training.py
• src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py
• src/nemo_safe_synthesizer/training/huggingface_backend.py
• tests/conftest.py
• tests/pii_replacer/test_detect.py
• tests/training/test_huggingface_backend.py

📜 Review details

⏰ Context from checks skipped due to timeout. (9)

• GitHub Check: Unit Tests (3.14)
• GitHub Check: Unit Tests (3.12)
• GitHub Check: Smoke Tests
• GitHub Check: Unit Tests (3.11)
• GitHub Check: Unit Tests (3.13)
• GitHub Check: End-user Wheel Install
• GitHub Check: Greptile Review
• GitHub Check: Analyze (Python)
• GitHub Check: Analyze (python)

🧰 Additional context used

📓 Path-based instructions (20)

**/*.{md,markdown,py}

📄 CodeRabbit inference engine (.cursor/rules/agent-markdown-style.mdc)

**/*.{md,markdown,py}: Avoid decorative bold (**text**) in list items, body text, and docstrings; use structural cues (headers, list markers, colons, backticks) for emphasis instead
Use backticks for code identifiers, paths, and CLI commands in markdown and docstrings

Files:

• tests/training/test_huggingface_backend.py
• src/nemo_safe_synthesizer/training/huggingface_backend.py
• tests/conftest.py
• tests/pii_replacer/test_detect.py
• src/nemo_safe_synthesizer/config/training.py
• src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py

tests/**

📄 CodeRabbit inference engine (.cursor/rules/repo-navigation.mdc)

tests/**: Mirror src/ directory structure in tests/ directory for test organization
Auto-mark tests by directory: tests/e2e/ → e2e, tests/smoke/ → smoke, otherwise default to unit

Files:

• tests/training/test_huggingface_backend.py
• tests/conftest.py
• tests/pii_replacer/test_detect.py

**/*.py

📄 CodeRabbit inference engine (AGENTS.md)

**/*.py: Use uv for Python dependency management and execution; never use pip or raw python. Target Python 3.11–3.14 and use modern syntax such as X | Y, list[str], and Self.
Run project tooling through mise tasks or wrapper scripts in tools/; do not invoke ruff or ty directly. Use uv run for Python execution.
Put durable implementation guidance in public function and class docstrings; put local invariants in source comments.
Use Python 3.11–3.14-compatible modern type syntax, including X | Y, list[str], and Self.

**/*.py: Use American English spelling in Python code, comments, and documentation; use from __future__ import annotations in every module.
Use BaseSettings for environment/CLI settings; prefer AliasChoices for fields accepting both Python and environment-variable names.
Pydantic model fields must include Field(description=...); prefer assignment-style Field() and use Annotated only for additional metadata or constraints.
Use @dataclass(frozen=True) for immutable value objects and validators, and field(default_factory=...) for mutable defaults; never use mutable default values directly.
Use StrEnum for string-valued configuration or serialization enums and plain Enum for internal-only constants.
Obtain loggers with observability.get_logger(__name__); do not call logging.getLogger() or structlog.get_logger() directly.
Do not use print() for operational library output; use the approved logger, click.echo() for CLI output, or sys.stdout.write() for raw tool output.
Use logger extra={} for metrics, counts, durations, and other data intended for machine querying; use f-strings for human-readable context.
Raise known errors through the Safe Synthesizer custom hierarchy, using dual inheritance where callers should also catch a built-in exception.
Use X | Y, built-in generic types, Self for fluent returns, collection ABCs for arguments, Protocol for structural boundaries, and avoid Any.
Pre...

Files:

• tests/training/test_huggingface_backend.py
• src/nemo_safe_synthesizer/training/huggingface_backend.py
• tests/conftest.py
• tests/pii_replacer/test_detect.py
• src/nemo_safe_synthesizer/config/training.py
• src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py

**/tests/**/*.py

📄 CodeRabbit inference engine (AGENTS.md)

Use the unit pytest marker instead of the deprecated unit_test marker; async tests do not need @pytest.mark.asyncio because asyncio_mode = auto.

Files:

• tests/training/test_huggingface_backend.py
• tests/conftest.py
• tests/pii_replacer/test_detect.py

**/*

📄 CodeRabbit inference engine (AGENTS.md)

**/*: Read AGENTS.local.md if it exists and give its instructions top priority.
When a task matches a repository-specific skill, read the corresponding skill under .agents/skills/ instead of duplicating its workflow instructions.
Do not commit unless the user explicitly asks for a commit or PR work.
When committing, require DCO sign-off and GPG signing using git commit --signoff --gpg-sign (or -s -S); never manually add Signed-off-by or use --no-gpg-sign.
Use feature branches based on main; branch names commonly include an issue-number prefix such as <author>/123-short-name.
For recurring testing, building, syncing, bootstrapping, worktree, and GitHub workflows, use the matching skill under .agents/skills/.
For a full GPU/development environment, use uv sync --frozen --extra cu129 --extra engine --group dev; bare uv sync --frozen is incomplete and can cause ty, import-check, and GPU-test failures.

End files with a newline, contain no trailing whitespace, use one space between sentences, and keep code, comments, and docstrings within 120 characters.

Files:

• tests/training/test_huggingface_backend.py
• src/nemo_safe_synthesizer/training/huggingface_backend.py
• tests/conftest.py
• script/brev/setup.sh
• tests/pii_replacer/test_detect.py
• src/nemo_safe_synthesizer/config/training.py
• src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py
• docs/tutorials/safe-synthesizer-101.ipynb

---

⚙️ CodeRabbit configuration file

**/*: Review as a senior maintainer for NeMo Safe Synthesizer. Prioritize issues that can change behavior, break user workflows, weaken privacy guarantees, hide failures, make tests unreliable, or create maintenance risk. Avoid generic style commentary unless it points to a concrete project convention that automated tools will not catch.
Comment only when the finding is actionable and tied to changed code. For each finding, state the impact, the condition that triggers it, and the smallest practical fix. Prefer one precise comment over broad advice. Do not ask for refactors outside the PR scope unless the changed code creates the problem.
Review type guidance: - Potential issue: use for correctness bugs, data loss, privacy leaks,
security risks, broken public APIs, invalid config behavior, missing
validation, hidden failures, nondeterministic tests, or CI breakage.

• Refactor suggestion: use for local maintainability problems introduced
  by the diff when they have clear future cost, such as duplicated setup,
  unclear boundaries, over-mocking, avoidable complexity, or opaque test
  helpers.
• Nitpick: avoid in chill mode. Do not emit formatting, import-order,
  wording, or style-only comments unless automated tools cannot catch the
  issue and it affects maintainability.

Severity guidance: - Critical: security/privacy leaks, data loss, training/test/holdout
contamination, or broken release/package/core pipeline execution.

• Major: incorrect generation/training/evaluation behavior, broken
  CLI/SDK public API, invalid config defaults or validators, or GPU/vLLM
  cleanup and process-isolation bugs likely to fail CI or production
  runs.
• Minor: localized bugs, missing focused tests for changed behavior, or
  bad test patterns that weaken regression coverage.
• Trivial: small cleanup with no behavior impact. Usually suppress in
  chill mode.
• Info: context only. Avoid unless it helps reviewers understand risk.
  Safe-Synthesizer-specific review focus: - Data ...

Files:

• tests/training/test_huggingface_backend.py
• src/nemo_safe_synthesizer/training/huggingface_backend.py
• tests/conftest.py
• script/brev/setup.sh
• tests/pii_replacer/test_detect.py
• src/nemo_safe_synthesizer/config/training.py
• src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py
• docs/tutorials/safe-synthesizer-101.ipynb

**/*.{md,py,sh,toml,yml,yaml,Dockerfile}

📄 CodeRabbit inference engine (AGENTS.md)

Follow the detailed language and file-format conventions defined in STYLE_GUIDE.md.

Files:

• tests/training/test_huggingface_backend.py
• src/nemo_safe_synthesizer/training/huggingface_backend.py
• tests/conftest.py
• script/brev/setup.sh
• tests/pii_replacer/test_detect.py
• src/nemo_safe_synthesizer/config/training.py
• src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py

tests/**/*.py

📄 CodeRabbit inference engine (STYLE_GUIDE.md)

tests/**/*.py: Use absolute imports in tests.
Name test files test_*.py, classes Test*, and functions test_<module>_<expected_behavior>; prefix fixtures with fixture_ and give each fixture a one-line purpose docstring.
Use function-scoped fixtures by default, bare assert, pytest.raises(match=...), pytest.approx(), and pytest.mark.parametrize for input combinations.
Use tmp_path for file operations, mock only external boundaries, avoid shared mutable state and order dependencies, and mark CUDA-dependent tests appropriately.

All existing tests must pass before submitting a pull request; new features must include tests and bug fixes must include regression tests.

tests/**/*.py: Every test must have exactly one category marker: unit, smoke, or e2e; use modifier markers such as slow and requires_gpu only in addition to a category marker.
Use pytest.importorskip for optional dependencies that require specific extras, such as sentence_transformers and vllm.
Use fixture_mock_processor or fixture_mock_processor_without_valid_records for ParsedResponse test data, with valid_records, invalid_records, errors, and integer prompt_number fields.
Use the shared load_test_dataset(filename) and load_test_dataframe(filename) helpers for loading test datasets where applicable.
Tests should mirror the source structure, such as tests/training/, tests/generation/, and corresponding source modules.
print() is permitted in tests for debug output; Ruff rule T201 is suppressed for the tests/ directory.
Do not import helpers directly from another file under tests/; for shared methods, use a relative import from conftest.py (pytest fixtures are available automatically).
When adding a vLLM GPU smoke-test file, add pytest.mark.vllm, create a dedicated test:smoke:gpu:* mise task for per-file process isolation, and include that task in test:smoke:gpu.
GPU tests must use the requires_gpu marker; vLLM tests additionally r...

Files:

• tests/training/test_huggingface_backend.py
• tests/conftest.py
• tests/pii_replacer/test_detect.py

---

⚙️ CodeRabbit configuration file

Review tests against tests/TESTING.md. Check marker usage, fixture naming, tmp_path usage, determinism, and GPU/vLLM process-isolation requirements. Flag slop tests that only check that code runs, assert result is not None when stronger invariants exist, over-mock internal implementation details, patch around the bug instead of reproducing it, or add broad snapshot/golden churn without a clear contract. Flag change detector tests that fail on harmless refactors, formatting, record ordering, incidental wording, or private implementation details without demonstrating a behavior regression. Prefer existing fixtures or focused new fixtures for repeated setup; keep tests DRY when reasonable without making the behavior under test opaque. print() is allowed in tests.

Files:

• tests/training/test_huggingface_backend.py
• tests/conftest.py
• tests/pii_replacer/test_detect.py

**/*.{py,sh,yaml,yml,md}

📄 CodeRabbit inference engine (STYLE_GUIDE.md)

Include the required SPDX copyright and Apache-2.0 license headers, using comment syntax appropriate to the file format.

Include SPDX copyright headers in all source files, except files explicitly listed in .copyrightignore.

Files:

• tests/training/test_huggingface_backend.py
• src/nemo_safe_synthesizer/training/huggingface_backend.py
• tests/conftest.py
• script/brev/setup.sh
• tests/pii_replacer/test_detect.py
• src/nemo_safe_synthesizer/config/training.py
• src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py

**/*.{py,pyi}

📄 CodeRabbit inference engine (CONTRIBUTING.md)

**/*.{py,pyi}: Keep shared Python package code compatible with Python 3.11; do not use Python 3.12-only syntax such as PEP 695 type statements or bracketed generic parameters.
Use the repository's pinned Ruff tasks for Python formatting, import sorting, and linting rather than invoking unpinned tools directly.
Run the repository's pinned ty type checker and maintain type-correct Python code.

Files:

• tests/training/test_huggingface_backend.py
• src/nemo_safe_synthesizer/training/huggingface_backend.py
• tests/conftest.py
• tests/pii_replacer/test_detect.py
• src/nemo_safe_synthesizer/config/training.py
• src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py

**/*.{py,sh,yaml,yml,toml,md}

📄 CodeRabbit inference engine (CONTRIBUTING.md)

Use mise tasks with the repository's pinned tool versions for formatting, checking, and testing before submitting changes.

Files:

• tests/training/test_huggingface_backend.py
• src/nemo_safe_synthesizer/training/huggingface_backend.py
• tests/conftest.py
• script/brev/setup.sh
• tests/pii_replacer/test_detect.py
• src/nemo_safe_synthesizer/config/training.py
• src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py

src/**/*.py

📄 CodeRabbit inference engine (STYLE_GUIDE.md)

src/**/*.py: Keep shared package code compatible with Python 3.11; do not use Python 3.12-only syntax such as PEP 695 type statements or bracketed generic parameters.
Use TYPE_CHECKING guards for heavy imports such as pandas, torch, and transformers; use relative imports in src/.
Do not add __init__.py to test directories; every directory under src/ containing Python files must have one.
Comments should explain why rather than narrate what; do not add redundant docstrings, defensive except Exception around trusted internal calls, unjustified type ignores, or casts/Any to hide type errors.
Use pathlib.Path instead of os.path, and use explicit if/raise validation instead of assert in library code.

Files:

• src/nemo_safe_synthesizer/training/huggingface_backend.py
• src/nemo_safe_synthesizer/config/training.py
• src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py

---

⚙️ CodeRabbit configuration file

Review library code against STYLE_GUIDE.md. Focus on behavior, API contracts, error handling, resource cleanup, typing, logging, and user-facing failures. Public APIs and nontrivial functions need Google-style docstrings.

Files:

• src/nemo_safe_synthesizer/training/huggingface_backend.py
• src/nemo_safe_synthesizer/config/training.py
• src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py

src/nemo_safe_synthesizer/training/**/*.py

⚙️ CodeRabbit configuration file

Review training changes for dataset preprocessing, model path handling, artifact writes, LoRA/DP behavior, GPU memory usage, reproducibility, and cleanup on failure.

Files:

• src/nemo_safe_synthesizer/training/huggingface_backend.py

tests/conftest.py

📄 CodeRabbit inference engine (.cursor/rules/repo-navigation.mdc)

Store shared test fixtures in tests/conftest.py

Files:

• tests/conftest.py

tests/**/*conftest.py

📄 CodeRabbit inference engine (tests/TESTING.md)

Use the fixture_ prefix for dataset and tokenizer fixtures; CLI helpers may use descriptive names such as mock_workdir.

Files:

• tests/conftest.py

**/*.sh

📄 CodeRabbit inference engine (AGENTS.md)

Never use ~ inside double-quoted shell strings because it does not expand; use $HOME or an absolute path instead.

Use #!/usr/bin/env bash, set -euo pipefail unless incompatible, quote variables, use ${VAR:-default}, use snake_case functions with _ for private helpers, and lint with ShellCheck.

Files:

• script/brev/setup.sh

---

⚙️ CodeRabbit configuration file

Review shell scripts for #!/usr/bin/env bash, set -euo pipefail where appropriate, quoting, repo root detection, and shellcheck compliance.

Files:

• script/brev/setup.sh

script/**

⚙️ CodeRabbit configuration file

Review standalone scripts for reproducibility and operational safety. Check argument validation, quoting, repo-root detection, environment variables, generated artifacts, external commands, GPU/cluster assumptions, and whether the script should be wired through Makefile or documented in README/docs.

Files:

• script/brev/setup.sh

src/**/config/**/*.py

📄 CodeRabbit inference engine (STYLE_GUIDE.md)

Use NSSBaseModel for user-facing configuration and parameter models in config/; use BaseModel or module-specific bases for DTOs and internal structures.

Files:

• src/nemo_safe_synthesizer/config/training.py

src/nemo_safe_synthesizer/config/**/*.py

⚙️ CodeRabbit configuration file

Treat config changes as user-facing API changes. Check Pydantic field descriptions, defaults, validators, aliases, override behavior, CLI help text impact, YAML compatibility, and documented parameter semantics.

Files:

• src/nemo_safe_synthesizer/config/training.py

src/nemo_safe_synthesizer/pii_replacer/**/*.py

⚙️ CodeRabbit configuration file

Treat PII replacement changes as high-risk. Check entity coverage, replacement determinism, leakage of original values, handling of empty or multilingual text, and compatibility with optional dependencies.

Files:

• src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py

docs/**

⚙️ CodeRabbit configuration file

Review documentation as MkDocs Material content. Check Diataxis fit, accurate commands, internal links, code fences, and markdown style from STYLE_GUIDE.md.

Files:

• docs/tutorials/safe-synthesizer-101.ipynb

🧠 Learnings (4)

📚 Learning: 2026-05-27T22:20:37.354Z

Learnt from: kendrickb-nvidia
Repo: NVIDIA-NeMo/Safe-Synthesizer PR: 520
File: tests/generation/test_vllm_backend.py:556-587
Timestamp: 2026-05-27T22:20:37.354Z
Learning: In NVIDIA-NeMo/Safe-Synthesizer, `tests/conftest.py`’s `pytest_collection_modifyitems` hook applies pytest category markers automatically based on each test file’s path: tests under `/e2e/` get `pytest.mark.e2e`, tests under `/smoke/` get `pytest.mark.smoke`, and all other tests get `pytest.mark.unit`. Therefore, when reviewing pytest tests outside `tests/e2e/` and `tests/smoke/`, do not flag missing explicit `pytest.mark.unit` decorators on test classes/functions as an issue (the hook will add them during collection). If a new test directory/category is introduced, ensure the hook is updated so it’s categorized correctly.

Applied to files:

• tests/training/test_huggingface_backend.py
• tests/conftest.py
• tests/pii_replacer/test_detect.py

📚 Learning: 2026-07-27T22:07:22.590Z

Learnt from: binaryaaron
Repo: NVIDIA-NeMo/Safe-Synthesizer PR: 673
File: tests/pii_replacer/test_edit.py:327-327
Timestamp: 2026-07-27T22:07:22.590Z
Learning: When tests read structured logging context from Python `logging.LogRecord` instances, don’t access `record.ctx` directly (it isn’t declared on `LogRecord` and will break static typing). Instead, use `getattr(record, "ctx", default)` (or an appropriate fallback) to safely handle cases where `ctx` may or may not be attached. This applies even if Ruff rule `B009` isn’t enabled in the repo.

Applied to files:

• tests/training/test_huggingface_backend.py
• tests/conftest.py
• tests/pii_replacer/test_detect.py

📚 Learning: 2026-07-29T17:12:32.642Z

Learnt from: zywind
Repo: NVIDIA-NeMo/Safe-Synthesizer PR: 654
File: tests/config/test_parameters.py:116-118
Timestamp: 2026-07-29T17:12:32.642Z
Learning: In Pydantic validation tests (e.g., models configured with `from_attributes`), when asserting failures from `model_validate(...)`, assert the structured error details (such as `ValidationError.errors()[0]["type"]`, e.g. `"model_attributes_type"`) rather than relying on the human-readable error message text. This keeps tests stable even if wording changes, while still verifying the correct validation rule is triggered.

Applied to files:

• tests/training/test_huggingface_backend.py
• tests/conftest.py
• tests/pii_replacer/test_detect.py

📚 Learning: 2026-08-05T19:07:15.856Z

Learnt from: binaryaaron
Repo: NVIDIA-NeMo/Safe-Synthesizer PR: 679
File: tests/conftest.py:89-96
Timestamp: 2026-08-05T19:07:15.856Z
Learning: In the Safe Synthesizer test suite, treat CPU and CUDA installation profiles as the supported pytest collection profiles because they install PyTorch. A bare installation without PyTorch is an incomplete, unsupported profile. GPU-marked test modules may import PyTorch before tests/conftest.py::pytest_collection_modifyitems executes, so do not require that collection hook to prevent such imports.

Applied to files:

• tests/training/test_huggingface_backend.py
• tests/conftest.py
• tests/pii_replacer/test_detect.py

🔇 Additional comments (12)

src/nemo_safe_synthesizer/training/huggingface_backend.py (1)

424-424: LGTM!

tests/conftest.py (1)

139-139: LGTM!

tests/training/test_huggingface_backend.py (1)

136-141: LGTM!

Also applies to: 147-152

script/brev/setup.sh (1)

271-274: LGTM!

docs/tutorials/safe-synthesizer-101.ipynb (2)

2-148: LGTM!

Also applies to: 149-154, 155-230, 249-251

---

231-248: 🗄️ Data Integrity & Integration

No change required. setup.sh registers the kernelspec in the python3 directory and sets BNB_CUDA_VERSION=128. The notebook selects python3.

			> Likely an incorrect or invalid review comment.

src/nemo_safe_synthesizer/pii_replacer/data_editor/detect.py (3)

8-8: LGTM!

---

629-632: 🎯 Functional Correctness

No API compatibility issue

The lockfile selects GLiNER 0.2.24, which provides inference. Older releases without inference also provide predict_entities, which this fallback supports.

			> Likely an incorrect or invalid review comment.

---

589-595: 🎯 Functional Correctness

Do not change this warning filter for the current dependency set. The project requires huggingface-hub>=1.3.0 and locks version 1.17.0; resume_download was removed in version 1.0.0. This deprecation warning does not apply to the pinned dependency.

			> Likely an incorrect or invalid review comment.

tests/pii_replacer/test_detect.py (3)

69-69: LGTM!

Also applies to: 87-87

---

106-113: LGTM!

---

149-149: LGTM!

Also applies to: 166-166

[coderabbitai]

Actionable comments posted: 1

♻️ Duplicate comments (1)

src/nemo_safe_synthesizer/config/training.py (1)

214-225: ⚠️ Potential issue | 🟠 Major

Enforce the documented warmup domain for both field names.

warmup_steps accepts 1.5 and float("inf"), although the description permits only an integer step count or a ratio in (0, 1). warmup_ratio has no equivalent constraint, and Line 244 copies it after normal field validation. Invalid legacy values can reach the backend, which forwards warmup_steps directly to the trainer.

Use one finite, range, and integrality validator for both input paths. Add regression tests for both names. This repeats the unresolved validator finding from the previous review.

As per path instructions, configuration changes must check “default values, override merging, validators, aliases, and backwards-compatible user-facing config changes,” and new behavior needs focused tests.

#!/usr/bin/env bash
set -euo pipefail

rg -n -C 5 'warmup_(ratio|steps)|validate_assignment' src tests pyproject.toml uv.lock

Also applies to: 227-244

Source: Path instructions

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: e68240ed-a2d5-4879-883c-30e6a1918bbb

📥 Commits

Reviewing files that changed from the base of the PR and between 22e36e2 and f48a520.

📒 Files selected for processing (2)

• docs/tutorials/safe-synthesizer-101.ipynb
• src/nemo_safe_synthesizer/config/training.py

🚧 Files skipped from review as they are similar to previous changes (1)

• docs/tutorials/safe-synthesizer-101.ipynb

📜 Review details

⏰ Context from checks skipped due to timeout. (8)

• GitHub Check: Analyze (Python)
• GitHub Check: conventional-commit / semantic-pull-request
• GitHub Check: Greptile Review
• GitHub Check: Detect changes
• GitHub Check: Format
• GitHub Check: Typecheck
• GitHub Check: Analyze (javascript-typescript)
• GitHub Check: Analyze (python)

🧰 Additional context used

📓 Path-based instructions (10)

**/*.{md,markdown,py}

📄 CodeRabbit inference engine (.cursor/rules/agent-markdown-style.mdc)

**/*.{md,markdown,py}: Avoid decorative bold (**text**) in list items, body text, and docstrings; use structural cues (headers, list markers, colons, backticks) for emphasis instead
Use backticks for code identifiers, paths, and CLI commands in markdown and docstrings

Files:

• src/nemo_safe_synthesizer/config/training.py

**/*.py

📄 CodeRabbit inference engine (AGENTS.md)

**/*.py: Use uv for Python dependency management and execution; never use pip or raw python. Target Python 3.11–3.14 and use modern syntax such as X | Y, list[str], and Self.
Run project tooling through mise tasks or wrapper scripts in tools/; do not invoke ruff or ty directly. Use uv run for Python execution.
Put durable implementation guidance in public function and class docstrings; put local invariants in source comments.
Use Python 3.11–3.14-compatible modern type syntax, including X | Y, list[str], and Self.

**/*.py: Use American English spelling in Python code, comments, and documentation; use from __future__ import annotations in every module.
Use BaseSettings for environment/CLI settings; prefer AliasChoices for fields accepting both Python and environment-variable names.
Pydantic model fields must include Field(description=...); prefer assignment-style Field() and use Annotated only for additional metadata or constraints.
Use @dataclass(frozen=True) for immutable value objects and validators, and field(default_factory=...) for mutable defaults; never use mutable default values directly.
Use StrEnum for string-valued configuration or serialization enums and plain Enum for internal-only constants.
Obtain loggers with observability.get_logger(__name__); do not call logging.getLogger() or structlog.get_logger() directly.
Do not use print() for operational library output; use the approved logger, click.echo() for CLI output, or sys.stdout.write() for raw tool output.
Use logger extra={} for metrics, counts, durations, and other data intended for machine querying; use f-strings for human-readable context.
Raise known errors through the Safe Synthesizer custom hierarchy, using dual inheritance where callers should also catch a built-in exception.
Use X | Y, built-in generic types, Self for fluent returns, collection ABCs for arguments, Protocol for structural boundaries, and avoid Any.
Pre...

Files:

• src/nemo_safe_synthesizer/config/training.py

**/*

📄 CodeRabbit inference engine (AGENTS.md)

**/*: Read AGENTS.local.md if it exists and give its instructions top priority.
When a task matches a repository-specific skill, read the corresponding skill under .agents/skills/ instead of duplicating its workflow instructions.
Do not commit unless the user explicitly asks for a commit or PR work.
When committing, require DCO sign-off and GPG signing using git commit --signoff --gpg-sign (or -s -S); never manually add Signed-off-by or use --no-gpg-sign.
Use feature branches based on main; branch names commonly include an issue-number prefix such as <author>/123-short-name.
For recurring testing, building, syncing, bootstrapping, worktree, and GitHub workflows, use the matching skill under .agents/skills/.
For a full GPU/development environment, use uv sync --frozen --extra cu129 --extra engine --group dev; bare uv sync --frozen is incomplete and can cause ty, import-check, and GPU-test failures.

End files with a newline, contain no trailing whitespace, use one space between sentences, and keep code, comments, and docstrings within 120 characters.

Files:

• src/nemo_safe_synthesizer/config/training.py

---

⚙️ CodeRabbit configuration file

**/*: Review as a senior maintainer for NeMo Safe Synthesizer. Prioritize issues that can change behavior, break user workflows, weaken privacy guarantees, hide failures, make tests unreliable, or create maintenance risk. Avoid generic style commentary unless it points to a concrete project convention that automated tools will not catch.
Comment only when the finding is actionable and tied to changed code. For each finding, state the impact, the condition that triggers it, and the smallest practical fix. Prefer one precise comment over broad advice. Do not ask for refactors outside the PR scope unless the changed code creates the problem.
Review type guidance: - Potential issue: use for correctness bugs, data loss, privacy leaks,
security risks, broken public APIs, invalid config behavior, missing
validation, hidden failures, nondeterministic tests, or CI breakage.

• Refactor suggestion: use for local maintainability problems introduced
  by the diff when they have clear future cost, such as duplicated setup,
  unclear boundaries, over-mocking, avoidable complexity, or opaque test
  helpers.
• Nitpick: avoid in chill mode. Do not emit formatting, import-order,
  wording, or style-only comments unless automated tools cannot catch the
  issue and it affects maintainability.

Severity guidance: - Critical: security/privacy leaks, data loss, training/test/holdout
contamination, or broken release/package/core pipeline execution.

• Major: incorrect generation/training/evaluation behavior, broken
  CLI/SDK public API, invalid config defaults or validators, or GPU/vLLM
  cleanup and process-isolation bugs likely to fail CI or production
  runs.
• Minor: localized bugs, missing focused tests for changed behavior, or
  bad test patterns that weaken regression coverage.
• Trivial: small cleanup with no behavior impact. Usually suppress in
  chill mode.
• Info: context only. Avoid unless it helps reviewers understand risk.
  Safe-Synthesizer-specific review focus: - Data ...

Files:

• src/nemo_safe_synthesizer/config/training.py

**/*.{md,py,sh,toml,yml,yaml,Dockerfile}

📄 CodeRabbit inference engine (AGENTS.md)

Follow the detailed language and file-format conventions defined in STYLE_GUIDE.md.

Files:

• src/nemo_safe_synthesizer/config/training.py

src/**/config/**/*.py

📄 CodeRabbit inference engine (STYLE_GUIDE.md)

Use NSSBaseModel for user-facing configuration and parameter models in config/; use BaseModel or module-specific bases for DTOs and internal structures.

Files:

• src/nemo_safe_synthesizer/config/training.py

src/**/*.py

📄 CodeRabbit inference engine (STYLE_GUIDE.md)

src/**/*.py: Keep shared package code compatible with Python 3.11; do not use Python 3.12-only syntax such as PEP 695 type statements or bracketed generic parameters.
Use TYPE_CHECKING guards for heavy imports such as pandas, torch, and transformers; use relative imports in src/.
Do not add __init__.py to test directories; every directory under src/ containing Python files must have one.
Comments should explain why rather than narrate what; do not add redundant docstrings, defensive except Exception around trusted internal calls, unjustified type ignores, or casts/Any to hide type errors.
Use pathlib.Path instead of os.path, and use explicit if/raise validation instead of assert in library code.

Files:

• src/nemo_safe_synthesizer/config/training.py

---

⚙️ CodeRabbit configuration file

Review library code against STYLE_GUIDE.md. Focus on behavior, API contracts, error handling, resource cleanup, typing, logging, and user-facing failures. Public APIs and nontrivial functions need Google-style docstrings.

Files:

• src/nemo_safe_synthesizer/config/training.py

**/*.{py,sh,yaml,yml,md}

📄 CodeRabbit inference engine (STYLE_GUIDE.md)

Include the required SPDX copyright and Apache-2.0 license headers, using comment syntax appropriate to the file format.

Include SPDX copyright headers in all source files, except files explicitly listed in .copyrightignore.

Files:

• src/nemo_safe_synthesizer/config/training.py

**/*.{py,pyi}

📄 CodeRabbit inference engine (CONTRIBUTING.md)

**/*.{py,pyi}: Keep shared Python package code compatible with Python 3.11; do not use Python 3.12-only syntax such as PEP 695 type statements or bracketed generic parameters.
Use the repository's pinned Ruff tasks for Python formatting, import sorting, and linting rather than invoking unpinned tools directly.
Run the repository's pinned ty type checker and maintain type-correct Python code.

Files:

• src/nemo_safe_synthesizer/config/training.py

**/*.{py,sh,yaml,yml,toml,md}

📄 CodeRabbit inference engine (CONTRIBUTING.md)

Use mise tasks with the repository's pinned tool versions for formatting, checking, and testing before submitting changes.

Files:

• src/nemo_safe_synthesizer/config/training.py

src/nemo_safe_synthesizer/config/**/*.py

⚙️ CodeRabbit configuration file

Treat config changes as user-facing API changes. Check Pydantic field descriptions, defaults, validators, aliases, override behavior, CLI help text impact, YAML compatibility, and documented parameter semantics.

Files:

• src/nemo_safe_synthesizer/config/training.py

🔇 Additional comments (1)

src/nemo_safe_synthesizer/config/training.py (1)

7-15: LGTM!