Linting And Formatting
Code style and quality rules for NeMo AutoModel — ruff configuration, naming conventions, type hints, docstrings, copyright headers, and the code review checklist.
Score breakdown
Estimated from the available content and source signals.
Model compatibility
Inferred fit is not the same as a recorded hands-on test.
Overview
Linting and Formatting
Single source of truth for code style in NeMo AutoModel. Read this before writing new code or reviewing PRs.
Formatting and Linting
Run before every commit:
# Auto-format all source files (line length, quotes, trailing commas, etc.)
ruff format .
# Lint and auto-fix what can be fixed (unused imports, isort, etc.)
ruff check --fix .
To check without modifying files:
ruff format --check . # exits non-zero if any file would change
ruff check . # exits non-zero on lint violations
bandit -r app.py nemo_automodel examples scripts tools tutorials -t B614
To lint a single file or directory:
ruff format nemo_automodel/components/models/llama/model.py
ruff check --fix nemo_automodel/components/models/llama/
What linting enforces (from pyproject.toml)
| Rule | ID | Description |
|---|---|---|
| Line length | — | 120 characters (formatter) |
| Quote style | — | Double quotes |
| Unused imports | F401 | Auto-removed by --fix (ignored in __init__.py) |
| Unused variables | F841 | Auto-removed by --fix |
| Undefined names | F821 | Error |
| f-string without placeholders | F541 | Error |
| Import sorting | I | isort-compatible ordering, auto-fixed |
| Docstring convention | D101/D103 | Google style (currently ignored — selected then suppressed) |
| No pickle | S301/S403 | Security: forbids pickle.load |
| Restricted PyTorch load | B614 | Requires explicit weights_only=True for torch.load |
| Ambiguous variable names | E741 | Error (e.g., l, O, I) |
Tests (tests/) are excluded from Ruff and Bandit checks. Docstring rules
(D) are also relaxed in test files.
Type Hints
Required on all public API functions and methods.
- Use
T | Noneinstead ofOptional[T] - Use
X | Yinstead ofUnion[X, Y] - Use built-in generics (
list,dict,tuple) instead oftypingequivalents
Docstrings
Google-style where docstrings are added:
def build_model(config: dict) -> torch.nn.Module:
"""Instantiate and shard the model from config.
Args:
config: Mapping with _target_ and model hyperparameters.
Returns:
Fully initialized model ready for distributed training.
"""
NVIDIA Copyright Header
Every Python file must start with the NVIDIA copyright block. Do not remove or modify it. Use the current year (2026).
# Copyright (c) 2026, NVIDIA CORPORATION. All rights reserved.
#
# 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.
Additional Style Conventions
- Explicit over implicit. Inline logic where possible; avoid hiding behavior behind unnecessary layers of indirection.
- No speculative abstractions. Do not add features, parameters, or generalization beyond what is explicitly asked for.
- Optional dependencies must be guarded with
safe_import()fromnemo_automodel.shared.import_utils. Never let an optional import crash module loading. - Components must not import each other — enforced by
import-linter(seepyproject.toml).
Code Review Checklist
- Copyright header present on all new Python files
- Type hints on all public functions and methods
- Docstrings on public classes and functions (Google style)
- Double quotes for strings
- No bare
print()— uselogging.getLogger(__name__) - No commented-out code without explanation
- Optional imports guarded with
safe_import() - No cross-component imports between
components/subdirectories
Automated Review
The review-only maintainability heuristics and thresholds live in
.github/workflows/claude-review.yml. Keep repository-wide coding rules here
and automated-review prompt policy there so the detailed checklist has one
source of truth.
Best for
- Use Linting And Formatting when this documented workflow matches the task.
Tips and best practices
- Review the source instructions and adapt inputs before running the workflow.
What This Skill Can Do
AI-generated examples showing real capabilities
Was this skill useful?
Be the first to share a result.