Analyzes PyTorch, TensorFlow, scikit-learn code for ML/DL defects and risks.
npx skills add https://github.com/physics91/claude-vibe --skill ml-reviewer
Please help me install this Agent Skill. Skill: ml-reviewer Source: https://github.com/physics91/claude-vibe/tree/main/skills/ml-reviewer Command: npx skills add https://github.com/physics91/claude-vibe --skill ml-reviewer
This Skill reviews machine learning and deep learning codebases for model construction, training loops, data pipelines, GPU utilization, and MLOps practices to prevent common pitfalls and inefficiencies.
Run the ml-reviewer on a Python project with frameworks declared in requirements.txt or pyproject.toml to get a targeted ML code quality report.
High-intent search queries and answers about installing and using this skill.
Code review for ML frameworks detects architectural problems, training loop errors, and GPU misconfigurations by analyzing PyTorch, TensorFlow, and scikit-learn patterns. ml-reviewer scans model.py, train.py, and dataset.py files alongside requirements.txt or pyproject.toml to identify issues like missing gradient clipping, improper eval mode, and data pipeline problems with severity ratings and remediation steps.
GPU optimization checks flag non-optimal device usage and memory patterns in training scripts. ml-reviewer analyzes code patterns to identify inefficient GPU allocation and suggests corrections to improve performance across PyTorch and TensorFlow implementations.
MLOps review validates experiment tracking setup, reproducibility practices, and logging configurations. ml-reviewer ensures proper frameworks (PyTorch, TensorFlow, scikit-learn) integration with common signal files and detects misconfigurations in data pipelines, training loops, and experiment management.
Framework-aware review applies across research and production stages, inspecting model architecture, training patterns, data pipelines, and experiment-tracking practices. ml-reviewer targets both contexts using common file patterns (model.py, train.py, dataset.py) and dependency declarations.
Data pipeline checks validate data loading, preprocessing, and augmentation sanity by scanning dataset.py and related patterns. ml-reviewer detects common pitfalls in data handling that impact model training quality and provides structured findings with recommendations.
ml-reviewer detects frameworks through requirements.txt or pyproject.toml and analyzes model files (.pt, .pth, .h5, .pkl). It performs environment-aware checks on PyTorch, TensorFlow, Keras, and scikit-learn codebases to generate targeted quality reports.
KaiserWhoLearnsTrain and evaluate machine learning models across scikit-learn, PyTorch, and TensorFlow.
gokulkrishhAnalyze code changes for security, performance, accessibility, and quality issues.
twjarviszyra-webTrack ML experiments, metrics, and artifacts across training runs.
R2W1csLint and validate JavaScript and Python code for syntax and style issues.
