skills.rest
Orchestra-Research
Orchestra-ResearchOFFICIAL
Explore all skills by this author →
GitHub

mlflow

Track ML experiments, manage model registry versions, and deploy models with MLflow.

13.0k|930|Updated Nov 3, 2025
Search Tags:#mlops#machine learning#model deployment#experiment tracking#mlflow#model registry#model versioning
One-click install
npx skills add https://github.com/Orchestra-Research/AI-research-SKILLs --skill mlflow-orchestra-research
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: mlflow
Source: https://github.com/Orchestra-Research/AI-research-SKILLs/tree/main/13-mlops/mlflow
Command: npx skills add https://github.com/Orchestra-Research/AI-research-SKILLs --skill mlflow-orchestra-research

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires mlflow, sqlalchemy, boto3, and includes references (resource) components.

What problem does it solve? Managing machine learning experiments manually leads to lost hyperparameters, untracked metrics, and no clear path from training to production deployment. This Skill provides a framework-agnostic platform for tracking experiments, versioning models in a central registry, and deploying them to local servers, Docker, or cloud platforms. ## Core Features & Use Cases - Experiment Tracking: Log parameters, metrics, artifacts, and models for every run, with autologging support for PyTorch, scikit-learn, Keras, XGBoost, and HuggingFace Transformers. - Model Registry: Version models automatically, manage lifecycle stages or aliases (champion/challenger), and implement approval workflows with tags and annotations. - Deployment: Serve models locally via REST API, build Docker images, or deploy to AWS SageMaker, Azure ML, and Kubernetes with KServe. - Use Case: A data science team trains multiple classifier variants, compares runs by accuracy in the MLflow UI, registers the best model, promotes it through staging to production, and serves it as a Docker container behind a REST endpoint. ## Quick Start Use the mlflow skill to set up experiment tracking for my PyTorch training script and register the best model in the model registry.

Frequently Asked Questions about mlflow

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I track ML experiments with MLflow in Python?▼

Call mlflow.start_run() as a context manager, then use mlflow.log_param, mlflow.log_metric, and mlflow.log_artifact inside the block. Enable mlflow.autolog() to automatically capture parameters, metrics, and models for supported frameworks like scikit-learn and PyTorch.

How to register and version a model in MLflow Model Registry?▼

Pass registered_model_name when logging a model, or call mlflow.register_model with a runs:/ URI. Each registration creates an incrementing version, which you can promote between stages or assign aliases like champion using MlflowClient.

Does MLflow work with HuggingFace Transformers and PyTorch?▼

Yes, MLflow provides mlflow.transformers and mlflow.pytorch flavors with log_model and autolog support. Transformers models are logged with their tokenizer, and PyTorch Lightning training metrics and checkpoints are captured automatically.

Should I use MLflow stages or model aliases?▼

Use aliases for new projects since stages are deprecated in MLflow 2.9+. Aliases like champion and challenger support flexible A/B testing workflows, while stages (Staging, Production, Archived) remain for legacy compatibility.

How do I deploy an MLflow model to production?▼

Serve locally with mlflow models serve, build a Docker image with mlflow models build-docker, or deploy to AWS SageMaker and Azure ML using mlflow deployments create. Load production models via the models:/ URI with a stage or alias.

Why is my MLflow run not logging metrics automatically?▼

Autologging must be enabled before training starts with mlflow.autolog() or a framework-specific variant like mlflow.sklearn.autolog(). Verify your framework version is supported and that training occurs inside an active run context.

Related Skills

Explore More Software Engineering Skills→
cas-bigdatalabcas-bigdatalab
541

data_splitter

Split JSONL datasets into subsets by specified ratios.

Official
Intermediate
kalilurrahmankalilurrahman
1

model-deployment

Deploy ML models to production with FastAPI, TorchServe, monitoring, and A/B testing.

Community
Advanced
organvm-i-theoriaorganvm-i-theoria

ml-experiment-tracker

Track ML experiments, version datasets, and manage model registries with MLflow, W&B, and DVC.

Official
Intermediate
ehadziabdicehadziabdic

ml-experiment-tracking

Track and compare ML experiments with MLflow, Weights & Biases, Neptune, and CometML.

Community
Advanced