Przejdź do treści
Biblioteka skilli
AI i agenci · Integracja z narzędziem

mlflow

/mlflow

Wspiera pracę z MLflow: śledzenie eksperymentów, autologging, rejestr modeli i przepisy MLflow, przy wersjonowaniu artefaktów i serwowaniu modeli.

Repozytorium
Agent-Skills ↗
Licencja
MIT
Ostatnia zmiana repozytorium
01.10.2026

Ocena i bezpieczeństwo

100/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
1
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
3
Przestarzałość

Brak sygnałów. Repozytorium zmienione 01.10.2026, bez odwołań do wycofanych modeli i API.

Licencja

MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.

Co robi i kiedy

Wspiera pracę z MLflow: śledzenie eksperymentów, autologging, rejestr modeli i przepisy MLflow, przy wersjonowaniu artefaktów i serwowaniu modeli.

Oryginalny opis (EN)

Expert MLflow assistance covering experiment tracking, autologging, model registry, and MLflow recipes. Use when tracking ML metrics, versioning model artifacts, or serving production models.

Jak działa

  1. When to Use
  2. Quick Start
  3. Core Concepts
  4. Common Patterns
  5. Best Practices
  6. Troubleshooting
  7. References
Sekcje z nagłówków SKILL.md, w oryginale · 172 linii, 616 słów
Podgląd SKILL.md
---
name: mlflow
description: Expert MLflow assistance covering experiment tracking, autologging, model registry, and MLflow recipes. Use when tracking ML metrics, versioning model artifacts, or serving production models.
---

# MLflow

MLflow is an open-source platform for managing the end-to-end machine learning lifecycle, including experiment tracking, model registry, artifact storage, and LLM tracing evaluation.

## When to Use

- **Machine Learning Experiment Tracking**: Logging parameters, metrics, code versions, and artifacts across runs.
- **Model Registry & Governance**: Managing model lifecycle stages (Staging, Production, Archived) with lineage.
- **LLM Evaluation & Prompt Engineering**: Evaluating LLMs, RAG applications, and prompts using MLflow Evaluate.
- **Unified Model Deployment**: Packaging models into self-contained flavors (Python function, PyTorch, ONNX) for one-click deployment.

## Quick Start

```python
import mlflow
import mlflow.sklearn
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import mean_squared_error

mlflow.set_experiment("housing_price_prediction")

with mlflow.start_run():
    n_estimators = 50
    model = RandomForestRegressor(n_estimators=n_estimators)
    model.fit(X_train, y_train)

    preds = model.predict(X_val)
    mse = mean_squared_error(y_val, preds)

    mlflow.log_param("n_estimators", n_estimators)
    mlflow.log_metric("mse", mse)
    mlflow.sklearn.log_model(model, "random_forest_model")
```

## Core Concepts

### Experiment Tracking & Autologging

Logging parameters, evaluation metrics, and model weights automatically:

```python
import mlflow
import mlflow.sklearn
from sklearn.ensemble import RandomForestClassifier
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split

# Configure remote or local tracking URI
mlflow.set_tracking_uri("http://localhost:5000")
mlflow.set_experiment("iris_classification_prod")

# Enable automatic framework logging
mlflow.sklearn.autolog(log_model_signatures=True)

X, y = load_iris(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

with mlflow.start_run(run_name="rf_n100_d5") as run:
    params = {"n_estimators": 100, "max_depth": 5, "random_state": 42}
    mlflow.log_params(params)

    clf = RandomForestClassifier(**params)
    clf.fit(X_train, y_train)

    score = clf.score(X_test, y_test)
    mlflow.log_metric("accuracy", score)
    print(f"Logged run: {run.info.run_id} with Accuracy: {score:.4f}")
```

### Model Registry & Production Staging

Registering and promoting versioned models:

```python
from mlflow import MlflowClient

client = MlflowClient()

# Register model from a completed run artifact
model_uri = f"runs:/{run.info.run_id}/model"
registered_model = mlflow.register_model(model_uri, "CustomerChurnPredictor")

# Assign an alias (MLflow 2.x recommended pattern)
client.set_registered_model_alias(
    name="CustomerChurnPredictor",
    alias="champion",
    version=registered_model.version
)

# Load champion model in production serving microservice
champion_model = mlflow.pyfunc.load_model("models:/CustomerChurnPredictor@champion")
predictions = champion_model.predict(X_test)
```

### LLM Evaluation with mlflow.evaluate

Benchmarking RAG outputs against ground truth datasets:

```python
import mlflow
import pandas as pd

eval_df = pd.DataFrame({
    "inputs": ["What is MLflow?", "How does autologging work?"],
    "ground_truth": [
        "MLflow is an open-source platform for managing the end-to-end ML lifecycle.",
        "Autologging automatically captures metrics, parameters, and models without explicit log statements."
    ],
    "predictions": [
        "MLflow manages machine learning experiments, models, and deployments.",
        "It automatically tracks parameters and metrics during training."
    ]
})

with mlflow.start_run():
    results = mlflow.evaluate(
        data=eval_df,
        targets="ground_truth",
        predictions="predictions",
        model_type="text-summarization",
        evaluators="default"
    )
    print("Evaluation Metrics:", results.metrics)
```

## Common Patterns

### Autologging Framework Integrations

**Problem**: Writing manual `mlflow.log_metric` lines for every epoch and hyperparameter.

**Solution**:
Enable automatic framework-level logging:

```python
import mlflow

# Autolog PyTorch, TensorFlow, Scikit-Learn, LightGBM, or XGBoost
mlflow.autolog()

# Subsequent model.fit() calls automatically log all parameters, metrics, and models
model.fit(X_train, y_train)
```

## Best Practices

**Do**:

- Target MLflow 2.15+ utilizing model aliases (`@champion`, `@challenger`) rather than legacy stage transitions.
- Log model signatures (`mlflow.models.infer_signature`) to ensure input/output schema validation at deployment time.
- Use `mlflow.start_run()` inside Python context managers to ensure runs are reliably closed on errors.
- Back up the remote backend store (PostgreSQL) and artifact repository (S3/GCS) regularly.

**Don't**:

- Store large training datasets directly as artifacts; log dataset hashes and S3 URIs via `mlflow.data`.
- Use local filesystem tracking URIs in production or collaborative team environments.
- Hardcode tracking URIs; configure via environment variable `MLFLOW_TRACKING_URI`.

## Troubleshooting

| Error                                                | Cause                                                                | Solution                                                                       |
| :--------------------------------------------------- | :------------------------------------------------------------------- | :----------------------------------------------------------------------------- |
| `MlflowException: Could not find experiment with ID` | Experiment deleted or connecting to mismatched tracking URI.         | Set explicit tracking URI: `mlflow.set_tracking_uri("http://localhost:5000")`. |
| `Artifact transfer failed`                           | S3 / GCS bucket permissions missing on client runner.                | Ensure AWS/GCP credentials with write permissions are active in environment.   |
| `Schema enforcement error on model load`             | Input DataFrame columns mismatch model signature logged at training. | Align DataFrame columns and types with logged `ModelSignature`.                |

## References

- [MLflow Documentation](https://mlflow.org/)

Wersje

RepozytoriumWynikLicencja
Ta strona · G1Joshi/Agent-Skills100MIT
TerminalSkills/skills100Apache-2.0

Migawka bazy z 13.11.2026 · wersja paczki 6160175772d5 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 100mlflow-experiment-tracking-and-model-registry
    Konfiguruje i obsługuje MLflow: serwer śledzenia, magazyny backendu i artefaktów, Model Registry z etapami i aliasami, definicje MLproject, autologowanie oraz wdrożenia własne lub zarządzane w Databricks.
  • 100mlflow-tracking-setup
    Prawdopodobnie konfiguruje śledzenie eksperymentów MLflow w ramach kategorii ML Training.
  • 100mlflow-mlops-pipeline
    Buduje i zarządza kompletnymi potokami MLOps z MLflow: śledzenie eksperymentów, rejestr modeli, logowanie artefaktów, serwery z backendem S3 lub GCS, modele PyFunc, Optuna i usługi inferencji FastAPI w Dockerze.
  • 100mlops-mlflow-experiment-tracking
    Zapisuje w MLflow hiperparametry, metryki straty, krzywe ewaluacji i artefakty modeli podczas treningu.
  • 95ml4t-registry-system
    Śledzi eksperymenty modeli handlowych ML z adresowaniem po treści, przy wersjonowaniu modeli, cech i artefaktów eksperymentów dla powtarzalności.
  • 90ml-experiment-tracking
    Konfiguruje śledzenie eksperymentów ML w MLflow, W&B, Neptune lub DVC: logowanie parametrów, metryk, artefaktów i rejestru modeli.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.