Examples¶

Learn Tracelet through practical examples in the /examples directory. Each category demonstrates different aspects of experiment tracking.

Quick Examples¶

Basic Tracking¶

from tracelet import Experiment

# Create experiment with automatic detection
exp = Experiment(name="my_model", backend=["wandb"], automagic=True)
exp.start()

# Your existing training code - metrics logged automatically
for epoch in range(10):
    loss = model.train_one_epoch()
    print(f"Loss: {loss}")

exp.stop()

Artifact Tracking¶

from tracelet.core.artifacts import ArtifactType

# Log a trained model
model_artifact = exp.create_artifact(
    name="classifier",
    artifact_type=ArtifactType.MODEL
).add_file("model.pth", "model/classifier.pth")
exp.log_artifact(model_artifact)

Detailed Examples¶

Manual Tracking ¶

Step-by-step introduction to manual metric logging and parameter tracking.

Multi-Backend ¶

Compare MLflow, ClearML, and W&B backends with the same experiment.

Notebooks ¶

Jupyter notebook examples with visualizations and interactive exploration.

Repository Examples¶

The /examples directory contains runnable examples organized by complexity:

01_manual_tracking/ - Basic manual tracking examples
02_automagic_tracking/ - Zero-config automatic tracking
03_backend_integrations/ - Backend-specific features
04_advanced_features/ - Production-ready patterns
05_lightning_automagic/ - PyTorch Lightning integration
06_artifacts/ - Artifact management examples

Quick Start Path¶

New to tracking: examples/01_manual_tracking/01_basic_manual.py
Want zero config: examples/02_automagic_tracking/01_basic_automagic.py
Use PyTorch Lightning: examples/05_lightning_automagic/simple_lightning_example.py
Compare backends: examples/03_backend_integrations/compare_all_backends.py

Running Examples¶

All examples use synthetic data and can be run immediately:

cd examples
python 01_manual_tracking/01_basic_manual.py

For multi-backend examples, install the required backends:

pip install tracelet[mlflow,wandb,clearml]