Use W&B Artifacts to store and keep track of datasets, models, and evaluation results across machine learning pipelines. Think of an artifact as a versioned folder of data. You can store entire datasets directly in artifacts, or use artifact references to point to data in other systems.
Explore an example of Artifacts for dataset versioning and model management with a quick, interactive notebook hosted in Google Colab.
Initialize a run and log an artifact, for example a dataset version that you're using to train a model.
run = wandb.init(job_type="dataset-creation")artifact = wandb.Artifact('my-dataset', type='dataset')artifact.add_file('my-dataset.txt')run.log_artifact(artifact)
Start a new run and pull down that saved artifact, for example to use the dataset to train a model.
run = wandb.init(job_type="model-training")artifact = run.use_artifact('my-dataset:latest')artifact_dir = artifact.download()
If an artifact changes, re-run the same artifact creation script. In this case, imagine the data in the my-dataset.txt file changes. This same script will capture the new version neatly — we'll checksum the artifact, identify that something changed, and track the new version. If nothing changes, we don't reupload any data or create a new version.
run = wandb.init(job_type="dataset-creation")artifact = wandb.Artifact('my-dataset', type='dataset')# Imagine more lines of text were added to this text file:artifact.add_file('my-dataset.txt')# Log that artifact, and we identify the changed filerun.log_artifact(artifact)# Now you have a new version of the artifact, tracked in W&B
Using our Artifacts API, you can log artifacts as outputs of W&B runs, or use artifacts as input to runs.
Since a run can use another run’s output artifact as input, artifacts and runs together form a directed graph. You don’t need to define pipelines ahead of time. Just use and log artifacts, and we’ll stitch everything together.
Here's an example artifact where you can see the summary view of the DAG, as well as the zoomed-out view of every execution of each step and every artifact version.
To learn how to use Artifacts, check out the Artifacts API Docs →
Follow along with our interactive tutorial and learn how to track your machine learning pipeline with W&B Artifacts.