Skip to main content

Artifacts

Store, inspect, and download files attached to runs.

Artifacts are files attached to a run: model weights, exported models, plots, logs, or reports. Use them to inspect a result in detail, retrieve a model for further evaluation, or share a report with a colleague.

Open a file#

Open a run and find Artifacts. For example, an evaluation report can show which classes performed well and where the model made mistakes, beyond the overall accuracy metric. A model page also collects artifacts from its associated runs, with links back to their source.

Use Preview (1) to inspect a supported file, or choose Download from its actions (2).

Upload a file#

  1. Select Add beside Artifacts on the run details page.
  2. Select Choose file and pick a file from your computer.
  3. Select Upload, then check that the file appears in the list.

To upload files as your code runs, use Direct tracking or the CLI.

Troubleshoot uploads#

Open Storage from your account to see usage and any quota. For upload errors, see Troubleshooting.

For files uploaded by a script, check its upload errors before assuming the run contains every output. See Safe integration.

Delete files#

Open Storage from your account to delete uploaded files, including in bulk.

Warning

Deletion is permanent. Download anything you need first, and consider whether colleagues or evidence reviews still rely on the file.

Next steps#