Skip to content

Artifacts#

Previewing artifact metadata in Neptune

This section displays tracked artifacts. The contents of these files and folders are not uploaded; only their metadata.

Click on the files in the list to view the artifact metadata, such as hash, size, and path.

See example in Neptune 

What is an artifact in Neptune?#

You create an Artifact field by using the track_files() method and passing a path to a file or folder.

Example: Single file
run["train/dataset"].track_files("./datasets/train.csv")
Example: Folder
run["train/images"].track_files("./datasets/images")

The field name is up to you to specify. In the above example, we nest the artifact fields dataset and images under the namespace train.

Related

For the full artifact logging guide, see Track artifacts

Compare artifacts between runs#

You can compare artifact metadata between two runs at once.

To make runs visible in the source and target drop-down lists, toggle the eye icons ( ).

Differences shown for each artifact:

  • Path
  • Size
  • MD5 hash
  • Which other runs use the given artifact versions

See artifact comparison example in Neptune 

Related