MissingLink lets you capture artifacts that are generated as output from your experiments. These artifacts are attached to the experiments themselves and can be inspected or queried in the same way data is on a Data Volume That's because all saved artifacts are automatically added to their own Data Volume that is linked to the project you are working with.
You can view the artifacts in the MissingLink dashboard by clicking the Artifacts tab in the experiment that generated it.
By default, anything you put in a root output folder,
/output, will be automatically collected as an artifact of the experiment. For example, by running this command:
ml run xp --command 'ls -al > /output/ls.txt'
You can specify additional locations for artifacts by using the
--output-paths flag with
ml run commands, so:
ml run xp --command 'ls -al > /results/ls.txt' --output-paths '/results'
Once the job has completed or has been stopped, any files found in your
output paths will be saved to the associated data volume. Note that it might take a few minutes for the artifacts to be fully displayed.
Dedicated data volumes for artifacts
For artifacts to be successfully captured and stored, they must have a dedicated data volume that is fully configured.
The first time you run an experiment that produces artifacts, a data volume is created and is accessible in the Data Volumes area of the dashboard. To complete the configuration, access the data volume and click Complete Wizard. Go through the necessary steps, as prompted. Then rerun the experiment.