Worker runs
Worker runs¶
Processes are made up of worker runs. A worker run is a unique combination of:
- one process
- one worker version
- one (optional) model version
- one (optional) configuration.
Info
Worker runs must be unique within a process. It is possible to run the same worker version multiple times in a single process, but only if it is used with different models and/or configurations.
When configuring a process, you create these worker runs by selecting a worker version and adding it to your process. In the process configuration view you then have the option to select a model version if the selected worker version supports or requires it, and a configuration.
Worker configurations¶
Workers can allow for some parameters to be set by a user when using them in a process, by defining a user configuration. These configurations can be used in inference processes (as well as templates) and in training processes.
When configuring a worker run, you can either select an existing configuration or create a new one.
If you do not select any configuration, a default configuration defined by the worker is used. These default configuration values are also those that pre-fill the new configuration creation form. You can only create a new configuration if it is different from the default configuration, and if a configuration with the same parameter values does not already exist.
Archived configurations¶
Like worker and model versions, worker configurations cannot be deleted if results produced by worker runs they are a part of still exist on the Arkindex instance. They can however be archived: archived configurations can no longer be used in processes, and only instance administrators can un-archive a configuration.
Configurations access rights¶
Info
This only concerns Arkindex Enterprise Edition: in the Community Edition, there are no access rights.
Worker configuration access rights depend on worker access rights: any user with access to a worker has access to its configurations. A user must have administrator access to a worker to be able to create new worker configurations.
Worker results¶
Worker runs identify the objects (elements, transcriptions, classifications, metadata etc) that they have produced. You can use a worker run UUID to filter elements inside a project or a folder.
When an object was created as a result of a process (as opposed to a manual annotation) a summary of the worker run is shown at the top of the element details panel, below the element’s name.
Clicking on this worker run summary opens a modal that displays the details of this worker run: the worker version, model, configuration and process.
You can delete all the results produced by a given worker run.