ML Models

Currently supporting a feature in Label Studio Enterprise.

ModelInterface

Defines a "type signature" for the inputs and outputs for an LLM call as related to a Label Studio project.

ModelVersion

Defines an "implementation" of an LLM call, given any user input needed. Currently, the only one implemented is a third-party model (such as OpenAI) with a chat prompt.

ModelRun

Defines a run of a ModelVersion on a chosen set of tasks on a project. Adds that ModelVersion's outputs to the project as predictions.

Diagram

---
Django ER Diagram
---
erDiagram
ModelInterface {
    AutoField id
    CharField title
    TextField description
    ForeignKey created_by
    DateTimeField created_at
    DateTimeField updated_at
    ForeignKey organization
    JSONField input_fields
    JSONField output_classes
    ManyToManyField associated_projects
}
ThirdPartyModelVersion {
    AutoField id
    CharField title
    ForeignKey parent_model
    TextField prompt
    CharField provider
    CharField provider_model_id
    ForeignKey created_by
    DateTimeField created_at
    DateTimeField updated_at
    ForeignKey organization
}
ModelRun {
    AutoField id
    ForeignKey organization
    ForeignKey project
    ForeignKey model_version
    ForeignKey created_by
    CharField project_subset
    CharField status
    CharField job_id
    DateTimeField created_at
    DateTimeField triggered_at
    DateTimeField predictions_updated_at
    DateTimeField completed_at
}
ModelInterface }|--|| User : created_by
ModelInterface }|--|| Organization : organization
ModelInterface }|--|{ Project : associated_projects
ThirdPartyModelVersion }|--|| ModelInterface : parent_model
ThirdPartyModelVersion }|--|| User : created_by
ThirdPartyModelVersion }|--|| Organization : organization
ModelRun }|--|| Organization : organization
ModelRun }|--|| Project : project
ModelRun }|--|| ThirdPartyModelVersion : model_version
ModelRun }|--|| User : created_by

Loading

---

generated by:

pip install django-diagram
python -m django_diagram --settings=core.settings.label_studio --app=ml_models --output=ml_models/README.md