Documentation

This documentation details the different logics and functionalities of Active Tigger. It is a work in progress. It will be updated as the project evolves.

General comments

Only one process allowed in the same time by user. There is two types of processes in Active Tigger:

CPU processes: they are used to prepare data, train quick models, etc.
GPU processes: these processes are run on the GPU. The server can only run a limited number of GPU processes at the same time.

In both case, if the number of processes is too high, they will be queued.

Accounts

A user can have the following status :

root: can create and manage projects, users, and all data. Can access all projects and the /monitor page.
manager: can create and manage projects where he.she has complete rights. Can add users to the projects.
annotator: can only annotate on projects where he.she have been added. The frontend is simplified for this role.

User have also a relational status for a specific project:

manager: can manage the project, add users, and access all data.
annotator: can only annotate on the project without changing the project settings (schemes, labels, models, etc.).

For the moment, there is no management at the scheme level.

Create a project

Prepare labels and features

Explore the data

Annotate

Fine-tune a BERT model

Test model

The test set: - Created on the beginning of the project - Uploaded latter

Once activated, the test mode : - Deactivate for the user the choice of scheme, label management - Allow only annotation for the test set - Allow to explore the test set