Project - Creation & collaboration
A Project
is somehow an AI Lab where data scientists can perform as many Experiment
as they want in order to create the best ModelVersion
possible.
1. Projects list
You can access the Projects features from the Navigation Bar:
From this Project list, you can visualize the list of all the Project
you can access in the frame of the current Organization. For each Project
, you can visualize the number of Experiment
, DatasetVersion
, and contextual information.
This view also allows you to create a brand-new Project
.
It is recommended to create a new Project
in your Organization each time you work on a new use case. Basically, you should have one Project
per ModelVersion
, you are planning to train.
A
Project
is not displayed ?You can only visualize only the
Project
you are defined as Contributor in.
2. Create a new Project
Project
From the Projects list, you can click on + New Project
The Project
creation form will then open. Through this form, you will be able to give for traceability purposes a name and a description of your new Project
.
Then, you can select the members of your Organization that will be authorized to visualize and access this new Project
. To do so, you just need to click on a member's name to add or remove him/her to the Contributors list.
Default Contributors
By default, all members of your Organization are pre-selected as Contributors in the Project creation form.
To finalize the Project
creation, click on Create.
3. A collaborative lab
A Project
is a space dedicated to the training of a ModelVersion
for a dedicated purpose.
As we know the process of ModelVersion
creation is a long process that needs several tries, experimentations, and adjustment, a Project
allows you to perform all your ModelVersion
training attempts in a structured and trackable space.
To do so, in a Project
, you'll be able to store and gather all the DatasetVersion
needed in the frame of the current Project
and retrieve all the Experiment
performed until now in a single view.
4. Project overview
Once created, a Project
offers three different tabs:
- Overview
- Experiments
- Settings
A. Overview
a. Attach DatasetVersion
DatasetVersion
The Project overview is divided into two sections, the DatasetVersion panel and the Last Experiment list.
Obviously, after initialization, those two placeholders are empty.
So the very first step will be to attach to your Project
, the DatasetsetVersion
you plan to work within the frame of the ModelVersion
creation. This can be done by clicking on the + button next to Attached datasets. A modal will open in which you'll be able to browse first among your Dataset
and then among related DatasetVersion
to select the ones to be attached to the Project
.
Once attached, your Project
is ready to get its first Experiment
.
B. Last Experiments
The Last Experiments list is quite self-explanatory, it allows you to quickly visualize the main properties related to the last Experiment
created in the frame of the Project and create a new Experiment
as detailed here.
Clicking on any Experiment
name will lead you to the Experiment overview page.
5. Experiments
By accessing the Experiments tab you will access the exhaustive list of Experiment
created in the frame of the current Project
and visualize the main properties of those Experiment
.
You also have the possibility to create a new Experiment
by clicking on the + New Experiment button, more details about Experiment
creation are available here.
In case, you have many Experiment
in your Project
, you can leverage the Search bar to search for particular Experiment
. The Search bar embeds an auto-completion engine that allows you to search for Experiment
on particular properties such as Name, Status, and Creation date ..
6. Settings
The Settings page, basically allows authorized users to modify the name of the Project
and delete it through the General tab and manage the Users that can access this Project
though the Collaborators tab.
Updated about 1 year ago