Skip to content

Create a Neptune project#

A Neptune project typically represents one machine learning task. A project can contain runs, models, model versions, and project-level metadata.

Before you start#

If you haven't already, sign up for a Neptune account at neptune.ai/register.

Creating a project#

  1. In the left sidebar:

    • Click , or
    • Click and then Create new project.

  2. Enter the project information and set a privacy level:

    1. Project name: Must consist of 3 to 50 alphanumerical characters and hyphens (-).

      The name is case-insensitive. For example, CV-Project and cv-project will be treated as the same.

    2. Project key: Must consist of 1 to 10 capital letters.

      The project key can't be changed after the project is created. It's used to identify runs and other Neptune objects.

    3. Project privacy: Level of visibility for the project. Users with access effectively become project owners.

      • Private: Only people assigned to the project can see it.

        Note on plans

        On plans without project-level access control, private projects (accessible only to some workspace members) are not available. You need to set the privacy to "workspace" or "public".

      • Workspace: Any workspace members can access these projects.

      • Public: Freely available to view by anyone on the internet.

  3. (optional) Set a project color and enter a description.

  4. To finish, click Create.

Creating a new project in Neptune

API tip

You can also create projects programmatically. For details, see Create and delete projects via API.

Projects limit#

Note that you can only create a new project if your quota of projects isn't full.

Project-based plans

On older project-based subscription plans, the projects quota is managed by your workspace admin.

If the limit of projects is reached, you need to upgrade your subscription or archive another project first.

Next steps#

You can now record metadata to the project by setting the NEPTUNE_PROJECT environment variable to the full project name (workspace-name/project-name).

Example: If your workspace name is ml-team and your project name is classification:

export NEPTUNE_PROJECT="ml-team/classification"

export NEPTUNE_PROJECT="ml-team/classification"
System-wide
setx NEPTUNE_PROJECT "ml-team/classification"
Current session only
set NEPTUNE_PROJECT="ml-team/classification"

Alternatively, you can specify the project name when initializing Neptune:

Create a run
run = neptune.init_run(project="ml-team/classification")
params = {
    "max_epochs": 10,
    "optimizer": "Adam",
    "dropout": 0.2,
}
run["parameters"] = params
...
Create a model
model = neptune.init_model(project="ml-team/classification", key="FOREST")
model["signature"].upload("model_signature.json")
...
Add or fetch project-level metadata
project = neptune.init_project(project="ml-team/classification")
project["dataset/v0.1"].track_files("s3://datasets/images")
...

Related