In CML on Private Cloud, the ML Workspace is where data scientists get their work
done. After your Admin has created or given you access to an environment, you can set up a
workspace.
The first user to access the ML workspace after it is created must have the
EnvironmentAdmin role assigned.
Log in to the CDP Private Cloud web interface using your corporate credentials
or other credentials that you received from your CDP administrator.
Click ML Workspaces.
Click Provision Workspace. The Provision
Workspace panel displays.
In Provision Workspace, fill out the following
fields.
Workspace Name - Give the ML workspace a name.
For example, test-cml. Do not use capital letters in the workspace
name.
Select Environment - From the dropdown, select
the environment where the ML workspace must be provisioned. If you do
not have any environments available to you in the dropdown, contact your
CDP admin to gain access.
Namespace - Enter the namespace to use for the
ML workspace.
NFS Server - Select Internal to use an NFS
server that is integrated into the Kubernetes cluster. This is the
recommended selection at this time.
The path to the internal NFS server is already set in the
environment.
In Production Machine Learning, select to enable the
following features.
Enable Governance - Enables advanced lineage and
governance features.
Governance Principal Name - If
Enable Governance is selected, you can use
the default value of mlgov, or enter an alternative
name. The alternative name must be present in your environment and be
given permissions in Ranger to allow the MLGovernance service deliver
events to Atlas.
Enable Model Metrics - Enables exporting metrics
for models to a PostgreSQL database.
In Other Settings, select to enable the following
features.
Enable TLS - Select this to enable https access
to the workspace.
Enable Monitoring - Administrators (users with
the EnvironmentAdmin role) can use a Grafana dashboard to monitor
resource usage in the provisioned workspace.
CML Static Subdomain - This is a custom name for the workspace
endpoint, and it is also used for the URLs of models, applications, and
experiments. Only one workspace with the specific subdomain endpoint
name can be running at a time. You can create a wildcard certificate for
this endpoint in advance. The workspace name has this format:
<static subdomain name>.<environment
name>.<workload subdomain>.<base
domain></cmd>
Click Provision Workspace. The new workspace
provisioning process takes several minutes.
After the workspace is provisioned, you can log in by clicking the workspace name on
the Machine Learning Workspaces page. The first user to log
in must be the administrator.