Workspace
A workspace refers to an isolated environment or set of resources that allows users or teams to organize and manage their machine learning (ML) experiments, pipelines, and resources effectively.
The idea of a workspace is commonly applied in practice and relates to managing different namespaces, notebooks, pipelines, and other resources in an isolated, organized manner. vIn general, a workspace can be thought of as an environment where users can:
- Collaborate on machine learning projects.
- Access shared or isolated resources such as notebooks, datasets, models, and pipelines.
- Manage resources in a secure, scalable manner powered using Kubernetes namespaces in the host cluster (i.e. infrastructure)
Benefits¶
Workspaces provide organizations with the following benefits:
Isolation of Resources¶
By leveraging namespaces and RBAC, the platform ensures that resources such as pipelines, notebooks, and datasets are isolated between teams or users. This provides security and prevents interference in a multi-user setup.
Collaboration and Sharing¶
Workspaces facilitate collaboration by allowing users to share resources like datasets and models within a team or project, while keeping sensitive data isolated across different namespaces.
Scalability¶
Workspaces can be easily scaled to fit the requirements of different users. For example, different teams can have different resource quotas depending on the scale of their workloads, whether they need more GPUs for model training or memory for data processing.
Customizable Environments¶
Users can customize their workspace environments by selecting specific notebook images, container configurations, and resources such as CPUs or GPUs.
Dashboard¶
A personalized web based dashboard is available for every authenticated user. The dashboard is the control center, where the user can manage components such as pipelines, experiments, and notebook servers.