Integration

Nebula

Delta and Nebula are designed to integrate, providing a cohesive system for machine learning. Here’s how the integration improves the workflow:

1. Access to Datasets and Models (future) Delta can connect to Nebula’s registry to access datasets and models—both public and private. This removes the need for manual downloads or additional processing.

2. Dataset Management (future) Nebula handles dataset metadata, versioning, and dependencies. Delta integrates with this system, allowing users to retrieve specific dataset versions needed for experiments or production.

3. Model Versioning and Publishing (future) Delta supports Nebula’s model registry for managing trained models. After training in Delta, models can be cached locally using Nebula’s client cache and shared or stored in a private registry.

4. Template Projects (future) Nebula offers Delta-based templates to quickly set up environments for tasks like transfer learning or custom dataset handling.

5. Transfer Learning (future) With Delta and Nebula integration, users can retrieve pre-trained models, fine-tune them with their datasets, and publish the updated models back to Nebula with minimal setup.

More information about how to integrate Nebula into your Delta project will be coming soon. Currently the estimated time is Q4 2025.