Blog / Simplify and automate the machine learning model lifecycle
Simplify and automate the machine learning model lifecycle

Simplify and automate the machine learning model lifecycle

by Tarek Oraby | on September 18, 2024

We're pleased to introduce Model Hub, Valohai's centralized solution for tracking and managing machine learning models throughout their entire lifecycle.

Model Hub offers more than just model registry capabilities — it's packed with advanced features, all within a streamlined interface. It simplifies collaboration, model iteration, and regulatory compliance for data science teams.

The key features of the Model Hub include:

  • A unified view of all models
  • Versioning and lineage tracking
  • Lifecycle management
  • Training performance tracking
  • Automated workflows
  • Access control and documentation
  • Easy model iteration and refinement

Minimal manual intervention

What makes the Model Hub unique is that it builds on Valohai's core design philosophy of reproducibility. In Valohai, the data, code, parameters, and the computing environment of every ML job are automatically tracked, ensuring that workflows are always reproducible. Model Hub leverages this automation to model management. Simply select your model files, and Model Hub will link them to all relevant elements—data, code, pipelines, and parameters—without requiring manual input.

This approach empowers data science teams to manage models with minimal manual effort, reducing the risk of errors, improving consistency, and ensuring all models are fully documented and reproducible throughout their lifecycle.

Single pane of glass for all models

Model Hub offers a centralized view of all your models. The UI allows you to easily access detailed information, such as model version history, training performance, and lineage. You can compare versions, download artifacts, or trigger workflows without needing to search through multiple tools.

Model overview in Valohai's Model Hub

Simplified model versioning

With Model Hub, creating and tracking model versions is easy. A single click or API call generates a new version and tracks its lineage, ensuring consistency. Integration with ML pipelines allows for automatic version registration, reducing the risk of errors.

Automated lineage tracking

Model Hub automatically tracks the entire lineage of each model version—from data and parameters to code and pipeline steps. By leveraging Valohai's reproducibility-first design, every model's origin can be traced back through the pipeline that generated it, ensuring transparency and the ability to reproduce past results.

Lineage tracking in Valohai's Model Hub

Lifecycle management

Model Hub helps teams manage model lifecycles. You can approve or reject versions based on performance, intended use, or other criteria. It also allows tagging and note-taking of model versions for better organization and collaboration.

Lifecycle management in Valohai's Model Hub

Performance tracking and comparison

Valohai makes performance tracking seamless by capturing training metrics automatically from print() statements, without the need for custom logging libraries. Model Hub leverages this to provide an easy-to-understand view of metrics like accuracy, loss, and F1 scores across model versions. This enables teams to monitor model progress, identify trends, and make informed decisions about iteration and refinement.

Prformance tracking in Valohai's Model Hub

Triggers for automated workflows

To make deployment smoother, Model Hub supports triggers that can be used to automate workflows, like deploying a new model version or rolling back to a previous one. This automation support ensures seamless transitions from development to production.

Access control

Admins can control who can view or modify models and their versions by setting access permissions. This prevents unauthorized changes and ensures team members have the right access to the right models.

Documentation

Clear documentation is essential for collaboration and compliance. Model Hub allows you to document each model and model versions intended use, limitations, and other relevant information, ensuring that all team members have the context they need.

Model iteration and refinement

Model Hub makes it easy to refine models by enabling you to re-trigger pipelines, adjust parameters, and experiment with different configurations. This accelerates model development and ensures continuous improvement.

Conclusion

Valohai's Model Hub simplifies and automates machine learning model management across their entire lifecycle. Its comprehensive feature set improves collaboration, ensures compliance, and enhances operational efficiency. By centralizing and automating model tracking, teams can focus more on developing models and less on manual tasks.

If you’re not a Valohai user yet, you can get started by booking a meeting with our Customer Team. Alternatively, you can get a preview of Valohai’s capabilities using our self-service trial.

Free eBookPractical MLOpsHow to get started with MLOps?