What are modelcatalogd and modelmanagerd?

modelcatalogd and modelmanagerd are processes linked to on-device machine learning models.

What are these processes?

• modelcatalogd: keeps a catalogue of machine learning models on your Mac. It tracks which models are installed, which versions are current, and which need updating.
• modelmanagerd: manages the life of ML models. It downloads, updates, and loads models as needed.

What models do they manage?

macOS includes many on-device ML models for features such as:

• Siri speech recognition
• Live Text (text recognition in images)
• Visual Look Up (object identification)
• Photo face recognition and scene sorting
• Dictation
• Predictive text and autocorrect
• Apple Intelligence features (on supported hardware)

Why separate model management?

ML models can be large, from hundreds of megabytes to several gigabytes. Rather than bundling every model with the operating system, macOS downloads models as needed and updates them apart from macOS updates. modelmanagerd handles this.

Do they use many resources?

They can use a lot of bandwidth when downloading new models. CPU use is small. They manage models but do not run them. That work is done by aned and other framework processes.

Should you worry?

No. They are normal macOS processes that keep your Mac stocked with the latest ML models for on-device intelligence features.

Enjoyed this post?

Well, you could share the post with others, follow me with RSS Feeds and/or send me a comment via email.

Tags