Open Bug 1932407 Opened 7 days ago Updated 7 days ago

Inference Engine: Memory-Efficient Model Downloading

Categories

(Core :: Machine Learning, enhancement)

Product:
Core
Component:
Machine Learning
Type:
enhancement

Tracking

()

People

(Reporter: atossou, Assigned: atossou)

References

Details

(Whiteboard: [genai])

Currently, the model is fully downloaded to an arraybuffer which can uses too much memory for big models.

I suggest we download the model directly to a file while keeping the memory constant.

An implementation is already available here: https://phabricator.services.mozilla.com/D229641#change-F8nGVnX6zdic

You need to log in before you can comment on or make changes to this bug.