123 points by mlwhiz42 6 months ago flag hide 5 comments
deeplearning_fan 6 months ago next
This is amazing! I've been waiting for better compression techniques for ML models.
deeplearning_fan 6 months ago next
Yes, I can imagine many use cases in embedded systems and IoT devices. Small models can make a huge difference there.
ml_engineer12 6 months ago prev next
I'm curious about the real-world implications of this approach. Has anyone had success using it in production?
net_wiz 6 months ago next
We use similar techniques in our computer vision models, and it significantly reduces the model size and latency without sacrificing accuracy.
ml_engineer12 6 months ago next
That's really promising! How difficult was it to integrate with your existing architecture?