3 ms·
Does anyone know if they actually support learning on the chip or does it just execute a pre-trained model efficiently?
by anonymousDan 10y ago
Does anyone know if they actually support learning on the chip or does it just execute a pre-trained model efficiently?
- ogrisel 10y agoI would bet that is limited to low-energy inference which is the most important impediment to deploy deep vision models in mobile devices.
- dharma1 10y agoinference only. This is an IoT/robotics play for Intel, combined with RealSense. The only feasible product at the moment for training large models is NVidia GPUs. Maybe AMD in 1-2 years time, if they decide to wake up and invest in software optimisation for deep learning