A Novel Memory-Scheduling Strategy for Large Convolutional Neural Network on Memory-Limited Devices

<div>mmCNN strategy for VGG-S model under different constraints of GPU memory size. (a) VGG-S 200M. (b) VGG-S 100M. (c) VGG-S 50M. (d) VGG-S 10M. (e) VGG-S 5M.</div>

Computational Intelligence and Neuroscience

fig8

Figure 8

Figure 8: A Novel Memory-Scheduling Strategy for Large Convolutional Neural Network on Memory-Limited Devices 