1 paper · 1 filter
Gabriel Afriat, Xiang Meng, Shibal Ibrahim +2
Weight pruning is a common technique for compressing large neural networks. We focus on the challenging post-training one-shot setting, where a pre-trained model is compressed with…