Masked Structural Growth for 2x Faster Language Model Pre-training.

Yiqun Yao, Zheng Zhang, Jing Li, Yequan Wang

Browse the full ICLR paper archive.