For all experiments, we used 10% of the training data as batch size for the large-batch experiments and 256 data points for small-batch experiments. 512 batch_size应该不算大的
【在 x****u 的大作中提到】 : https://arxiv.org/abs/1609.04836 : On Large-Batch Training for Deep Learning: Generalization Gap and Sharp : Minima
【在 L****8 的大作中提到】 : For all experiments, we used : 10% of the training data as batch size for the large-batch experiments : and 256 data points for small-batch experiments. : 512 batch_size应该不算大的
【在 L****8 的大作中提到】 : For all experiments, we used : 10% of the training data as batch size for the large-batch experiments : and 256 data points for small-batch experiments. : 512 batch_size应该不算大的
s*r
13 楼
amex的apr现在是多少啊?
g*u
14 楼
你的 10% of train 和 256 谁好?
【在 L****8 的大作中提到】 : For all experiments, we used : 10% of the training data as batch size for the large-batch experiments : and 256 data points for small-batch experiments. : 512 batch_size应该不算大的