OOM when allocating tensor with shape[5184,4,4,2048] and type float on /job:localhost/replica:0/task

OOM when allocating tensor with shape[5184,4,4,2048] and type float on /job:localhost/replica:0/task

报错原因:GPU内存不够,减小batch,或者减小模型。

猜你喜欢

转载自blog.csdn.net/qq_32799915/article/details/80355249
今日推荐