Understanding the ResourceExhaustedError: OOM when allocating tensor with shape

Question

I\'m trying to implement a skip thought model using tensorflow and a current version is placed here.

Currently I using one GPU of my machine (total 2 GPUs) and the