Closed feynman233 closed 7 years ago
Hi! I think there should be some memory-efficient implementations, but I don't know exactly any repo. And in this one no memory efficient implementation and at this moment I don't intend to do this. But you are welcome to send pull request)
hi,@feynman233 @ikhlestov dose it mean that it can not train densenet_BC on 224224 pictures? my gpu is 12GB, and i try to train on 224224 images ,set the batchsize to 10, it comes the error 'OOM when allocating tensor with shape[10,224,224,108]'
Hi! It seems so. Unfortunately I have no access to the ImageNet dataset, so I not test the network on it.
Hi, I have found the same problem about the GPU memory,and is there any memory-efficient tensorflow Implementation? Thanks very much!