search

MarvinTeichmann / tensorflow-fcn

An Implementation of Fully Convolutional Networks in Tensorflow.
MIT License
1.1k stars 433 forks source link

Why the memory of weights is 1.6G? #15

Closed JackieLeeTHU11 closed 7 years ago

JackieLeeTHU11 commented 7 years ago

@MarvinTeichmann When I save the weights of this model (FCN), the memory is 1.6GB and the .meta file is 537.7MB. But actually, the memory of the weights is about 530MB. It sames strange here?

MarvinTeichmann commented 7 years ago

The reason is, that tensorflow stores weights much more inefficient then numpy. If you care for memory store the weights in an .npy file.

Alternatively discuss the issue with a tensorflow maintainer (e.g. @vrv), if you think this should be fixed.