Created by: tonyyang-svail

The hidden dimension of the last two layers should be 4096 (this gives ~400M model size). Ours is 512(this yields ~70M as model size).

https://github.com/PaddlePaddle/Paddle/blob/d43469b74e74b9c1b2a96fd731d2740367955adb/python/paddle/fluid/tests/book/test_image_classification.py#L87-L92

This will affect our benchmark with other frameworks.

The definition in other frameworks: https://github.com/ry/tensorflow-vgg16/blob/97a9e1ff663ac7238f986d8b8247f4ab81f83432/vgg16.py#L86-L89 , or https://gist.github.com/baraldilorenzo/07d7802847aaad0a35d3#file-vgg-16_keras-py-L46