replace `FLAGS.batch_size` by `images.get_shape()[0]`
We can directly get `batch_size` by `images.get_shape()[0]` in `inference` method, since maybe we will not use `cifar.inputs` method to build the input.
Showing
Please register or sign in to comment