GitLab

the number of batches per epoch also depends on the number of gpus. it should be `num_batches_per_epoch = (cifar10.NUM_EXAMPLES_PER_EPOCH_FOR_TRAIN / FLAGS.batch_size / FLAGS.num_gpus)`

While the original code works fine in practice, it technically
allows gradient application and moving average update to happen
in any order. This causes the behavior to deviate from pure
mathematical specifications.

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.

Updated cifar10_input.py to put the data augmentation into a name_scope named data_augmentation so the tensorboard graph is easier to read and more meaningful

tf.contrib.framework.get_or_create_global_step -> tf.train.get_or_create_global_step

CIFAR website is now using https.
Since urllib does not follow the http to https redirection and thus never retrieves any data, the data URL needs an update.

2. Fixed accuracy when calcuting logs more than 10 steps

Without the new variable_scope, creating apply_gradient_op raises
an error that additional moving average or slot variables could not
be created. This is because of the 'leaky reuse' of variable scope,
so we correct the problem by explicitly introducing a new variable scope.

Related issues: tensorflow/models#901, tensorflow/tensorflow#6220