Coverage for hypergan/discriminators/common.py : 59%

import tensorflow as tf 

import hyperchamber as hc 

def repeating_block(component, net, depth, filter=3): 

ops = component.ops 

config = component.config 

layer_regularizer = config.layer_regularizer 

ksize = [1,filter-1,filter-1,1] 

stride = [1,filter-1,filter-1,1] 

for i in range(config.block_repeat_count-1): 

net = config.activation(net) 

if layer_regularizer is not None: 

net = component.layer_regularizer(net) 

net = ops.conv2d(net, 3, 3, 1, 1, depth) 

print("[discriminator] hidden layer", net) 

net = tf.nn.avg_pool(net, ksize=ksize, strides=stride, padding='SAME') 

print('[discriminator] layer', net) 

return net 

def standard_block(component, net, depth, filter=3): 

ops = component.ops 

config = component.config 

stride_w = filter-1 

stride_h = filter-1 

ksize = [1,filter-1,filter-1,1] 

stride = [1,stride_w,stride_h,1] 

net = ops.conv2d(net, filter, filter, 1, 1, depth) 

net = tf.nn.avg_pool(net, ksize=ksize, strides=stride, padding='SAME') 

print('[discriminator] layer', net) 

return net 

def strided_block(component, net, depth, filter=3): 

ops = component.ops 

config = component.config 

net = ops.conv2d(net, filter, filter, 2, 2, depth) 

print('[discriminator] layer', net) 

return net