eladshabi

# source: https://github.com/hwalsuklee/tensorflow-generative-model-collections/blob/master/ACGAN.py

def generator(self, z, y, is_training=True, reuse=False):
    if self.mixed:
        with tf.variable_scope("generator", reuse=reuse,
			       custom_getter=float32_variable_storage_getter):
            # merge noise and code
            z = concat([z, y], 1)
            net = fc(z, 1024, scope='g_fc1', activation_fn=None)

eladshabi

# source: https://github.com/hwalsuklee/tensorflow-generative-model-collections/blob/master/ACGAN.py

with tf.variable_scope("generator", reuse=reuse):
			
    # merge noise and code
    z = concat([z, y], 1)

    net = fc(z, 1024, scope='g_fc1', activation_fn=None)
    net = bn(net, is_training=is_training, scope='g_bn1')
    net = tf.nn.relu(net)

eladshabi

# source: https://github.com/hwalsuklee/tensorflow-generative-model-collections/blob/master/ops.py

def conv2d(input_, output_dim, k_h=5, k_w=5, d_h=2, d_w=2, stddev=0.02, name="conv2d",
           data_type=tf.float32):
    with tf.variable_scope(name):
        w = tf.get_variable('w', [k_h, k_w, input_.get_shape()[-1], output_dim],
              initializer=tf.truncated_normal_initializer(stddev=stddev),
                            dtype=data_type)
        conv = tf.nn.conv2d(input_, w, strides=[1, d_h, d_w, 1], padding='SAME')

eladshabi

# source: https://docs.nvidia.com/deeplearning/sdk/mixed-precision-training/index.html

with tf.device('/gpu:0'), \
    tf.variable_scope('fp32_storage',custom_getter=float32_variable_storage_getter):
        data, target, loss = create_simple_model(nbatch, nin, nout, dtype)
        variables = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES)
        
	# Traning variables
	lr = 0.0002 
	beta = 0.5

eladshabi

# source: https://github.com/hwalsuklee/tensorflow-generative-model-collections/blob/master/ACGAN.py

with tf.variable_scope("generator", reuse=reuse, custom_getter=float32_variable_storage_getter):
    # merge noise and code
    z = concat([z, y], 1)
    net = fc(z, 1024, scope='g_fc1', activation_fn=None)

    # Batch normalization should be calculated as type of float32
    net = tf.cast(net, tf.float32)
    net = bn(net, is_training=is_training, scope='g_bn1')

eladshabi

# source: https://github.com/hwalsuklee/tensorflow-generative-model-collections/blob/master/ACGAN.py

t_vars = tf.trainable_variables()

d_vars = [var for var in t_vars if 'd_' in var.name]
g_vars = [var for var in t_vars if 'g_' in var.name]
q_vars = [var for var in t_vars if ('d_' in var.name) or ('c_' in var.name) or ('g_' in var.name)]

with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
	

eladshabi

# source: https://docs.nvidia.com/deeplearning/sdk/mixed-precision-training/index.html

def float32_variable_storage_getter(getter, name, shape=None, dtype=None,
                                    initializer=None, regularizer=None,
                                    trainable=True,
                                    *args, **kwargs):

    storage_dtype = tf.float32 if trainable else dtype
    variable = getter(name, shape, dtype=storage_dtype,
                      initializer=initializer, regularizer=regularizer,

eladshabi

# source: https://docs.nvidia.com/deeplearning/sdk/mixed-precision-training/index.html
import tensorflow as tf

def create_simple_model(nbatch, nin, nout, dtype):
    """A simple softmax model."""
    data = tf.placeholder(dtype, shape=(nbatch, nin))
    weights = tf.get_variable('weights', (nin, nout), dtype)
    biases = tf.get_variable('biases', nout, dtype, initializer=tf.zeros_initializer())
    logits = tf.matmul(data, weights) + biases
    target = tf.placeholder(tf.float32, shape=(nbatch, nout))