def generate_np(self, x_val, **kwargs):
"""
Facilitates testing this attack.
"""
fixed, feedable, hash_key = self.construct_variables(kwargs)
if hash_key not in self.graphs:
with tf.variable_scope(None, 'attack_%d' % len(self.graphs)):
# x is a special placeholder we always want to have
with tf.device('/gpu:0'):
x = tf.placeholder(tf.float32, shape=x_val.shape, name='x')
inputs, outputs = self.generate(x, **kwargs)
from runner import RunnerMultiGPU
runner = RunnerMultiGPU(inputs, outputs, sess=self.sess)
self.graphs[hash_key] = runner
runner = self.graphs[hash_key]
feed_dict = {'x': x_val}
for name in feedable:
feed_dict[name] = feedable[name]
fvals = runner.run(feed_dict)
while not runner.is_finished():
fvals = runner.run()
return fvals['adv_x']
def setUp(self):
super(TestRunnerMultiGPU, self).setUp()
self.sess = tf.Session()
inputs = []
outputs = []
self.niter = 10
niter = self.niter
# A Simple graph with `niter` sub-graphs.
with tf.variable_scope(None, 'runner'):
for i in range(niter):
v = tf.get_variable('v%d' % i, shape=(100, 10))
w = tf.get_variable('w%d' % i, shape=(100, 1))
inputs += [{'v': v, 'w': w}]
outputs += [{'v': v, 'w': w}]
self.runner = RunnerMultiGPU(inputs, outputs, sess=self.sess)
def __init__(self, *args, **kwargs): super(TrainerMultiGPU, self).__init__(*args, **kwargs) from runner import RunnerMultiGPU self.runner = RunnerMultiGPU(self.inputs, self.outputs, sess=self.sess)