def run(self, start, end):
rs = []
for filenames, xs, ys, ys_target in load_batches_imagenet_test(batch_size=1,
x_min=MODEL.x_min,
x_max=MODEL.x_max,
x_shape=MODEL.x_shape,
x_dtype=MODEL.x_dtype,
y_dtype=MODEL.y_dtype,
start=start,
end=end,
label_offset=LABEL_OFFSET,
return_target_class=True):
print(filenames)
rs.append(self._run(xs, ys, ys_target))
return rs
def run(self, start, end):
xs_adv = np.zeros((end - start,) + MODEL.x_shape)
idx = 0
for filenames, xs, ys, ys_target in load_batches_imagenet_test(batch_size=1,
x_min=MODEL.x_min,
x_max=MODEL.x_max,
x_shape=MODEL.x_shape,
x_dtype=MODEL.x_dtype,
y_dtype=MODEL.y_dtype,
start=start,
end=end,
label_offset=LABEL_OFFSET,
return_target_class=True):
print(filenames)
xs_adv[idx] = self._run(self, xs, ys, ys_target)
idx += 1
return xs_adv
'alpha': ALPHA_L_2,
'session': SESSION,
})
builder.iteration(ITERATION)
builder.batch_size(BATCH_SIZE)
builder.no_batch_pred(True)
benchmark = builder.build(SESSION, MODEL, args.method, args.goal,
args.distance_metric)
os.makedirs(args.output_dir, exist_ok=True)
for count, (filenames, xs, ys, ys_target) in enumerate(
load_batches_imagenet_test(batch_size=BATCH_SIZE,
x_min=MODEL.x_min,
x_max=MODEL.x_max,
x_shape=MODEL.x_shape,
x_dtype=MODEL.x_dtype,
y_dtype=MODEL.y_dtype,
start=0,
end=1000,
label_offset=LABEL_OFFSET,
return_target_class=True)):
print(count * BATCH_SIZE, (count + 1) * BATCH_SIZE)
output_filename = os.path.join(args.output_dir, '%d_rs.npy' % count)
rs = benchmark.run(xs, ys, ys_target)
np.save(output_filename, rs)
SESSION.close()
