def test_ftz(self):
daz.set_ftz()
np.testing.assert_equal(self.normal / self.scale, 0)
np.testing.assert_equal(self.denormal * self.scale, self.normal)
assert np.all(self.denormal != 0)
daz.unset_ftz()
self.check_normal()
def test_ftz(self):
daz.set_ftz()
assert self.normal / self.scale == 0
assert self.denormal * self.scale == self.normal
assert self.denormal != 0
daz.unset_ftz()
self.check_normal()
),
training=MiniBatchTraining(batch_size=32,
optimizer=AdamOptimizer(0.0001),
listener=CostListener()))
def show_image(i, x, y):
plt.imshow(x[i])
plt.show()
print("y = " + str(np.squeeze(y[:, i])))
if __name__ == "__main__":
import daz
daz.set_ftz()
np.seterr(under='warn')
plt.rcParams['figure.figsize'] = (7.0, 4.0)
plt.rcParams['image.interpolation'] = 'nearest'
plt.rcParams['image.cmap'] = 'gray'
# load image dataset: blue/red dots in circles
train_x_orig, train_y_orig, test_x_orig, test_y_orig, classes = \
load_dataset()
# show_image(0, train_x_orig, train_y_orig)
# Pre-processing
train_x_flat = train_x_orig.reshape(train_x_orig.shape[0], -1).T
test_x_flat = test_x_orig.reshape(test_x_orig.shape[0], -1).T
train_x = train_x_flat / 255.