def test_model(num_classes): bounds = (0, 255) channels = num_classes def mean_brightness_net(images): logits = mx.symbol.mean(images, axis=(2, 3)) return logits images = mx.symbol.Variable('images') logits = mean_brightness_net(images) model = MXNetModel( images, logits, {}, ctx=mx.cpu(), num_classes=num_classes, bounds=bounds, channel_axis=1) test_images = np.random.rand(2, channels, 5, 5).astype(np.float32) test_label = 7 # Tests assert model.forward(test_images).shape \ == (2, num_classes) test_logits = model.forward_one(test_images[0]) assert test_logits.shape == (num_classes,) test_gradient = model.gradient_one(test_images[0], test_label) assert test_gradient.shape == test_images[0].shape np.testing.assert_almost_equal( model.forward_and_gradient_one(test_images[0], test_label)[0], test_logits) np.testing.assert_almost_equal( model.forward_and_gradient_one(test_images[0], test_label)[1], test_gradient) assert model.num_classes() == num_classes
def test_model_gradient(num_classes): bounds = (0, 255) channels = num_classes def mean_brightness_net(images): logits = mx.symbol.mean(images, axis=(2, 3)) return logits images = mx.symbol.Variable("images") logits = mean_brightness_net(images) preprocessing = ( np.arange(num_classes)[:, None, None], np.random.uniform(size=(channels, 5, 5)) + 1, ) model = MXNetModel( images, logits, {}, ctx=mx.cpu(), num_classes=num_classes, bounds=bounds, preprocessing=preprocessing, channel_axis=1, ) test_images = np.random.rand(2, channels, 5, 5).astype(np.float32) test_image = test_images[0] test_label = 7 epsilon = 1e-2 _, g1 = model.forward_and_gradient_one(test_image, test_label) l1 = model._loss_fn(test_image - epsilon / 2 * g1, test_label) l2 = model._loss_fn(test_image + epsilon / 2 * g1, test_label) assert 1e4 * (l2 - l1) > 1 # make sure that gradient is numerically correct np.testing.assert_array_almost_equal(1e4 * (l2 - l1), 1e4 * epsilon * np.linalg.norm(g1)**2, decimal=1)