⬆️ Linting dependencies (#2663)

Author: lgeiger

pyproject.toml

@@ -1,5 +1,5 @@
 [tool.black]
-target-version = ['py36', 'py37', 'py38']
+target-version = ['py37', 'py38', 'py39', 'py310']
 exclude = '''
 (
   /(

tensorflow_addons/image/filters.py

@@ -204,8 +204,8 @@ def _get_gaussian_kernel(sigma, filter_shape):
     """Compute 1D Gaussian kernel."""
     sigma = tf.convert_to_tensor(sigma)
     x = tf.range(-filter_shape // 2 + 1, filter_shape // 2 + 1)
-    x = tf.cast(x ** 2, sigma.dtype)
-    x = tf.nn.softmax(-x / (2.0 * (sigma ** 2)))
+    x = tf.cast(x**2, sigma.dtype)
+    x = tf.nn.softmax(-x / (2.0 * (sigma**2)))
     return x
 
 

tensorflow_addons/losses/metric_learning.py

@@ -31,15 +31,12 @@ def pairwise_distance(feature: TensorLike, squared: bool = False):
     Returns:
       pairwise_distances: 2-D Tensor of size `[number of data, number of data]`.
     """
-    pairwise_distances_squared = (
-        tf.math.add(
-            tf.math.reduce_sum(tf.math.square(feature), axis=[1], keepdims=True),
-            tf.math.reduce_sum(
-                tf.math.square(tf.transpose(feature)), axis=[0], keepdims=True
-            ),
-        )
-        - 2.0 * tf.matmul(feature, tf.transpose(feature))
-    )
+    pairwise_distances_squared = tf.math.add(
+        tf.math.reduce_sum(tf.math.square(feature), axis=[1], keepdims=True),
+        tf.math.reduce_sum(
+            tf.math.square(tf.transpose(feature)), axis=[0], keepdims=True
+        ),
+    ) - 2.0 * tf.matmul(feature, tf.transpose(feature))
 
     # Deal with numerical inaccuracies. Set small negatives to zero.
     pairwise_distances_squared = tf.math.maximum(pairwise_distances_squared, 0.0)

tensorflow_addons/metrics/matthews_correlation_coefficient.py

@@ -102,9 +102,9 @@ def result(self):
         # covariance true-pred
         cov_ytyp = num_correct * num_samples - tf.tensordot(true_sum, pred_sum, axes=1)
         # covariance pred-pred
-        cov_ypyp = num_samples ** 2 - tf.tensordot(pred_sum, pred_sum, axes=1)
+        cov_ypyp = num_samples**2 - tf.tensordot(pred_sum, pred_sum, axes=1)
         # covariance true-true
-        cov_ytyt = num_samples ** 2 - tf.tensordot(true_sum, true_sum, axes=1)
+        cov_ytyt = num_samples**2 - tf.tensordot(true_sum, true_sum, axes=1)
 
         mcc = cov_ytyp / tf.math.sqrt(cov_ytyt * cov_ypyp)
 

tensorflow_addons/optimizers/conditional_gradient.py

@@ -123,7 +123,7 @@ def _prepare_local(self, var_device, var_dtype, apply_state):
 
     @staticmethod
     def _frobenius_norm(m):
-        return tf.reduce_sum(m ** 2) ** 0.5
+        return tf.reduce_sum(m**2) ** 0.5
 
     @staticmethod
     def _top_singular_vector(m):

tensorflow_addons/optimizers/cyclical_learning_rate.py

@@ -309,7 +309,7 @@ def __init__(
             initial_learning_rate=initial_learning_rate,
             maximal_learning_rate=maximal_learning_rate,
             step_size=step_size,
-            scale_fn=lambda x: gamma ** x,
+            scale_fn=lambda x: gamma**x,
             scale_mode=scale_mode,
             name=name,
         )

tensorflow_addons/optimizers/tests/conditional_gradient_test.py

@@ -85,7 +85,7 @@ def loss():
     grads0_0 = 32 * 1.0 + 40 * 2.0
     grads0_1 = 40 * 1.0 + 50 * 2.0
     grads0 = tf.constant([[grads0_0, grads0_1]], dtype=dtype)
-    norm0 = tf.math.reduce_sum(grads0 ** 2) ** 0.5
+    norm0 = tf.math.reduce_sum(grads0**2) ** 0.5
 
     learning_rate = 0.1
     lambda_ = 0.1
@@ -117,8 +117,8 @@ def test_basic_frobenius(dtype, use_resource):
         var1 = tf.Variable([3.0, 4.0], dtype=dtype[0])
     grads0 = tf.constant([0.1, 0.1], dtype=dtype[0])
     grads1 = tf.constant([0.01, 0.01], dtype=dtype[0])
-    norm0 = tf.math.reduce_sum(grads0 ** 2) ** 0.5
-    norm1 = tf.math.reduce_sum(grads1 ** 2) ** 0.5
+    norm0 = tf.math.reduce_sum(grads0**2) ** 0.5
+    norm1 = tf.math.reduce_sum(grads1**2) ** 0.5
 
     def learning_rate():
         return 0.5
@@ -440,7 +440,7 @@ def loss():
 
     # the gradient for this loss function:
     grads0 = tf.constant([[0, 0], [1, 1]], dtype=tf.float32)
-    norm0 = tf.math.reduce_sum(grads0 ** 2) ** 0.5
+    norm0 = tf.math.reduce_sum(grads0**2) ** 0.5
 
     learning_rate = 0.1
     lambda_ = 0.1
@@ -503,8 +503,8 @@ def test_tensor_learning_rate_and_conditional_gradient_frobenius(dtype):
     var1 = tf.Variable([3.0, 4.0], dtype=dtype)
     grads0 = tf.constant([0.1, 0.1], dtype=dtype)
     grads1 = tf.constant([0.01, 0.01], dtype=dtype)
-    norm0 = tf.math.reduce_sum(grads0 ** 2) ** 0.5
-    norm1 = tf.math.reduce_sum(grads1 ** 2) ** 0.5
+    norm0 = tf.math.reduce_sum(grads0**2) ** 0.5
+    norm1 = tf.math.reduce_sum(grads1**2) ** 0.5
     ord = "fro"
     cg_opt = cg_lib.ConditionalGradient(
         learning_rate=tf.constant(0.5), lambda_=tf.constant(0.01), ord=ord
@@ -943,8 +943,8 @@ def test_sharing_frobenius(dtype):
     var1 = tf.Variable([3.0, 4.0], dtype=dtype)
     grads0 = tf.constant([0.1, 0.1], dtype=dtype)
     grads1 = tf.constant([0.01, 0.01], dtype=dtype)
-    norm0 = tf.math.reduce_sum(grads0 ** 2) ** 0.5
-    norm1 = tf.math.reduce_sum(grads1 ** 2) ** 0.5
+    norm0 = tf.math.reduce_sum(grads0**2) ** 0.5
+    norm1 = tf.math.reduce_sum(grads1**2) ** 0.5
     learning_rate = 0.1
     lambda_ = 0.1
     ord = "fro"

tensorflow_addons/optimizers/tests/cyclical_learning_rate_test.py

@@ -99,7 +99,7 @@ def test_exponential_cyclical_learning_rate(serialize):
         non_bounded_value = np.abs(i / 2000.0 - 2 * np.floor(1 + i / (2 * 2000)) + 1)
         expected = initial_learning_rate + (
             maximal_learning_rate - initial_learning_rate
-        ) * np.maximum(0, (1 - non_bounded_value)) * (gamma ** i)
+        ) * np.maximum(0, (1 - non_bounded_value)) * (gamma**i)
         computed = exponential_cyclical_lr(step).numpy()
         np.testing.assert_allclose(computed, expected, 1e-6)
         step += 1

tensorflow_addons/optimizers/tests/weight_decay_optimizers_test.py

@@ -152,9 +152,9 @@ def adamw_update_numpy(
         for v in (learning_rate, beta_1, beta_2, epsilon, weight_decay)
     )
     t = slot_vars.get("t", 0) + 1
-    lr_t = lr * np.sqrt(1 - beta2 ** t) / (1 - beta1 ** t)
+    lr_t = lr * np.sqrt(1 - beta2**t) / (1 - beta1**t)
     slot_vars["m"] = beta1 * slot_vars.get("m", 0) + (1 - beta1) * grad_t
-    slot_vars["v"] = beta2 * slot_vars.get("v", 0) + (1 - beta2) * grad_t ** 2
+    slot_vars["v"] = beta2 * slot_vars.get("v", 0) + (1 - beta2) * grad_t**2
     param_t = param * (1 - wd) - lr_t * slot_vars["m"] / (np.sqrt(slot_vars["v"]) + eps)
     slot_vars["t"] = t
     return param_t, slot_vars

tensorflow_addons/optimizers/tests/yogi_test.py

@@ -59,7 +59,7 @@ def yogi_update_numpy(
     beta1 = np.array(beta1, dtype=param.dtype)
     beta2 = np.array(beta2, dtype=param.dtype)
 
-    alpha_t = alpha * np.sqrt(1 - beta2 ** t) / (1 - beta1 ** t)
+    alpha_t = alpha * np.sqrt(1 - beta2**t) / (1 - beta1**t)
 
     m_t = beta1 * m + (1 - beta1) * g_t
     g2_t = g_t * g_t
@@ -128,8 +128,8 @@ def do_test_sparse(beta1=0.0, l1reg=0.0, l2reg=0.0):
         # Run 3 steps of Yogi.
         for t in range(1, 4):
             beta1_power, beta2_power = get_beta_accumulators(opt, dtype)
-            test_utils.assert_allclose_according_to_type(beta1 ** t, beta1_power)
-            test_utils.assert_allclose_according_to_type(0.999 ** t, beta2_power)
+            test_utils.assert_allclose_according_to_type(beta1**t, beta1_power)
+            test_utils.assert_allclose_according_to_type(0.999**t, beta2_power)
             opt.apply_gradients(zip([grads0, grads1], [var0, var1]))
 
             var0_np, m0, v0 = yogi_update_numpy(
@@ -224,8 +224,8 @@ def do_test_basic(beta1=0.0, l1reg=0.0, l2reg=0.0):
         # Run 3 steps of Yogi.
         for t in range(1, 4):
             beta1_power, beta2_power = get_beta_accumulators(opt, dtype)
-            test_utils.assert_allclose_according_to_type(beta1 ** t, beta1_power)
-            test_utils.assert_allclose_according_to_type(0.999 ** t, beta2_power)
+            test_utils.assert_allclose_according_to_type(beta1**t, beta1_power)
+            test_utils.assert_allclose_according_to_type(0.999**t, beta2_power)
 
             opt.apply_gradients(zip([grads0, grads1], [var0, var1]))
 
@@ -284,8 +284,8 @@ def test_tensor_learning_rate():
         # Run 3 steps of Yogi.
         for t in range(1, 4):
             beta1_power, beta2_power = get_beta_accumulators(opt, dtype)
-            test_utils.assert_allclose_according_to_type(0.9 ** t, beta1_power)
-            test_utils.assert_allclose_according_to_type(0.999 ** t, beta2_power)
+            test_utils.assert_allclose_according_to_type(0.9**t, beta1_power)
+            test_utils.assert_allclose_according_to_type(0.999**t, beta2_power)
 
             opt.apply_gradients(zip([grads0, grads1], [var0, var1]))
 
@@ -320,8 +320,8 @@ def test_sharing():
         # Run 3 steps of intertwined Yogi1 and Yogi2.
         for t in range(1, 4):
             beta1_power, beta2_power = get_beta_accumulators(opt, dtype)
-            test_utils.assert_allclose_according_to_type(0.9 ** t, beta1_power)
-            test_utils.assert_allclose_according_to_type(0.999 ** t, beta2_power)
+            test_utils.assert_allclose_according_to_type(0.9**t, beta1_power)
+            test_utils.assert_allclose_according_to_type(0.999**t, beta2_power)
             opt.apply_gradients(zip([grads0, grads1], [var0, var1]))
             var0_np, m0, v0 = yogi_update_numpy(var0_np, grads0_np, t, m0, v0)
             var1_np, m1, v1 = yogi_update_numpy(var1_np, grads1_np, t, m1, v1)