Make separable conv backend tests efficient (#9570)

tests/keras/backend/backend_test.py

@@ -257,6 +257,11 @@ def ref_depthwise_conv(x, w, padding, data_format):
     return y
 
 
+def ref_separable_conv(x, w1, w2, padding, data_format):
+    x2 = ref_depthwise_conv(x, w1, padding, data_format)
+    return ref_conv(x2, w2, padding, data_format)
+
+
 def ref_rnn(x, w, init, go_backwards=False, mask=None, unroll=False, input_length=None):
     w_i, w_h, w_o = w
     h = []
@@ -1086,27 +1091,30 @@ class TestBackend(object):
                                        BACKENDS, cntk_dynamicity=True,
                                        data_format=data_format)
 
-    def test_separable_conv2d(self):
-        for (input_shape, data_format) in [
-                ((2, 3, 4, 5), 'channels_first'),
-                ((2, 3, 5, 6), 'channels_first'),
-                ((1, 6, 5, 3), 'channels_last')]:
-            input_depth = input_shape[1] if data_format == 'channels_first' else input_shape[-1]
-            _, x_val = parse_shape_or_val(input_shape)
-            x_tf = KTF.variable(x_val)
-            for kernel_shape in [(2, 2), (4, 3)]:
-                for depth_multiplier in [1, 2]:
-                    _, depthwise_val = parse_shape_or_val(kernel_shape + (input_depth, depth_multiplier))
-                    _, pointwise_val = parse_shape_or_val((1, 1) + (input_depth * depth_multiplier, 7))
-
-                    z_tf = KTF.eval(KTF.separable_conv2d(x_tf, KTF.variable(depthwise_val),
-                                                         KTF.variable(pointwise_val),
-                                                         data_format=data_format))
-                    z_c = cntk_func_three_tensor('separable_conv2d', input_shape,
-                                                 depthwise_val,
-                                                 pointwise_val,
-                                                 data_format=data_format)([x_val])[0]
-                    assert_allclose(z_tf, z_c, 1e-3)
+    @pytest.mark.skipif(K.backend() == 'theano', reason='Not supported.')
+    @pytest.mark.parametrize('op,input_shape,kernel_shape,depth_multiplier,padding,data_format', [
+        ('separable_conv2d', (2, 3, 4, 5), (3, 3), 1, 'same', 'channels_first'),
+        ('separable_conv2d', (2, 3, 5, 6), (4, 3), 2, 'valid', 'channels_first'),
+        ('separable_conv2d', (1, 6, 5, 3), (3, 4), 1, 'valid', 'channels_last'),
+        ('separable_conv2d', (1, 7, 6, 3), (3, 3), 2, 'same', 'channels_last'),
+    ])
+    def test_separable_conv2d(self, op, input_shape, kernel_shape, depth_multiplier, padding, data_format):
+        input_depth = input_shape[1] if data_format == 'channels_first' else input_shape[-1]
+        _, x = parse_shape_or_val(input_shape)
+        _, depthwise = parse_shape_or_val(kernel_shape + (input_depth, depth_multiplier))
+        _, pointwise = parse_shape_or_val((1, 1) + (input_depth * depth_multiplier, 7))
+        y1 = ref_separable_conv(x, depthwise, pointwise, padding, data_format)
+        if K.backend() == 'cntk':
+            y2 = cntk_func_three_tensor(
+                op, input_shape,
+                depthwise, pointwise,
+                padding=padding, data_format=data_format)([x])[0]
+        else:
+            y2 = K.eval(getattr(K, op)(
+                K.variable(x),
+                K.variable(depthwise), K.variable(pointwise),
+                padding=padding, data_format=data_format))
+        assert_allclose(y1, y2, atol=1e-05)
 
     def test_pool2d(self):
         check_single_tensor_operation('pool2d', (5, 10, 12, 3),