all cases use large shape (#22106)

enhanced ops: conv2d, conv3d
elementwise_pow: change to a reasonable shape

python/paddle/fluid/tests/unittests/test_conv2d_op.py

@@ -430,8 +430,14 @@ class TestWithStride(TestConv2dOp):
 
 
 class TestWithGroup(TestConv2dOp):
-    def init_group(self):
-        self.groups = 3
+    def init_test_case(self):
+        self.pad = [0, 0]
+        self.stride = [1, 1]
+        self.input_size = [2, 3, 5, 5]  # NCHW
+        self.group = 3
+        assert np.mod(self.input_size[1], self.groups) == 0
+        f_c = self.input_size[1] // self.groups
+        self.filter_size = [18, f_c, 3, 3]
 
 
 class TestWith1x1(TestConv2dOp):
@@ -441,7 +447,7 @@ class TestWith1x1(TestConv2dOp):
         self.input_size = [2, 3, 5, 5]  # NCHW
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 1, 1]
+        self.filter_size = [120, f_c, 1, 1]
 
     def init_group(self):
         self.groups = 3
@@ -454,7 +460,7 @@ class TestWithDepthWise3x3(TestConv2dOp):
         self.input_size = [3, 4, 10, 10]  # NCHW
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [8, f_c, 3, 3]
+        self.filter_size = [12, f_c, 3, 3]
 
     def init_dilation(self):
         self.dilations = [2, 2]
@@ -496,7 +502,7 @@ class TestWithDilation(TestConv2dOp):
         self.input_size = [2, 3, 10, 10]  # NCHW
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [12, f_c, 3, 3]
 
     def init_dilation(self):
         self.dilations = [2, 2]
@@ -509,10 +515,10 @@ class TestWithInput1x1Filter1x1(TestConv2dOp):
     def init_test_case(self):
         self.pad = [0, 0]
         self.stride = [1, 1]
-        self.input_size = [2, 3, 1, 1]  # NCHW
+        self.input_size = [100, 3, 1, 1]  # NCHW
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 1, 1]
+        self.filter_size = [120, f_c, 1, 1]
 
     def init_group(self):
         self.groups = 3
@@ -548,7 +554,7 @@ class TestDepthwiseConv(TestConv2dOp):
         self.groups = 3
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [3, f_c, 3, 3]
+        self.filter_size = [12, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
 
@@ -561,7 +567,7 @@ class TestDepthwiseConv2(TestConv2dOp):
         self.groups = 3
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [3, f_c, 3, 3]
+        self.filter_size = [12, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
 
@@ -574,7 +580,7 @@ class TestDepthwiseConv3(TestConv2dOp):
         self.groups = 3
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
 
@@ -588,7 +594,7 @@ class TestDepthwiseConvWithDilation(TestConv2dOp):
         self.dilations = [2, 2]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
 
@@ -602,7 +608,7 @@ class TestDepthwiseConvWithDilation2(TestConv2dOp):
         self.dilations = [2, 2]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
 
@@ -616,7 +622,7 @@ class TestDepthwiseConvandFuse(TestConv2dOp):
         self.groups = 3
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [3, f_c, 3, 3]
+        self.filter_size = [12, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
 
@@ -630,7 +636,7 @@ class TestDepthwiseConv2andFuse(TestConv2dOp):
         self.groups = 3
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [3, f_c, 3, 3]
+        self.filter_size = [12, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
 
@@ -644,7 +650,7 @@ class TestDepthwiseConv3andFuse(TestConv2dOp):
         self.groups = 3
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
 
@@ -659,7 +665,7 @@ class TestDepthwiseConvWithDilationandFuse(TestConv2dOp):
         self.dilations = [2, 2]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
 
@@ -674,7 +680,7 @@ class TestDepthwiseConvWithDilation2andFuse(TestConv2dOp):
         self.dilations = [2, 2]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
 
@@ -878,8 +884,14 @@ class TestWithStride_AsyPadding(TestConv2dOp_v2):
 
 
 class TestWithGroup_AsyPadding(TestConv2dOp_v2):
-    def init_group(self):
-        self.groups = 3
+    def init_test_case(self):
+        self.pad = [0, 0]
+        self.stride = [1, 2]
+        self.input_size = [2, 3, 5, 5]  # NCHW
+        self.group = 3
+        assert np.mod(self.input_size[1], self.groups) == 0
+        f_c = self.input_size[1] // self.groups
+        self.filter_size = [24, f_c, 4, 3]
 
 
 class TestWith1x1_AsyPadding(TestConv2dOp_v2):
@@ -888,7 +900,7 @@ class TestWith1x1_AsyPadding(TestConv2dOp_v2):
         self.input_size = [2, 3, 5, 5]  # NCHW
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 1, 1]
+        self.filter_size = [120, f_c, 1, 1]
 
     def init_group(self):
         self.groups = 3
@@ -904,7 +916,7 @@ class TestWithDepthWise3x3_AsyPadding(TestConv2dOp_v2):
         self.input_size = [3, 4, 10, 10]  # NCHW
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [8, f_c, 3, 3]
+        self.filter_size = [16, f_c, 3, 3]
 
     def init_dilation(self):
         self.dilations = [2, 2]
@@ -955,7 +967,7 @@ class TestWithDilation_AsyPadding(TestConv2dOp_v2):
         self.input_size = [2, 3, 10, 10]  # NCHW
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
 
     def init_dilation(self):
         self.dilations = [2, 2]
@@ -971,10 +983,10 @@ class TestWithDilation_AsyPadding(TestConv2dOp_v2):
 class TestWithInput1x1Filter1x1_AsyPadding(TestConv2dOp_v2):
     def init_test_case(self):
         self.stride = [1, 1]
-        self.input_size = [2, 3, 1, 1]  # NCHW
+        self.input_size = [40, 3, 1, 1]  # NCHW
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 1, 1]
+        self.filter_size = [120, f_c, 1, 1]
 
     def init_group(self):
         self.groups = 3
@@ -1000,7 +1012,7 @@ class TestDepthwiseConv_AsyPadding(TestConv2dOp_v2):
         self.groups = 3
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [3, f_c, 3, 3]
+        self.filter_size = [12, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
     def init_paddings(self):
@@ -1016,7 +1028,7 @@ class TestDepthwiseConv2_AsyPadding(TestConv2dOp_v2):
         self.groups = 3
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [3, f_c, 3, 3]
+        self.filter_size = [12, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
     def init_paddings(self):
@@ -1032,7 +1044,7 @@ class TestDepthwiseConv3_AsyPadding(TestConv2dOp_v2):
         self.groups = 3
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
     def init_paddings(self):
@@ -1050,7 +1062,7 @@ class TestDepthwiseConvWithDilation_AsyPadding(TestConv2dOp_v2):
         self.dilations = [2, 2]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
     def init_paddings(self):
@@ -1068,7 +1080,7 @@ class TestDepthwiseConvWithDilation2_AsyPadding(TestConv2dOp_v2):
         self.dilations = [2, 2]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
     def init_paddings(self):
@@ -1086,7 +1098,7 @@ class TestDepthwiseConvandFuse_AsyPadding(TestConv2dOp_v2):
         self.groups = 3
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [3, f_c, 3, 3]
+        self.filter_size = [12, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
     def init_paddings(self):
@@ -1104,7 +1116,7 @@ class TestDepthwiseConv2andFuse_AsyPadding(TestConv2dOp_v2):
         self.groups = 3
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [3, f_c, 3, 3]
+        self.filter_size = [12, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
     def init_paddings(self):
@@ -1122,7 +1134,7 @@ class TestDepthwiseConv3andFuse_AsyPadding(TestConv2dOp_v2):
         self.groups = 3
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
     def init_paddings(self):
@@ -1141,7 +1153,7 @@ class TestDepthwiseConvWithDilationandFuse_AsyPadding(TestConv2dOp_v2):
         self.dilations = [2, 2]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
     def init_paddings(self):
@@ -1160,7 +1172,7 @@ class TestDepthwiseConvWithDilation2andFuse_AsyPadding(TestConv2dOp_v2):
         self.dilations = [2, 2]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 3, 3]
+        self.filter_size = [24, f_c, 3, 3]
         self.op_type = "depthwise_conv2d"
 
     def init_paddings(self):

python/paddle/fluid/tests/unittests/test_conv3d_op.py

@@ -361,7 +361,7 @@ class TestWith1x1(TestConv3dOp):
         self.input_size = [2, 3, 4, 4, 4]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 1, 1, 1]
+        self.filter_size = [120, f_c, 1, 1, 1]
 
     def init_dilation(self):
         self.dilations = [1, 1, 1]
@@ -374,10 +374,10 @@ class TestWithInput1x1Filter1x1(TestConv3dOp):
     def init_test_case(self):
         self.pad = [0, 0, 0]
         self.stride = [1, 1, 1]
-        self.input_size = [2, 3, 1, 1, 1]
+        self.input_size = [40, 3, 1, 1, 1]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 1, 1, 1]
+        self.filter_size = [120, f_c, 1, 1, 1]
 
     def init_dilation(self):
         self.dilations = [1, 1, 1]
@@ -393,7 +393,7 @@ class TestWithDilation(TestConv3dOp):
         self.input_size = [2, 3, 6, 6, 6]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 2, 2, 2]
+        self.filter_size = [24, f_c, 2, 2, 2]
 
     def init_dilation(self):
         self.dilations = [2, 2, 2]
@@ -699,7 +699,7 @@ class TestWith1x1_AsyPadding(TestConv3dOp_2):
         self.input_size = [2, 3, 4, 4, 4]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 1, 1, 1]
+        self.filter_size = [120, f_c, 1, 1, 1]
 
     def init_dilation(self):
         self.dilations = [1, 1, 1]
@@ -718,7 +718,7 @@ class TestWithDilation_AsyPadding(TestConv3dOp_2):
         self.input_size = [2, 3, 6, 6, 6]
         assert np.mod(self.input_size[1], self.groups) == 0
         f_c = self.input_size[1] // self.groups
-        self.filter_size = [6, f_c, 2, 2, 2]
+        self.filter_size = [24, f_c, 2, 2, 2]
 
     def init_dilation(self):
         self.dilations = [2, 2, 2]

python/paddle/fluid/tests/unittests/test_elementwise_pow_op.py

@@ -39,8 +39,8 @@ class TestElementwisePowOp_big_shape_1(TestElementwisePowOp):
     def setUp(self):
         self.op_type = "elementwise_pow"
         self.inputs = {
-            'X': np.random.uniform(0.1, 1, [100, 100]).astype("float64"),
-            'Y': np.random.uniform(0.1, 1, [100, 100]).astype("float64")
+            'X': np.random.uniform(0.1, 1, [10, 10]).astype("float64"),
+            'Y': np.random.uniform(0.1, 1, [10, 10]).astype("float64")
         }
         self.outputs = {'Out': np.power(self.inputs['X'], self.inputs['Y'])}
 
@@ -49,8 +49,8 @@ class TestElementwisePowOp_big_shape_2(TestElementwisePowOp):
     def setUp(self):
         self.op_type = "elementwise_pow"
         self.inputs = {
-            'X': np.random.uniform(0.1, 1, [100, 100]).astype("float64"),
-            'Y': np.random.uniform(0.1, 1, [100, 100]).astype("float64") * 20
+            'X': np.random.uniform(0.1, 1, [10, 10]).astype("float64"),
+            'Y': np.random.uniform(0.1, 1, [10, 10]).astype("float64") * 20
         }
         self.outputs = {'Out': np.power(self.inputs['X'], self.inputs['Y'])}