replace cross_entropy in python/paddle/fluid/tests/unittests/*/*.py except unittests/*.py (#48920)

python/paddle/fluid/tests/unittests/asp/asp_pruning_base.py

@@ -60,7 +60,12 @@ class TestASPHelperPruningBase(unittest.TestCase):
     def run_training_pruning_test(self, get_mask_gen_func, get_mask_check_func):
         with fluid.program_guard(self.main_program, self.startup_program):
             loss = paddle.mean(
-                fluid.layers.cross_entropy(input=self.predict, label=self.label)
+                paddle.nn.functional.cross_entropy(
+                    input=self.predict,
+                    label=self.label,
+                    reduction='none',
+                    use_softmax=False,
+                )
             )
             optimizer = paddle.incubate.asp.decorate(
                 fluid.optimizer.SGD(learning_rate=0.01)

python/paddle/fluid/tests/unittests/asp/test_asp_customized_pruning.py

@@ -269,7 +269,12 @@ class TestASPStaticCustomerizedPruneFunc(unittest.TestCase):
     def test_training_pruning(self):
         with fluid.program_guard(self.main_program, self.startup_program):
             loss = paddle.mean(
-                fluid.layers.cross_entropy(input=self.predict, label=self.label)
+                paddle.nn.functional.cross_entropy(
+                    input=self.predict,
+                    label=self.label,
+                    reduction='none',
+                    use_softmax=False,
+                )
             )
             optimizer = sparsity.decorate(
                 fluid.optimizer.SGD(learning_rate=0.01)

python/paddle/fluid/tests/unittests/asp/test_asp_optimize_static.py

@@ -45,7 +45,12 @@ class TestASPStaticOptimize(unittest.TestCase):
         with fluid.program_guard(self.main_program, self.startup_program):
             self.img, self.label, predict = build_model()
             self.loss = paddle.mean(
-                fluid.layers.cross_entropy(input=predict, label=self.label)
+                paddle.nn.functional.cross_entropy(
+                    input=predict,
+                    label=self.label,
+                    reduction='none',
+                    use_softmax=False,
+                )
             )
             self.optimizer = fluid.optimizer.SGD(learning_rate=0.01)
 

python/paddle/fluid/tests/unittests/asp/test_asp_pruning_static.py

@@ -65,7 +65,12 @@ class TestASPStaticPruningBase(unittest.TestCase):
     def test_training_pruning(self):
         with fluid.program_guard(self.main_program, self.startup_program):
             loss = paddle.mean(
-                fluid.layers.cross_entropy(input=self.predict, label=self.label)
+                paddle.nn.functional.cross_entropy(
+                    input=self.predict,
+                    label=self.label,
+                    reduction='none',
+                    use_softmax=False,
+                )
             )
             optimizer = paddle.incubate.asp.decorate(
                 fluid.optimizer.SGD(learning_rate=0.01)

python/paddle/fluid/tests/unittests/asp/test_asp_save_load.py

@@ -146,7 +146,12 @@ class TestASPStaticOptimize(unittest.TestCase):
         with fluid.program_guard(self.main_program, self.startup_program):
             self.img, self.label, predict = build_model()
             self.loss = paddle.mean(
-                fluid.layers.cross_entropy(input=predict, label=self.label)
+                paddle.nn.functional.cross_entropy(
+                    input=predict,
+                    label=self.label,
+                    reduction='none',
+                    use_softmax=False,
+                )
             )
             self.optimizer = fluid.optimizer.SGD(learning_rate=0.01)
             self.optimizer = paddle.incubate.asp.decorate(self.optimizer)

python/paddle/fluid/tests/unittests/asp/test_fleet_with_asp_sharding.py

@@ -60,7 +60,12 @@ class TestFleetWithASPSharding(unittest.TestCase):
             fc_3 = fluid.layers.fc(input=fc_2, size=64, act='tanh')
             fc_4 = fluid.layers.fc(input=fc_3, size=64, act='tanh')
             prediction = fluid.layers.fc(input=fc_4, size=2, act='softmax')
-            cost = fluid.layers.cross_entropy(input=prediction, label=input_y)
+            cost = paddle.nn.functional.cross_entropy(
+                input=prediction,
+                label=input_y,
+                reduction='none',
+                use_softmax=False,
+            )
             avg_cost = paddle.mean(x=cost)
 
             dist_strategy = paddle.distributed.fleet.DistributedStrategy()

python/paddle/fluid/tests/unittests/asp/test_fleet_with_asp_static.py

@@ -49,7 +49,12 @@ class TestFleetWithASPStatic(unittest.TestCase):
 
             fc_1 = fluid.layers.fc(input=input_x, size=64, act='tanh')
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
-            cost = fluid.layers.cross_entropy(input=prediction, label=input_y)
+            cost = paddle.nn.functional.cross_entropy(
+                input=prediction,
+                label=input_y,
+                reduction='none',
+                use_softmax=False,
+            )
             avg_cost = paddle.mean(x=cost)
 
             strategy = paddle.distributed.fleet.DistributedStrategy()
@@ -122,7 +127,12 @@ class TestFleetWithASPAMPStatic(unittest.TestCase):
 
             fc_1 = fluid.layers.fc(input=input_x, size=64, act='tanh')
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
-            cost = fluid.layers.cross_entropy(input=prediction, label=input_y)
+            cost = paddle.nn.functional.cross_entropy(
+                input=prediction,
+                label=input_y,
+                reduction='none',
+                use_softmax=False,
+            )
             avg_cost = paddle.mean(x=cost)
 
             strategy = paddle.distributed.fleet.DistributedStrategy()

python/paddle/fluid/tests/unittests/dygraph_to_static/ifelse_simple_func.py

@@ -22,7 +22,9 @@ def add_fn(x):
 
 
 def loss_fn(x, lable):
-    loss = fluid.layers.cross_entropy(x, lable)
+    loss = paddle.nn.functional.cross_entropy(
+        x, lable, reduction='none', use_softmax=False
+    )
     return loss
 
 
@@ -45,7 +47,9 @@ def dyfunc_with_if_else(x_v, label=None):
         x_v = x_v + 1
     # plain if in python
     if label is not None:
-        loss = fluid.layers.cross_entropy(x_v, label)
+        loss = paddle.nn.functional.cross_entropy(
+            x_v, label, reduction='none', use_softmax=False
+        )
         return loss
     return x_v
 
@@ -302,7 +306,9 @@ def if_with_and_or(x_v, label=None):
         x_v = x_v + 1
 
     if label is not None:
-        loss = fluid.layers.cross_entropy(x_v, label)
+        loss = paddle.nn.functional.cross_entropy(
+            x_v, label, reduction='none', use_softmax=False
+        )
         return loss
     return x_v
 

python/paddle/fluid/tests/unittests/dygraph_to_static/test_mnist.py

@@ -107,7 +107,9 @@ class MNIST(fluid.dygraph.Layer):
         x = self.inference(inputs)
         if label is not None:
             acc = paddle.static.accuracy(input=x, label=label)
-            loss = fluid.layers.cross_entropy(x, label)
+            loss = paddle.nn.functional.cross_entropy(
+                x, label, reduction='none', use_softmax=False
+            )
             avg_loss = paddle.mean(loss)
 
             return x, acc, avg_loss

python/paddle/fluid/tests/unittests/dygraph_to_static/test_program_translator.py

@@ -109,7 +109,9 @@ class StaticCode1:
 
         def true_fn_1():
             nonlocal __return_0, __return_1, __return_value_0, loss
-            loss = fluid.layers.cross_entropy(x_v, label)
+            loss = paddle.nn.functional.cross_entropy(
+                x_v, label, reduction='none', use_softmax=False
+            )
             __return_0 = _jst.create_bool_as_type(label is not None, True)
             __return_value_0 = loss
             return
@@ -178,7 +180,9 @@ class StaticCode2:
 
         def true_fn_3():
             nonlocal __return_2, __return_3, __return_value_1, loss
-            loss = fluid.layers.cross_entropy(x_v, label)
+            loss = paddle.nn.functional.cross_entropy(
+                x_v, label, reduction='none', use_softmax=False
+            )
             __return_2 = _jst.create_bool_as_type(label is not None, True)
             __return_value_1 = loss
             return

python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet.py

@@ -272,7 +272,12 @@ class ResNetHelper:
                     img, label = data
 
                     pred = resnet(img)
-                    loss = fluid.layers.cross_entropy(input=pred, label=label)
+                    loss = paddle.nn.functional.cross_entropy(
+                        input=pred,
+                        label=label,
+                        reduction='none',
+                        use_softmax=False,
+                    )
                     avg_loss = paddle.mean(x=loss)
                     acc_top1 = paddle.static.accuracy(
                         input=pred, label=label, k=1

python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet_amp.py

@@ -74,7 +74,12 @@ def train(to_static, build_strategy=None):
                     # FIXME(Aurelius84): The followding cross_entropy seems to bring out a
                     # precision problem, need to figure out the underlying reason.
                     # If we remove it, the loss between dygraph and dy2stat is exactly same.
-                    loss = fluid.layers.cross_entropy(input=pred, label=label)
+                    loss = paddle.nn.functional.cross_entropy(
+                        input=pred,
+                        label=label,
+                        reduction='none',
+                        use_softmax=False,
+                    )
                 avg_loss = paddle.mean(x=pred)
                 acc_top1 = paddle.static.accuracy(input=pred, label=label, k=1)
                 acc_top5 = paddle.static.accuracy(input=pred, label=label, k=5)

python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet_pure_fp16.py

@@ -75,7 +75,9 @@ def train(to_static, build_strategy=None):
                 level='O2',
             ):
                 pred = resnet(img)
-                loss = fluid.layers.cross_entropy(input=pred, label=label)
+                loss = paddle.nn.functional.cross_entropy(
+                    input=pred, label=label, reduction='none', use_softmax=False
+                )
             avg_loss = paddle.mean(x=pred)
             acc_top1 = paddle.static.accuracy(input=pred, label=label, k=1)
             acc_top5 = paddle.static.accuracy(input=pred, label=label, k=5)

python/paddle/fluid/tests/unittests/dygraph_to_static/test_se_resnet.py

@@ -340,7 +340,9 @@ class SeResNeXt(fluid.dygraph.Layer):
         out = self.out(y)
 
         softmax_out = paddle.nn.functional.softmax(out)
-        loss = fluid.layers.cross_entropy(input=softmax_out, label=label)
+        loss = paddle.nn.functional.cross_entropy(
+            input=softmax_out, label=label, reduction='none', use_softmax=False
+        )
         avg_loss = paddle.mean(x=loss)
 
         acc_top1 = paddle.static.accuracy(input=softmax_out, label=label, k=1)

python/paddle/fluid/tests/unittests/dygraph_to_static/test_sentiment.py

@@ -106,7 +106,9 @@ class CNN(fluid.dygraph.Layer):
         prediction = self._fc_prediction(fc_1)
         prediction = self._fc1_act(prediction)
 
-        cost = fluid.layers.cross_entropy(input=prediction, label=label)
+        cost = paddle.nn.functional.cross_entropy(
+            input=prediction, label=label, reduction='none', use_softmax=False
+        )
         avg_cost = paddle.mean(x=cost)
         acc = paddle.static.accuracy(input=prediction, label=label)
         return avg_cost, prediction, acc
@@ -149,7 +151,9 @@ class BOW(fluid.dygraph.Layer):
         prediction = self._fc_prediction(fc_2)
         prediction = paddle.nn.functional.softmax(prediction)
 
-        cost = fluid.layers.cross_entropy(input=prediction, label=label)
+        cost = paddle.nn.functional.cross_entropy(
+            input=prediction, label=label, reduction='none', use_softmax=False
+        )
         avg_cost = paddle.mean(x=cost)
         acc = paddle.static.accuracy(input=prediction, label=label)
         return avg_cost, prediction, acc
@@ -195,7 +199,9 @@ class GRU(fluid.dygraph.Layer):
         fc_2 = paddle.tanh(fc_2)
         prediction = self._fc_prediction(fc_2)
         prediction = paddle.nn.functional.softmax(prediction)
-        cost = fluid.layers.cross_entropy(input=prediction, label=label)
+        cost = paddle.nn.functional.cross_entropy(
+            input=prediction, label=label, reduction='none', use_softmax=False
+        )
         avg_cost = paddle.mean(x=cost)
         acc = paddle.static.accuracy(input=prediction, label=label)
         return avg_cost, prediction, acc
@@ -254,7 +260,9 @@ class BiGRU(fluid.dygraph.Layer):
         prediction = paddle.nn.functional.softmax(prediction)
         # TODO(Aurelius84): Uncomment the following codes when we support return variable-length vars.
         # if label is not None:
-        cost = fluid.layers.cross_entropy(input=prediction, label=label)
+        cost = paddle.nn.functional.cross_entropy(
+            input=prediction, label=label, reduction='none', use_softmax=False
+        )
         avg_cost = paddle.mean(x=cost)
         acc = paddle.static.accuracy(input=prediction, label=label)
         return avg_cost, prediction, acc

python/paddle/fluid/tests/unittests/ipu/test_dy2static_fp16_ipu.py

@@ -34,7 +34,9 @@ class SimpleLayer(paddle.nn.Layer):
         x = paddle.flatten(x, 1, -1)
         if target is not None:
             x = paddle.nn.functional.softmax(x)
-            loss = paddle.fluid.layers.cross_entropy(x, target)
+            loss = paddle.paddle.nn.functional.cross_entropy(
+                x, target, reduction='none', use_softmax=False
+            )
             if self.use_ipu:
                 loss = paddle.incubate.identity_loss(loss, 1)
             else:

python/paddle/fluid/tests/unittests/ipu/test_dy2static_ipu.py

@@ -52,7 +52,9 @@ class SimpleLayer(paddle.nn.Layer):
             if self.loss_op:
                 loss = self.loss_op(x, target)
             else:
-                loss = paddle.fluid.layers.cross_entropy(x, target)
+                loss = paddle.paddle.nn.functional.cross_entropy(
+                    x, target, reduction='none', use_softmax=False
+                )
             if self.use_reduction:
                 loss = paddle.mean(loss)
             if self.use_identity_loss:

python/paddle/fluid/tests/unittests/ipu/test_modelruntime_ipu.py

@@ -33,7 +33,9 @@ class SimpleLayer(paddle.nn.Layer):
         x = paddle.flatten(x, 1, -1)
         if target is not None:
             x = paddle.nn.functional.softmax(x)
-            loss = paddle.fluid.layers.cross_entropy(x, target)
+            loss = paddle.paddle.nn.functional.cross_entropy(
+                x, target, reduction='none', use_softmax=False
+            )
             return x, loss
         return x
 

python/paddle/fluid/tests/unittests/ipu/test_print_op_ipu.py

@@ -120,7 +120,9 @@ class SimpleLayer(paddle.nn.Layer):
         x = paddle.flatten(x, 1, -1)
         if target is not None:
             x = paddle.nn.functional.softmax(x)
-            loss = paddle.fluid.layers.cross_entropy(x, target)
+            loss = paddle.paddle.nn.functional.cross_entropy(
+                x, target, reduction='none', use_softmax=False
+            )
             loss = paddle.incubate.identity_loss(loss, 1)
             return x, loss
         return x

python/paddle/fluid/tests/unittests/ir/test_ir_subgraph_python_interface.py

@@ -35,7 +35,9 @@ class TestQuantizationSubGraph(unittest.TestCase):
             hidden = data
             for _ in range(num):
                 hidden = fluid.layers.fc(hidden, size=128, act='relu')
-            loss = fluid.layers.cross_entropy(input=hidden, label=label)
+            loss = paddle.nn.functional.cross_entropy(
+                input=hidden, label=label, reduction='none', use_softmax=False
+            )
             loss = paddle.mean(loss)
             return loss
 

python/paddle/fluid/tests/unittests/mlu/test_adam_op_mlu.py

@@ -263,7 +263,7 @@ class TestNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             adam = fluid.optimizer.Adam(learning_rate=0.01)
             adam.minimize(loss)

python/paddle/fluid/tests/unittests/mlu/test_adamw_op_mlu.py

@@ -214,7 +214,7 @@ class TestNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             adam = paddle.optimizer.AdamW(learning_rate=0.01, weight_decay=0.02)
             adam.minimize(loss)

python/paddle/fluid/tests/unittests/mlu/test_elementwise_max_op_mlu.py

@@ -343,7 +343,7 @@ class TestElementwiseMaxNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=c, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/mlu/test_elementwise_min_op_mlu.py

@@ -189,7 +189,7 @@ class TestElementwiseMinOpNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=c, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/mlu/test_gelu_op_mlu.py

@@ -112,7 +112,7 @@ class TestGeluNet(unittest.TestCase):
             fc_1_gelu = paddle.nn.functional.gelu(fc_1)
             prediction = fluid.layers.fc(input=fc_1_gelu, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/mlu/test_leaky_relu_op_mlu.py

@@ -106,7 +106,7 @@ class TestLeakyReluNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=y, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/mlu/test_relu6_op_mlu.py

@@ -125,7 +125,7 @@ class TestRelu6Net(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/mlu/test_relu_op_mlu.py

@@ -126,7 +126,7 @@ class TestReluNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/mlu/test_tanh_op_mlu.py

@@ -107,7 +107,7 @@ class TestTanhNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=d, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_adam_op_npu.py

@@ -263,7 +263,7 @@ class TestNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             adam = fluid.optimizer.Adam(learning_rate=0.01)
             adam.minimize(loss)
@@ -348,7 +348,7 @@ class TestNetWithEpsilonTensor(unittest.TestCase):
                     input=fc_1, size=2, param_attr=weight_attr2, act='softmax'
                 )
 
-                cost = fluid.layers.cross_entropy(input=prediction, label=label)
+                cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
                 loss = paddle.mean(cost)
                 beta1_init = 0.9
                 beta2_init = 0.999

python/paddle/fluid/tests/unittests/npu/test_adamw_op_npu.py

@@ -214,7 +214,7 @@ class TestNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             adam = paddle.optimizer.AdamW(learning_rate=0.01, weight_decay=0.02)
             adam.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_cos_op_npu.py

@@ -104,7 +104,7 @@ class TestCosNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=d, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_elementwise_div_op_npu.py

@@ -138,7 +138,7 @@ class TestElementwiseDivNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=g, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_elementwise_max_op_npu.py

@@ -302,7 +302,7 @@ class TestElementwiseMaxNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=c, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_elementwise_min_op_npu.py

@@ -189,7 +189,7 @@ class TestElementwiseMinOpNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=c, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_elementwise_pow_op_npu.py

@@ -313,7 +313,7 @@ class TestElementwisePowNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=c, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_elementwise_sub_op_npu.py

@@ -194,7 +194,7 @@ class TestSubtractNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_gelu_op_npu.py

@@ -112,7 +112,7 @@ class TestGeluNet(unittest.TestCase):
             fc_1_gelu = paddle.nn.functional.gelu(fc_1)
             prediction = fluid.layers.fc(input=fc_1_gelu, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_leaky_relu_op_npu.py

@@ -106,7 +106,7 @@ class TestLeakyReluNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=y, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_log_op_npu.py

@@ -104,7 +104,7 @@ class TestLogNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=d, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_mul_op_npu.py

@@ -247,7 +247,7 @@ class TestMulNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=result, size=8)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)
@@ -324,7 +324,7 @@ class TestMulNet3_2(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=result, size=8)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)
@@ -404,7 +404,7 @@ class TestMulNet3_2_xc2(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=result_re, size=8)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)
@@ -485,7 +485,7 @@ class TestMulNet4_2(unittest.TestCase):
 
             prediction = fluid.layers.fc(input=result, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_pow_op_npu.py

@@ -104,7 +104,7 @@ class TestPowNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_reduce_sum_op_npu.py

@@ -112,7 +112,7 @@ class TestReduceSumNet(unittest.TestCase):
 
             prediction = fluid.layers.fc(input=z_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_relu6_op_npu.py

@@ -125,7 +125,7 @@ class TestRelu6Net(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_relu_op_npu.py

@@ -118,7 +118,7 @@ class TestReluNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_rmsprop_op_npu.py

@@ -52,7 +52,7 @@ class TestNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             rmsprop = fluid.optimizer.RMSProp(learning_rate=0.01)
             rmsprop.minimize(loss)
@@ -115,7 +115,7 @@ class TestCenteredNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             rmsprop = fluid.optimizer.RMSProp(learning_rate=0.01, centered=True)
             rmsprop.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_sgd_op_npu.py

@@ -77,7 +77,7 @@ class TestNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=z, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_softmax_op_npu.py

@@ -81,7 +81,7 @@ class TestSoftmaxNet(unittest.TestCase):
             # 4 x 2
             prob = paddle.nn.functional.softmax(prediction, axis=1)
 
-            cost = fluid.layers.cross_entropy(input=prob, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prob, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_sqrt_op_npu.py

@@ -107,7 +107,7 @@ class TestSqrtNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=d, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_square_op_npu.py

@@ -104,7 +104,7 @@ class TestSquareNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=d, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)

python/paddle/fluid/tests/unittests/npu/test_tanh_op_npu.py

@@ -107,7 +107,7 @@ class TestTanhNet(unittest.TestCase):
             fc_1 = fluid.layers.fc(input=d, size=128)
             prediction = fluid.layers.fc(input=fc_1, size=2, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=prediction, label=label)
+            cost = paddle.nn.functional.cross_entropy(input=prediction, label=label, reduction='none', use_softmax=False)
             loss = paddle.mean(cost)
             sgd = fluid.optimizer.SGD(learning_rate=0.01)
             sgd.minimize(loss)