replace cross_entropy in python/paddle/fluid/tests/unittests/test_[a-n]*.py...

replace cross_entropy in python/paddle/fluid/tests/unittests/test_[a-n]*.py except test_dist_transpiler.py (#48913)

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

@@ -806,7 +806,12 @@ class TestAdamOptimizer(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
@@ -966,7 +971,9 @@ class TestAdamOptimizer(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(use_global_beta_pow=True)
         adam.minimize(loss)

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

@@ -57,7 +57,9 @@ def convolutional_neural_network(use_py_reader):
         )
 
         prediction = fluid.layers.fc(input=conv_pool_2, size=10, act='softmax')
-        loss = fluid.layers.cross_entropy(input=prediction, label=label)
+        loss = paddle.nn.functional.cross_entropy(
+            input=prediction, label=label, reduction='none', use_softmax=False
+        )
         avg_loss = paddle.mean(loss)
         acc = paddle.static.accuracy(input=prediction, label=label)
         i = fluid.layers.zeros(shape=[1], dtype='int64')

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

@@ -107,7 +107,9 @@ class TestCompiledProgramError(unittest.TestCase):
         )
         label = fluid.layers.data(name='label', shape=[1], dtype='int64')
         prediction = fluid.layers.fc(input=img, size=10, act='softmax')
-        loss = fluid.layers.cross_entropy(input=prediction, label=label)
+        loss = paddle.nn.functional.cross_entropy(
+            input=prediction, label=label, reduction='none', use_softmax=False
+        )
         avg_loss = paddle.mean(loss)
 
     def compile_program_not_compiled(self):

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

@@ -17,6 +17,7 @@ import unittest
 import numpy as np
 from op_test import OpTest, randomize_probability
 
+import paddle
 import paddle.fluid as fluid
 import paddle.fluid.core as core
 from paddle.fluid import Program, program_guard
@@ -419,7 +420,9 @@ class TestCrossEntropyOpError(unittest.TestCase):
                 lab1 = fluid.create_lod_tensor(
                     np.array([-1, 3, 5, 5]), [[1, 1, 1, 1]], fluid.CPUPlace()
                 )
-                fluid.layers.cross_entropy(x1, lab1)
+                paddle.nn.functional.cross_entropy(
+                    x1, lab1, reduction='none', use_softmax=False
+                )
 
             self.assertRaises(TypeError, test_Variable)
 
@@ -432,7 +435,9 @@ class TestCrossEntropyOpError(unittest.TestCase):
                 lab2 = fluid.layers.data(
                     name='lab2', shape=[3, 4, 5, 6], dtype="int32"
                 )
-                fluid.layers.cross_entropy(x2, lab2)
+                paddle.nn.functional.cross_entropy(
+                    x2, lab2, reduction='none', use_softmax=False
+                )
 
             self.assertRaises(TypeError, test_dtype)
 

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

@@ -67,7 +67,12 @@ def simple_fc_net(places, use_legacy_py_reader, use_double_buffer):
                 hidden, size=CLASS_NUM, act='softmax'
             )
             loss = paddle.mean(
-                fluid.layers.cross_entropy(input=predict_label, label=label)
+                paddle.nn.functional.cross_entropy(
+                    input=predict_label,
+                    label=label,
+                    reduction='none',
+                    use_softmax=False,
+                )
             )
 
             optimizer = fluid.optimizer.Adam()

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

@@ -73,7 +73,9 @@ def get_model(batch_size):
 
     # Train program
     predict = cnn_model(images)
-    cost = fluid.layers.cross_entropy(input=predict, label=label)
+    cost = paddle.nn.functional.cross_entropy(
+        input=predict, label=label, reduction='none', use_softmax=False
+    )
     avg_cost = paddle.mean(x=cost)
 
     # Evaluator
@@ -188,9 +190,11 @@ class TestCloneWithStopGradient(unittest.TestCase):
             hidden1 = fluid.layers.fc(input=img, size=200, act='relu')
             hidden1.stop_gradient = True
             hidden2 = fluid.layers.dropout(hidden1, dropout_prob=0.5)
-            loss = fluid.layers.cross_entropy(
+            loss = paddle.nn.functional.cross_entropy(
                 input=fluid.layers.fc(hidden2, size=10, act='softmax'),
                 label=fluid.layers.data(name='label', shape=[1], dtype='int64'),
+                reduction='none',
+                use_softmax=False,
             )
             avg_loss = paddle.mean(loss)
             test_program = train_program.clone(for_test=False)
@@ -226,9 +230,11 @@ class TestCloneWithStopGradientInSubBlock(unittest.TestCase):
 
             hidden2 = fluid.layers.cond(cond, true_fn, false_fn)
 
-            loss = fluid.layers.cross_entropy(
+            loss = paddle.nn.functional.cross_entropy(
                 input=fluid.layers.fc(hidden2, size=10, act='softmax'),
                 label=fluid.layers.data(name='label', shape=[1], dtype='int64'),
+                reduction='none',
+                use_softmax=False,
             )
             avg_loss = paddle.mean(loss)
             test_program = train_program.clone(for_test=False)
@@ -266,9 +272,11 @@ class TestCloneWithRaise(unittest.TestCase):
                 return hidden2
 
             hidden2 = fluid.layers.cond(cond, true_fn, false_fn)
-            loss = fluid.layers.cross_entropy(
+            loss = paddle.nn.functional.cross_entropy(
                 input=fluid.layers.fc(hidden2, size=10, act='softmax'),
                 label=fluid.layers.data(name='label', shape=[1], dtype='int64'),
+                reduction='none',
+                use_softmax=False,
             )
             avg_loss = paddle.mean(loss)
             test_program = train_program.clone(for_test=False)

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

@@ -52,8 +52,8 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
         fc_1 = paddle.fluid.layers.fc(input=input_x, size=64, act='tanh')
         fc_2 = paddle.fluid.layers.fc(input=fc_1, size=64, act='tanh')
         prediction = paddle.fluid.layers.fc(input=[fc_2], size=2, act='softmax')
-        cost = paddle.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)
 

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

@@ -68,8 +68,8 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
         fc_1 = paddle.fluid.layers.fc(input=x_embedding, size=64, act='tanh')
         fc_2 = paddle.fluid.layers.fc(input=fc_1, size=64, act='tanh')
         prediction = paddle.fluid.layers.fc(input=[fc_2], size=2, act='softmax')
-        cost = paddle.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)
 

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

@@ -56,8 +56,8 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
         fc_1 = paddle.fluid.layers.fc(input=emb, size=64, act='tanh')
         fc_2 = paddle.fluid.layers.fc(input=fc_1, size=64, act='tanh')
         prediction = paddle.fluid.layers.fc(input=[fc_2], size=2, act='softmax')
-        cost = paddle.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)
         os.environ["FLAGS_LAUNCH_BARRIER"] = "0"

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

@@ -51,8 +51,8 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
         fc_1 = paddle.fluid.layers.fc(input=input_x, size=64, act='tanh')
         fc_2 = paddle.fluid.layers.fc(input=fc_1, size=64, act='tanh')
         prediction = paddle.fluid.layers.fc(input=[fc_2], size=2, act='softmax')
-        cost = paddle.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)
 
@@ -84,8 +84,8 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
         fc_1 = paddle.fluid.layers.fc(input=input_x, size=64, act='tanh')
         fc_2 = paddle.fluid.layers.fc(input=fc_1, size=64, act='tanh')
         prediction = paddle.fluid.layers.fc(input=[fc_2], size=2, act='softmax')
-        cost = paddle.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)
 

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

@@ -150,7 +150,9 @@ class TestDistFleetHeterProgram(unittest.TestCase):
 
         with fluid.device_guard("gpu"):
             labels = fluid.layers.cast(inputs[-1], dtype="int64")
-            cost = fluid.layers.cross_entropy(input=predict, label=labels)
+            cost = paddle.nn.functional.cross_entropy(
+                input=predict, label=labels, reduction='none', use_softmax=False
+            )
             avg_cost = paddle.sum(cost)
 
         return avg_cost

python/paddle/fluid/tests/unittests/test_dygraph_mnist_fp16.py

@@ -112,7 +112,9 @@ class MNIST(fluid.dygraph.Layer):
         x = paddle.reshape(x, shape=[-1, self.pool_2_shape])
         cost = self._linear(x)
         cost = paddle.nn.functional.softmax(cost)
-        loss = fluid.layers.cross_entropy(cost, label)
+        loss = paddle.nn.functional.cross_entropy(
+            cost, label, reduction='none', use_softmax=False
+        )
         avg_loss = paddle.mean(loss)
         return avg_loss
 

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

@@ -138,7 +138,9 @@ class TestDygraphMultiForward(unittest.TestCase):
                     label.stop_gradient = True
 
                     cost = mnist(img)
-                    loss = fluid.layers.cross_entropy(cost, label)
+                    loss = paddle.nn.functional.cross_entropy(
+                        cost, label, reduction='none', use_softmax=False
+                    )
                     avg_loss = paddle.mean(loss)
 
                     dy_out = avg_loss.numpy()
@@ -167,7 +169,9 @@ class TestDygraphMultiForward(unittest.TestCase):
             )
             label = fluid.layers.data(name='label', shape=[1], dtype='int64')
             cost = mnist(img)
-            loss = fluid.layers.cross_entropy(cost, label)
+            loss = paddle.nn.functional.cross_entropy(
+                cost, label, reduction='none', use_softmax=False
+            )
             avg_loss = paddle.mean(loss)
 
             # initialize params and fetch them

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

@@ -45,7 +45,9 @@ def simple_fc_net():
             ),
         )
     prediction = fluid.layers.fc(hidden, size=10, act='softmax')
-    loss = fluid.layers.cross_entropy(input=prediction, label=label)
+    loss = paddle.nn.functional.cross_entropy(
+        input=prediction, label=label, reduction='none', use_softmax=False
+    )
     loss = paddle.mean(loss)
     optimizer = fluid.optimizer.Adam(learning_rate=1e-3)
     optimizer.minimize(loss)

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

@@ -43,7 +43,9 @@ def gru_net(
     gru_max_tanh = paddle.tanh(gru_max)
     fc1 = fluid.layers.fc(input=gru_max_tanh, size=hid_dim2, act='tanh')
     prediction = fluid.layers.fc(input=fc1, size=class_dim, 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
+    )
     avg_cost = paddle.mean(x=cost)
     return avg_cost
 

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

@@ -45,7 +45,9 @@ def lstm_net(
     lstm_max_tanh = paddle.tanh(lstm_max)
     fc1 = fluid.layers.fc(input=lstm_max_tanh, size=hid_dim2, act='tanh')
     prediction = fluid.layers.fc(input=fc1, size=class_dim, 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
+    )
     avg_cost = paddle.mean(x=cost)
     return avg_cost
 

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

@@ -67,7 +67,12 @@ class TestFeedData(unittest.TestCase):
 
         predict_label = fluid.layers.fc(hidden, size=class_num, act='softmax')
         loss = paddle.mean(
-            fluid.layers.cross_entropy(input=predict_label, label=label)
+            paddle.nn.functional.cross_entropy(
+                input=predict_label,
+                label=label,
+                reduction='none',
+                use_softmax=False,
+            )
         )
 
         optimizer = fluid.optimizer.Adam()

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

@@ -46,7 +46,9 @@ class TestFetchUnmerged(unittest.TestCase):
         )
         hidden = fluid.layers.fc(input=conv_pool_2, size=32, act='relu')
         prediction = fluid.layers.fc(input=hidden, size=10, act='softmax')
-        loss = fluid.layers.cross_entropy(input=prediction, label=label)
+        loss = paddle.nn.functional.cross_entropy(
+            input=prediction, label=label, reduction='none', use_softmax=False
+        )
         avg_loss = paddle.mean(loss)
         return avg_loss, prediction
 

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

@@ -40,8 +40,8 @@ class TestDistributedStrategyAuto(unittest.TestCase):
         fc_1 = paddle.fluid.layers.fc(input=input_x, size=64, act='tanh')
         fc_2 = paddle.fluid.layers.fc(input=fc_1, size=64, act='tanh')
         prediction = paddle.fluid.layers.fc(input=[fc_2], size=2, act='softmax')
-        cost = paddle.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)
 

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

@@ -203,7 +203,12 @@ class TestFleetBaseSingleError(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)
             fleet.init(is_collective=True)
 

python/paddle/fluid/tests/unittests/test_fleet_base_2.py

@@ -54,8 +54,8 @@ class TestFleetBase(unittest.TestCase):
         fc_1 = paddle.fluid.layers.fc(input=input_x, size=64, act='tanh')
         fc_2 = paddle.fluid.layers.fc(input=fc_1, size=64, act='tanh')
         prediction = paddle.fluid.layers.fc(input=[fc_2], size=2, act='softmax')
-        cost = paddle.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)
 

python/paddle/fluid/tests/unittests/test_fleet_base_3.py

@@ -40,8 +40,8 @@ class TestFleetBase_1(unittest.TestCase):
         fc_1 = paddle.fluid.layers.fc(input=input_x, size=64, act='tanh')
         fc_2 = paddle.fluid.layers.fc(input=fc_1, size=64, act='tanh')
         prediction = paddle.fluid.layers.fc(input=[fc_2], size=2, act='softmax')
-        cost = paddle.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)
 
@@ -71,8 +71,8 @@ class TestFleetBase(unittest.TestCase):
         fc_1 = paddle.fluid.layers.fc(input=input_x, size=64, act='tanh')
         fc_2 = paddle.fluid.layers.fc(input=fc_1, size=64, act='tanh')
         prediction = paddle.fluid.layers.fc(input=[fc_2], size=2, act='softmax')
-        cost = paddle.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)
 

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

@@ -85,7 +85,9 @@ class TestFleetBaseSingleRunCollective(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)
 
         fleet.init(is_collective=True)
@@ -124,7 +126,9 @@ class TestFleetBaseSingleRunPS(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)
 
         fleet.init()

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

@@ -53,7 +53,9 @@ class TestFuseBatchNormActPass(unittest.TestCase):
                 input=hidden3, act='relu', data_layout='NHWC'
             )
             prediction = fluid.layers.fc(input=hidden4, size=10, act='softmax')
-            loss = fluid.layers.cross_entropy(input=prediction, label=y)
+            loss = paddle.nn.functional.cross_entropy(
+                input=prediction, label=y, reduction='none', use_softmax=False
+            )
             loss = paddle.mean(loss)
             sgd = fluid.optimizer.SGD(learning_rate=0.001)
             if use_cuda:

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

@@ -106,7 +106,9 @@ class TestFusedBnAddActAPI(unittest.TestCase):
                 act='softmax',
                 param_attr=self.fc_param_attr,
             )
-            loss = fluid.layers.cross_entropy(input=prediction, label=y)
+            loss = paddle.nn.functional.cross_entropy(
+                input=prediction, label=y, reduction='none', use_softmax=False
+            )
             loss = paddle.mean(loss)
             sgd = fluid.optimizer.SGD(learning_rate=0.001)
             sgd = fluid.contrib.mixed_precision.decorate(
@@ -162,7 +164,9 @@ class TestFusedBnAddActAPI(unittest.TestCase):
             prediction = fluid.layers.fc(
                 input=out, size=10, act='softmax', param_attr=self.fc_param_attr
             )
-            loss = fluid.layers.cross_entropy(input=prediction, label=y)
+            loss = paddle.nn.functional.cross_entropy(
+                input=prediction, label=y, reduction='none', use_softmax=False
+            )
             loss = paddle.mean(loss)
             sgd = fluid.optimizer.SGD(learning_rate=0.001)
             sgd = fluid.contrib.mixed_precision.decorate(

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

@@ -61,7 +61,9 @@ def simple_depthwise_net(use_feed):
         hidden = sep_conv(hidden, channel=200, stride=2, filter=5)
         hidden = fluid.layers.relu(hidden)
     prediction = fluid.layers.fc(hidden, size=10, act='softmax')
-    loss = fluid.layers.cross_entropy(input=prediction, label=label)
+    loss = paddle.nn.functional.cross_entropy(
+        input=prediction, label=label, reduction='none', use_softmax=False
+    )
     loss = paddle.mean(loss)
     return loss
 

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

@@ -68,7 +68,12 @@ def simple_fc_net(places, use_legacy_py_reader, use_double_buffer):
                 hidden, size=CLASS_NUM, act='softmax'
             )
             loss = paddle.mean(
-                fluid.layers.cross_entropy(input=predict_label, label=label)
+                paddle.nn.functional.cross_entropy(
+                    input=predict_label,
+                    label=label,
+                    reduction='none',
+                    use_softmax=False,
+                )
             )
 
             optimizer = fluid.optimizer.Adam()

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

@@ -41,7 +41,9 @@ def bow_net(
     fc_1 = fluid.layers.fc(input=bow_tanh, size=hid_dim, act="tanh")
     fc_2 = fluid.layers.fc(input=fc_1, size=hid_dim2, act="tanh")
     prediction = fluid.layers.fc(input=[fc_2], size=class_dim, 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
+    )
     avg_cost = paddle.mean(x=cost)
 
     return avg_cost
@@ -83,7 +85,9 @@ class TestGradientClip(unittest.TestCase):
                 hidden = fluid.layers.fc(input=image, size=32, act='relu')
             predict = fluid.layers.fc(input=hidden, size=10, act='softmax')
 
-            cost = fluid.layers.cross_entropy(input=predict, label=label)
+            cost = paddle.nn.functional.cross_entropy(
+                input=predict, label=label, reduction='none', use_softmax=False
+            )
             avg_cost = paddle.mean(cost)
 
         prog_clip = prog.clone()

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

@@ -90,8 +90,10 @@ class AutoPruneLayer2(fluid.Layer):
         label = self.linear2(label)
         label = fluid.layers.cast(label, dtype="float32")
         label = fluid.layers.cast(label, dtype='int64')
-        # Note that the label is not persistable in fluid.layers.cross_entropy.
-        loss = fluid.layers.cross_entropy(input=feature, label=label)
+        # Note that the label is not persistable in paddle.nn.functional.cross_entropy.
+        loss = paddle.nn.functional.cross_entropy(
+            input=feature, label=label, reduction='none', use_softmax=False
+        )
         loss = paddle.mean(loss)
         return loss
 
@@ -107,7 +109,9 @@ class AutoPruneLayer3(fluid.Layer):
             feature, num_or_sections=[10, 10], dim=1
         )
         # Note that: part2 is not used.
-        loss = fluid.layers.cross_entropy(input=part1, label=label)
+        loss = paddle.nn.functional.cross_entropy(
+            input=part1, label=label, reduction='none', use_softmax=False
+        )
         loss = paddle.mean(loss)
         if test_num == 1:
             return loss, part2

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

@@ -159,7 +159,9 @@ class TestImperativeMnist(unittest.TestCase):
                         cost_static = traced_layer([img])
                         helper.assertEachVar(cost, cost_static)
 
-                    loss = fluid.layers.cross_entropy(cost, label)
+                    loss = paddle.nn.functional.cross_entropy(
+                        cost, label, reduction='none', use_softmax=False
+                    )
                     avg_loss = paddle.mean(loss)
 
                     dy_out = avg_loss.numpy()
@@ -199,7 +201,9 @@ class TestImperativeMnist(unittest.TestCase):
             )
             label = fluid.layers.data(name='label', shape=[1], dtype='int64')
             cost = mnist(img)
-            loss = fluid.layers.cross_entropy(cost, label)
+            loss = paddle.nn.functional.cross_entropy(
+                cost, label, reduction='none', use_softmax=False
+            )
             avg_loss = paddle.mean(loss)
             sgd.minimize(avg_loss)
 

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

@@ -62,7 +62,9 @@ class TestImperativeMnistSortGradient(unittest.TestCase):
                     label2.stop_gradient = True
 
                     cost2 = mnist2(img2)
-                    loss2 = fluid.layers.cross_entropy(cost2, label2)
+                    loss2 = paddle.nn.functional.cross_entropy(
+                        cost2, label2, reduction='none', use_softmax=False
+                    )
                     avg_loss2 = paddle.mean(loss2)
 
                     dy_out2 = avg_loss2.numpy()
@@ -102,7 +104,9 @@ class TestImperativeMnistSortGradient(unittest.TestCase):
             )
             label = fluid.layers.data(name='label', shape=[1], dtype='int64')
             cost = mnist(img)
-            loss = fluid.layers.cross_entropy(cost, label)
+            loss = paddle.nn.functional.cross_entropy(
+                cost, label, reduction='none', use_softmax=False
+            )
             avg_loss = paddle.mean(loss)
             sgd.minimize(avg_loss)
 

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

@@ -477,8 +477,11 @@ class TestDygraphOCRAttention(unittest.TestCase):
                     dy_prediction = paddle.reshape(
                         dy_prediction, [label_out.shape[0], -1]
                     )
-                    loss = fluid.layers.cross_entropy(
-                        input=dy_prediction, label=label_out
+                    loss = paddle.nn.functional.cross_entropy(
+                        input=dy_prediction,
+                        label=label_out,
+                        reduction='none',
+                        use_softmax=False,
                     )
                     avg_loss = paddle.sum(loss)
 
@@ -555,8 +558,11 @@ class TestDygraphOCRAttention(unittest.TestCase):
                 static_prediction, shape=[-1, Config.num_classes + 2]
             )
 
-            cost = fluid.layers.cross_entropy(
-                input=static_prediction, label=static_label_out
+            cost = paddle.nn.functional.cross_entropy(
+                input=static_prediction,
+                label=static_label_out,
+                reduction='none',
+                use_softmax=False,
             )
             static_avg_loss = paddle.sum(cost)
             # param_grad_list = fluid.backward.append_backward(static_avg_loss)

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

@@ -311,7 +311,9 @@ class TestDygraphResnet(unittest.TestCase):
                     helper.assertEachVar(out_dygraph, out_static)
                     resnet.train()
 
-                loss = fluid.layers.cross_entropy(input=out, label=label)
+                loss = paddle.nn.functional.cross_entropy(
+                    input=out, label=label, reduction='none', use_softmax=False
+                )
                 avg_loss = paddle.mean(x=loss)
 
                 dy_out = avg_loss.numpy()
@@ -364,7 +366,9 @@ class TestDygraphResnet(unittest.TestCase):
             )
             label = fluid.layers.data(name='label', shape=[1], dtype='int64')
             out = resnet(img)
-            loss = fluid.layers.cross_entropy(input=out, label=label)
+            loss = paddle.nn.functional.cross_entropy(
+                input=out, label=label, reduction='none', use_softmax=False
+            )
             avg_loss = paddle.mean(x=loss)
             optimizer.minimize(avg_loss)
 

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

@@ -118,7 +118,9 @@ class TestDygraphResnetSortGradient(unittest.TestCase):
                 label.stop_gradient = True
 
                 out = resnet(img)
-                loss = fluid.layers.cross_entropy(input=out, label=label)
+                loss = paddle.nn.functional.cross_entropy(
+                    input=out, label=label, reduction='none', use_softmax=False
+                )
                 avg_loss = paddle.mean(x=loss)
 
                 dy_out = avg_loss.numpy()
@@ -174,7 +176,9 @@ class TestDygraphResnetSortGradient(unittest.TestCase):
             )
             label = fluid.layers.data(name='label', shape=[1], dtype='int64')
             out = resnet(img)
-            loss = fluid.layers.cross_entropy(input=out, label=label)
+            loss = paddle.nn.functional.cross_entropy(
+                input=out, label=label, reduction='none', use_softmax=False
+            )
             avg_loss = paddle.mean(x=loss)
             optimizer.minimize(avg_loss)
 

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

@@ -373,8 +373,11 @@ class TestImperativeResneXt(unittest.TestCase):
 
                     out = se_resnext(img)
                     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)
 
@@ -453,7 +456,12 @@ class TestImperativeResneXt(unittest.TestCase):
             label = fluid.layers.data(name='label', shape=[1], dtype='int64')
             out = se_resnext(img)
             softmax_out = paddle.nn.function.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)
             optimizer.minimize(avg_loss)
 

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

@@ -49,7 +49,9 @@ def convolutional_neural_network(img):
 def static_train_net(img, label):
     prediction = convolutional_neural_network(img)
 
-    loss = fluid.layers.cross_entropy(input=prediction, label=label)
+    loss = paddle.nn.functional.cross_entropy(
+        input=prediction, label=label, reduction='none', use_softmax=False
+    )
     avg_loss = paddle.mean(loss)
 
     optimizer = fluid.optimizer.SGD(learning_rate=0.001)
@@ -172,7 +174,9 @@ class TestImperativeStaticModelRunnerMnist(unittest.TestCase):
 
                     cost = mnist(img)
 
-                    loss = fluid.layers.cross_entropy(cost, label)
+                    loss = paddle.nn.functional.cross_entropy(
+                        cost, label, reduction='none', use_softmax=False
+                    )
                     avg_loss = paddle.mean(loss)
 
                     avg_loss.backward()

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

@@ -78,7 +78,9 @@ class TestImperativeStaticModelRunnerWhile(unittest.TestCase):
 
         pred = while_softmax_regression(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(loss)
 
         optimizer = fluid.optimizer.SGD(learning_rate=0.001)
@@ -149,7 +151,9 @@ class TestImperativeStaticModelRunnerWhile(unittest.TestCase):
 
                 cost = while_net(img)
 
-                loss = fluid.layers.cross_entropy(cost, label)
+                loss = paddle.nn.functional.cross_entropy(
+                    cost, label, reduction='none', use_softmax=False
+                )
                 avg_loss = paddle.mean(loss)
 
                 avg_loss.backward()
@@ -174,7 +178,9 @@ class TestImperativeStaticModelRunnerWhile(unittest.TestCase):
 
             pred = while_softmax_regression(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(loss)
 
             optimizer = fluid.optimizer.SGD(learning_rate=0.001)

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

@@ -182,7 +182,9 @@ class TestSaveInferenceModel(unittest.TestCase):
             auc_var, batch_auc_var, auc_states = paddle.static.auc(
                 input=predict, label=y
             )
-            cost = fluid.layers.cross_entropy(input=predict, label=y)
+            cost = paddle.nn.functional.cross_entropy(
+                input=predict, label=y, reduction='none', use_softmax=False
+            )
             avg_cost = paddle.mean(x=cost)
 
         place = core.CPUPlace()

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

@@ -40,7 +40,9 @@ def fc_with_batchnorm(use_feed):
 
         hidden = paddle.static.nn.batch_norm(input=hidden)
     prediction = fluid.layers.fc(hidden, size=10, act='softmax')
-    loss = fluid.layers.cross_entropy(input=prediction, label=label)
+    loss = paddle.nn.functional.cross_entropy(
+        input=prediction, label=label, reduction='none', use_softmax=False
+    )
     loss = paddle.mean(loss)
     return loss
 

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

@@ -46,7 +46,9 @@ def lstm_net(
     lstm_max_tanh = paddle.tanh(lstm_max)
     fc1 = fluid.layers.fc(input=lstm_max_tanh, size=hid_dim2, act='tanh')
     prediction = fluid.layers.fc(input=fc1, size=class_dim, 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
+    )
     avg_cost = paddle.mean(x=cost)
     return avg_cost
 

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

@@ -35,7 +35,9 @@ def simple_fc_net(use_feed):
     for _ in range(hidden_layer):
         x = fluid.layers.fc(input=x, size=20, act='relu')
     y_predict = fluid.layers.fc(input=x, size=10, act='softmax')
-    cost = fluid.layers.cross_entropy(input=y_predict, label=y)
+    cost = paddle.nn.functional.cross_entropy(
+        input=y_predict, label=y, reduction='none', use_softmax=False
+    )
     avg_cost = paddle.mean(cost)
     return avg_cost
 
@@ -48,7 +50,9 @@ def fc_with_inplace_net(use_feed):
     reshape = paddle.reshape(x=fc, shape=[-1, 2, 5])
     reshape = paddle.reshape(x=reshape, shape=[-1, 5, 2])
     y_predict = fluid.layers.fc(input=reshape, size=10, act='softmax')
-    cost = fluid.layers.cross_entropy(input=y_predict, label=y)
+    cost = paddle.nn.functional.cross_entropy(
+        input=y_predict, label=y, reduction='none', use_softmax=False
+    )
     avg_cost = paddle.mean(cost)
     return avg_cost
 

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

@@ -94,7 +94,9 @@ class LinerNetWithLabel(paddle.nn.Layer):
     )
     def forward(self, x, label):
         out = self._linear(x)
-        loss = fluid.layers.cross_entropy(out, label)
+        loss = paddle.nn.functional.cross_entropy(
+            out, label, reduction='none', use_softmax=False
+        )
         avg_loss = paddle.mean(loss)
         return out, avg_loss
 
@@ -112,7 +114,9 @@ class LinerNetWithPruneInput(paddle.nn.Layer):
     )
     def forward(self, x, label):
         out = self._linear(x)
-        loss = fluid.layers.cross_entropy(out, label)
+        loss = paddle.nn.functional.cross_entropy(
+            out, label, reduction='none', use_softmax=False
+        )
         avg_loss = paddle.mean(loss)
         return out
 
@@ -312,7 +316,9 @@ def train(layer, input_size=784, label_size=1):
 
         cost = layer(img)
 
-        loss = fluid.layers.cross_entropy(cost, label)
+        loss = paddle.nn.functional.cross_entropy(
+            cost, label, reduction='none', use_softmax=False
+        )
         avg_loss = paddle.mean(loss)
 
         avg_loss.backward()

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

@@ -49,7 +49,9 @@ def convolutional_neural_network(img):
 def static_train_net(img, label):
     prediction = convolutional_neural_network(img)
 
-    loss = fluid.layers.cross_entropy(input=prediction, label=label)
+    loss = paddle.nn.functional.cross_entropy(
+        input=prediction, label=label, reduction='none', use_softmax=False
+    )
     avg_loss = paddle.mean(loss)
 
     optimizer = fluid.optimizer.SGD(learning_rate=0.001)

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

@@ -28,7 +28,9 @@ img_shape = [1, 28, 28]
 
 def loss_net(hidden, label):
     prediction = fluid.layers.fc(input=hidden, size=10, act='softmax')
-    loss = fluid.layers.cross_entropy(input=prediction, label=label)
+    loss = paddle.nn.functional.cross_entropy(
+        input=prediction, label=label, reduction='none', use_softmax=False
+    )
     avg_loss = paddle.mean(loss)
     return avg_loss
 

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

@@ -99,7 +99,9 @@ class TestDygraphDataLoader(unittest.TestCase):
                 step = 0
                 for image, label in dataloader():
                     out = fc_net(image)
-                    loss = fluid.layers.cross_entropy(out, label)
+                    loss = paddle.nn.functional.cross_entropy(
+                        out, label, reduction='none', use_softmax=False
+                    )
                     avg_loss = paddle.mean(loss)
                     avg_loss.backward()
                     optimizer.minimize(avg_loss)
@@ -169,7 +171,9 @@ class TestDygraphDataLoaderWithBatchedDataset(TestDygraphDataLoader):
                 step = 0
                 for image, label in dataloader():
                     out = fc_net(image)
-                    loss = fluid.layers.cross_entropy(out, label)
+                    loss = paddle.nn.functional.cross_entropy(
+                        out, label, reduction='none', use_softmax=False
+                    )
                     avg_loss = paddle.mean(loss)
                     avg_loss.backward()
                     optimizer.minimize(avg_loss)

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

@@ -99,7 +99,9 @@ class TestDygraphDataLoader(unittest.TestCase):
                 step = 0
                 for image, label in dataloader():
                     out = fc_net(image)
-                    loss = fluid.layers.cross_entropy(out, label)
+                    loss = paddle.nn.functional.cross_entropy(
+                        out, label, reduction='none', use_softmax=False
+                    )
                     avg_loss = paddle.mean(loss)
                     avg_loss.backward()
                     optimizer.minimize(avg_loss)
@@ -167,7 +169,9 @@ class TestDygraphDataLoaderWithBatchedDataset(TestDygraphDataLoader):
                 step = 0
                 for image, label in dataloader():
                     out = fc_net(image)
-                    loss = fluid.layers.cross_entropy(out, label)
+                    loss = paddle.nn.functional.cross_entropy(
+                        out, label, reduction='none', use_softmax=False
+                    )
                     avg_loss = paddle.mean(loss)
                     avg_loss.backward()
                     optimizer.minimize(avg_loss)

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

@@ -80,7 +80,12 @@ def simple_fc_net_static():
                 bias_attr=bias_attr,
             )
             loss = paddle.mean(
-                fluid.layers.cross_entropy(input=predict_label, label=label)
+                paddle.nn.functional.cross_entropy(
+                    input=predict_label,
+                    label=label,
+                    reduction='none',
+                    use_softmax=False,
+                )
             )
 
             optimizer = fluid.optimizer.Adam()

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

@@ -80,7 +80,12 @@ def simple_fc_net_static():
                 bias_attr=bias_attr,
             )
             loss = paddle.mean(
-                fluid.layers.cross_entropy(input=predict_label, label=label)
+                paddle.nn.functional.cross_entropy(
+                    input=predict_label,
+                    label=label,
+                    reduction='none',
+                    use_softmax=False,
+                )
             )
 
             optimizer = fluid.optimizer.Adam()