Fix layer & dygraph circular dependent (#22334)

* fix circular dependent

* try import layers.nn from dygraph
test=develop

python/paddle/fluid/dygraph/nn.py

@@ -17,7 +17,8 @@ from __future__ import print_function
 from six.moves import reduce
 from .. import core
 from ..layers import utils
-from ..dygraph import dygraph_utils
+from ..layers import nn
+from .. import dygraph_utils
 from . import layers
 from ..framework import Variable, in_dygraph_mode, OpProtoHolder, Parameter, _dygraph_tracer, _varbase_creator
 from ..param_attr import ParamAttr

python/paddle/fluid/dygraph/dygraph_utils.py → python/paddle/fluid/dygraph_utils.py

@@ -12,8 +12,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from .. import core
-from ..framework import dygraph_only
+from . import core
+from .framework import dygraph_only
 
 
 @dygraph_only

python/paddle/fluid/layers/learning_rate_scheduler.py

@@ -32,7 +32,7 @@ from . import tensor
 from ..initializer import init_on_cpu
 from ..framework import default_main_program, Parameter, unique_name, name_scope
 from ..framework import Variable
-from ..dygraph import base as imperative_base
+from ..framework import in_dygraph_mode
 from ..dygraph import learning_rate_scheduler as imperate_lr
 
 __all__ = [
@@ -88,7 +88,7 @@ def noam_decay(d_model, warmup_steps):
                          warmup_steps)
     """
     with default_main_program()._lr_schedule_guard():
-        if imperative_base.enabled():
+        if in_dygraph_mode():
             decay = imperate_lr.NoamDecay(d_model, warmup_steps)
             return decay
         else:
@@ -143,7 +143,7 @@ def exponential_decay(learning_rate, decay_steps, decay_rate, staircase=False):
 
     """
     with default_main_program()._lr_schedule_guard():
-        if imperative_base.enabled():
+        if in_dygraph_mode():
             decay = imperate_lr.ExponentialDecay(learning_rate, decay_steps,
                                                  decay_rate, staircase)
             return decay
@@ -199,7 +199,7 @@ def natural_exp_decay(learning_rate, decay_steps, decay_rate, staircase=False):
 
     """
     with default_main_program()._lr_schedule_guard():
-        if imperative_base.enabled():
+        if in_dygraph_mode():
             decay = imperate_lr.NaturalExpDecay(learning_rate, decay_steps,
                                                 decay_rate, staircase)
             return decay
@@ -255,7 +255,7 @@ def inverse_time_decay(learning_rate, decay_steps, decay_rate, staircase=False):
 		    staircase=True))
     """
     with default_main_program()._lr_schedule_guard():
-        if imperative_base.enabled():
+        if in_dygraph_mode():
             decay = imperate_lr.InverseTimeDecay(learning_rate, decay_steps,
                                                  decay_rate, staircase)
             return decay
@@ -311,7 +311,7 @@ def polynomial_decay(learning_rate,
 
     """
     with default_main_program()._lr_schedule_guard():
-        if imperative_base.enabled():
+        if in_dygraph_mode():
             decay = imperate_lr.PolynomialDecay(learning_rate, decay_steps,
                                                 end_learning_rate, power, cycle)
             return decay
@@ -380,7 +380,7 @@ def piecewise_decay(boundaries, values):
         if len(values) - len(boundaries) != 1:
             raise ValueError("len(values) - len(boundaries) should be 1")
 
-        if imperative_base.enabled():
+        if in_dygraph_mode():
             decay = imperate_lr.PiecewiseDecay(boundaries, values, 0)
             return decay
         else:
@@ -444,7 +444,7 @@ def cosine_decay(learning_rate, step_each_epoch, epochs):
     """
 
     with default_main_program()._lr_schedule_guard():
-        if imperative_base.enabled():
+        if in_dygraph_mode():
             decay = imperate_lr.CosineDecay(learning_rate, step_each_epoch,
                                             epochs)
             return decay
@@ -520,7 +520,7 @@ def linear_lr_warmup(learning_rate, warmup_steps, start_lr, end_lr):
     linear_step = float(end_lr) - float(start_lr)
     with default_main_program()._lr_schedule_guard():
 
-        if imperative_base.enabled():
+        if in_dygraph_mode():
             lr = imperate_lr.LinearLrWarmup(learning_rate, warmup_steps,
                                             start_lr, end_lr)
             return lr

python/paddle/fluid/layers/nn.py

@@ -25,8 +25,7 @@ import inspect
 from ..layer_helper import LayerHelper
 from ..initializer import Normal, Constant, NumpyArrayInitializer
 from ..framework import Variable, OpProtoHolder, in_dygraph_mode, dygraph_only, _dygraph_tracer, default_main_program
-from ..dygraph import base
-from ..dygraph import dygraph_utils
+from .. import dygraph_utils
 from ..param_attr import ParamAttr
 from .layer_function_generator import autodoc, templatedoc, _generate_doc_string_
 from .tensor import concat, assign, fill_constant, zeros
@@ -34,7 +33,6 @@ from . import utils
 from .. import unique_name
 from functools import reduce
 from .. import core
-from ..dygraph import layers
 from ..data_feeder import convert_dtype, check_type_and_dtype, check_type, check_dtype
 
 __all__ = [
@@ -10276,9 +10274,6 @@ def _elementwise_op(helper):
     op_type = helper.layer_type
     x = helper.kwargs.get('x', None)
     y = helper.kwargs.get('y', None)
-    if in_dygraph_mode():
-        x = base.to_variable(x)
-        y = base.to_variable(y)
 
     assert x is not None, 'x cannot be None in {}'.format(op_type)
     assert y is not None, 'y cannot be None in {}'.format(op_type)