reaplce fill_constant_batch_size_like

python/paddle/distributed/auto_parallel/static/cost/comp_op_cost.py

@@ -471,6 +471,14 @@ def __init__(self, op=None, op_desc=None, cluster=None):
         super().__init__(op=op, op_desc=op_desc, cluster=cluster)
 
 
+@register_op_cost
+class ShapeOpCost(CompOpCost):
+    OP_TYPE = "shape"
+
+    def __init__(self, op=None, op_desc=None, cluster=None):
+        super().__init__(op=op, op_desc=op_desc, cluster=cluster)
+
+
 @register_op_cost
 class SliceOpCost(CompOpCost):
     OP_TYPE = "slice"

python/paddle/distribution/bernoulli.py

@@ -18,7 +18,7 @@
 import paddle
 from paddle.distribution import exponential_family
 from paddle.fluid.data_feeder import check_type, convert_dtype
-from paddle.fluid.layers import tensor
+from paddle.fluid.framework import Variable
 from paddle.framework import in_dynamic_mode
 from paddle.nn.functional import (
     binary_cross_entropy_with_logits,
@@ -97,7 +97,7 @@ def __init__(self, probs, name=None):
             check_type(
                 probs,
                 'probs',
-                (float, tensor.Variable),
+                (float, Variable),
                 self.name,
             )
 
@@ -180,7 +180,7 @@ def sample(self, shape):
             check_type(
                 shape,
                 'shape',
-                (np.ndarray, tensor.Variable, list, tuple),
+                (np.ndarray, Variable, list, tuple),
                 name,
             )
 
@@ -259,7 +259,7 @@ def rsample(self, shape, temperature=1.0):
             check_type(
                 shape,
                 'shape',
-                (np.ndarray, tensor.Variable, list, tuple),
+                (np.ndarray, Variable, list, tuple),
                 name,
             )
             check_type(
@@ -318,7 +318,7 @@ def cdf(self, value):
         """
         name = self.name + '_cdf'
         if not in_dynamic_mode():
-            check_type(value, 'value', tensor.Variable, name)
+            check_type(value, 'value', Variable, name)
 
         value = self._check_values_dtype_in_probs(self.probs, value)
         probs, value = paddle.broadcast_tensors([self.probs, value])
@@ -356,7 +356,7 @@ def log_prob(self, value):
         """
         name = self.name + '_log_prob'
         if not in_dynamic_mode():
-            check_type(value, 'value', tensor.Variable, name)
+            check_type(value, 'value', Variable, name)
 
         value = self._check_values_dtype_in_probs(self.probs, value)
         logits, value = paddle.broadcast_tensors([self.logits, value])
@@ -395,7 +395,7 @@ def prob(self, value):
         """
         name = self.name + '_prob'
         if not in_dynamic_mode():
-            check_type(value, 'value', tensor.Variable, name)
+            check_type(value, 'value', Variable, name)
 
         return self.log_prob(value).exp(name=name)
 

python/paddle/distribution/categorical.py

@@ -17,7 +17,7 @@
 import paddle
 from paddle.distribution import distribution
 from paddle.fluid.data_feeder import check_type, convert_dtype
-from paddle.fluid.layers import tensor
+from paddle.fluid.framework import Variable
 from paddle.framework import in_dynamic_mode
 from paddle.tensor import multinomial
 
@@ -100,7 +100,7 @@ def __init__(self, logits, name=None):
             check_type(
                 logits,
                 'logits',
-                (np.ndarray, tensor.Variable, list, tuple),
+                (np.ndarray, Variable, list, tuple),
                 'Categorical',
             )
 

python/paddle/distribution/distribution.py

@@ -26,7 +26,7 @@
 import paddle
 from paddle import _C_ops
 from paddle.fluid.data_feeder import check_variable_and_dtype, convert_dtype
-from paddle.fluid.layers import tensor
+from paddle.fluid.framework import Variable
 from paddle.framework import in_dynamic_mode
 
 
@@ -150,7 +150,7 @@ def _validate_args(self, *args):
         is_variable = False
         is_number = False
         for arg in args:
-            if isinstance(arg, tensor.Variable):
+            if isinstance(arg, Variable):
                 is_variable = True
             else:
                 is_number = True
@@ -176,9 +176,7 @@ def _to_tensor(self, *args):
         tmp = 0.0
 
         for arg in args:
-            if not isinstance(
-                arg, (float, list, tuple, np.ndarray, tensor.Variable)
-            ):
+            if not isinstance(arg, (float, list, tuple, np.ndarray, Variable)):
                 raise TypeError(
                     "Type of input args must be float, list, tuple, numpy.ndarray or Tensor, but received type {}".format(
                         type(arg)

python/paddle/distribution/normal.py

@@ -20,7 +20,7 @@
 import paddle
 from paddle.distribution import distribution
 from paddle.fluid.data_feeder import check_type, convert_dtype
-from paddle.fluid.layers import tensor
+from paddle.fluid.framework import Variable
 from paddle.framework import in_dynamic_mode
 from paddle.tensor import random
 
@@ -91,13 +91,13 @@ def __init__(self, loc, scale, name=None):
             check_type(
                 loc,
                 'loc',
-                (int, float, np.ndarray, tensor.Variable, list, tuple),
+                (int, float, np.ndarray, Variable, list, tuple),
                 'Normal',
             )
             check_type(
                 scale,
                 'scale',
-                (int, float, np.ndarray, tensor.Variable, list, tuple),
+                (int, float, np.ndarray, Variable, list, tuple),
                 'Normal',
             )
 
@@ -174,9 +174,9 @@ def sample(self, shape=(), seed=0):
         name = self.name + '_sample'
         if -1 in batch_shape:
             output_shape = shape + batch_shape
-            zero_tmp = tensor.fill_constant_batch_size_like(
-                self.loc + self.scale, batch_shape + shape, self.dtype, 0.0
-            )
+            fill_shape = list(batch_shape + shape)
+            fill_shape[0] = paddle.shape(self.loc + self.scale)[0].item()
+            zero_tmp = paddle.full(fill_shape, 0.0, self.dtype)
             zero_tmp_reshape = paddle.reshape(zero_tmp, output_shape)
 
             zero_tmp_shape = paddle.shape(zero_tmp_reshape)
@@ -234,9 +234,10 @@ def entropy(self):
         name = self.name + '_entropy'
         batch_shape = list((self.loc + self.scale).shape)
         if -1 in batch_shape:
-            zero_tmp = tensor.fill_constant_batch_size_like(
-                self.loc + self.scale, batch_shape, self.dtype, 0.0
-            )
+            fill_shape = list(batch_shape)
+            fill_shape[0] = paddle.shape(self.loc + self.scale)[0].item()
+            fill_dtype = (self.loc + self.scale).dtype
+            zero_tmp = paddle.full(fill_shape, 0.0, fill_dtype)
         else:
             zero_tmp = paddle.full(batch_shape, 0.0, self.dtype)
         return paddle.add(

python/paddle/distribution/uniform.py

@@ -18,7 +18,7 @@
 from paddle import _C_ops
 from paddle.distribution import distribution
 from paddle.fluid.data_feeder import check_type, convert_dtype
-from paddle.fluid.layers import tensor
+from paddle.fluid.framework import Variable
 from paddle.framework import in_dynamic_mode
 from paddle.tensor import random
 
@@ -96,13 +96,13 @@ def __init__(self, low, high, name=None):
             check_type(
                 low,
                 'low',
-                (int, float, np.ndarray, tensor.Variable, list, tuple),
+                (int, float, np.ndarray, Variable, list, tuple),
                 'Uniform',
             )
             check_type(
                 high,
                 'high',
-                (int, float, np.ndarray, tensor.Variable, list, tuple),
+                (int, float, np.ndarray, Variable, list, tuple),
                 'Uniform',
             )
 
@@ -160,9 +160,9 @@ def sample(self, shape, seed=0):
         batch_shape = list((self.low + self.high).shape)
         if -1 in batch_shape:
             output_shape = shape + batch_shape
-            zero_tmp = tensor.fill_constant_batch_size_like(
-                self.low + self.high, batch_shape + shape, self.dtype, 0.0
-            )
+            fill_shape = list(batch_shape + shape)
+            fill_shape[0] = paddle.shape(self.low + self.high)[0].item()
+            zero_tmp = paddle.full(fill_shape, 0.0, self.dtype)
             uniform_random_tmp = random.uniform_random_batch_size_like(
                 zero_tmp,
                 zero_tmp.shape,

python/paddle/fluid/layers/__init__.py

@@ -16,8 +16,6 @@
 from .nn import *
 from . import io
 from .io import *
-from . import tensor
-from .tensor import *
 from . import math_op_patch
 from .math_op_patch import *
 from .learning_rate_scheduler import *
@@ -27,5 +25,4 @@
 __all__ = []
 __all__ += nn.__all__
 __all__ += io.__all__
-__all__ += tensor.__all__
 __all__ += learning_rate_scheduler.__all__

python/paddle/fluid/layers/learning_rate_scheduler.py

@@ -25,7 +25,6 @@
 
 import paddle
 from . import nn
-from . import tensor
 from ..framework import (
     default_main_program,
     Parameter,
@@ -488,7 +487,7 @@ def cosine_decay(learning_rate, step_each_epoch, epochs):
             learning_rate = base_lr, step_each_epoch=10000, epochs=120)
     """
     check_type(
-        learning_rate, 'learning_rate', (float, tensor.Variable), 'cosine_decay'
+        learning_rate, 'learning_rate', (float, Variable), 'cosine_decay'
     )
 
     with default_main_program()._lr_schedule_guard():

python/paddle/fluid/optimizer.py

@@ -48,7 +48,6 @@
     _LearningRateEpochDecay,
 )
 from paddle.fluid import core
-from paddle.fluid.layers import tensor
 from functools import reduce
 from functools import cmp_to_key
 from .wrapped_decorator import signature_safe_contextmanager