OSSForks/tensorflow
1
0
Fork 0
mirror of https://github.com/zebrajr/tensorflow.git synced 2025-12-07 00:20:20 +01:00
Code Issues Actions 5 Packages Projects Releases Wiki Activity

Internal cleanup.

Browse Source 
PiperOrigin-RevId: 170922297
This commit is contained in:
A. Unique TensorFlower 2017-10-03 15:14:22 -07:00 committed by TensorFlower Gardener
parent d0c76cd188
commit 5123f29718
4 changed files with 38 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
tensorflow/python/eager/backprop.py
View File

@ -324,12 +324,18 @@ def imperative_grad(
result.append(_aggregate_grads(g)) result.append(_aggregate_grads(g))
return result return result
_op_attr_type_cache = {}
def op_attr_type(op_type, attr_name): def op_attr_type(op_type, attr_name):
try:
return _op_attr_type_cache[(op_type, attr_name)]
except KeyError:
with errors.raise_exception_on_not_ok_status() as status: with errors.raise_exception_on_not_ok_status() as status:
h = context.context()._handle # pylint: disable=protected-access h = context.context()._handle # pylint: disable=protected-access
op = pywrap_tensorflow.TFE_NewOp(h, op_type, status) op = pywrap_tensorflow.TFE_NewOp(h, op_type, status)
attr_type = pywrap_tensorflow.TFE_OpGetAttrType(op, attr_name, status) attr_type = pywrap_tensorflow.TFE_OpGetAttrType(op, attr_name, status)
_op_attr_type_cache[(op_type, attr_name)] = attr_type
return attr_type return attr_type

4
tensorflow/python/layers/normalization.py
View File

@ -277,7 +277,9 @@ class BatchNormalization(base.Layer):
with ops.name_scope(None, 'AssignMovingAvg', with ops.name_scope(None, 'AssignMovingAvg',
[variable, value, one_minus_decay]) as scope: [variable, value, one_minus_decay]) as scope:
with ops.colocate_with(variable): with ops.colocate_with(variable):
update_delta = (variable.read_value() - value) * one_minus_decay update_delta = math_ops.multiply(
math_ops.subtract(variable.read_value(), value),
one_minus_decay)
if isinstance(variable, resource_variable_ops.ResourceVariable): if isinstance(variable, resource_variable_ops.ResourceVariable):
# state_ops.assign_sub does an extra read_variable_op after the # state_ops.assign_sub does an extra read_variable_op after the
# assign. We avoid that here. # assign. We avoid that here.

3
tensorflow/python/ops/array_grad.py
View File

@ -79,11 +79,12 @@ def _ConcatGradHelper(op, grad, start_value_index, end_value_index, dim_index):
def _ExtractInputShapes(inputs): def _ExtractInputShapes(inputs):
"""Extract the shapes of a set of input tensors.""" """Extract the shapes of a set of input tensors."""
if not context.in_graph_mode():
return array_ops.shape_n(inputs)
sizes = [] sizes = []
fully_known = True fully_known = True
for x in inputs: for x in inputs:
input_shape = array_ops.shape(x) input_shape = array_ops.shape(x)
if context.in_graph_mode():
if not isinstance(input_shape, if not isinstance(input_shape,
ops.Tensor) or input_shape.op.type != "Const": ops.Tensor) or input_shape.op.type != "Const":
fully_known = False fully_known = False

27
tensorflow/python/ops/nn_grad.py
View File

@ -460,16 +460,25 @@ def _SparseSoftmaxCrossEntropyWithLogitsGrad(op, grad_0, _):
@ops.RegisterGradient("Conv2D") @ops.RegisterGradient("Conv2D")
def _Conv2DGrad(op, grad): def _Conv2DGrad(op, grad):
return [nn_ops.conv2d_backprop_input( strides = op.get_attr("strides")
array_ops.shape(op.inputs[0]), op.inputs[1], grad, op.get_attr("strides"), padding = op.get_attr("padding")
op.get_attr("padding"), op.get_attr("use_cudnn_on_gpu"), use_cudnn_on_gpu = op.get_attr("use_cudnn_on_gpu")
op.get_attr("data_format")), data_format = op.get_attr("data_format")
shape_0, shape_1 = array_ops.shape_n([op.inputs[0], op.inputs[1]])
return [nn_ops.conv2d_backprop_input(shape_0,
op.inputs[1],
grad,
strides,
padding,
use_cudnn_on_gpu,
data_format),
nn_ops.conv2d_backprop_filter(op.inputs[0], nn_ops.conv2d_backprop_filter(op.inputs[0],
array_ops.shape(op.inputs[1]), grad, shape_1,
op.get_attr("strides"), grad,
op.get_attr("padding"), strides,
op.get_attr("use_cudnn_on_gpu"), padding,
op.get_attr("data_format"))] use_cudnn_on_gpu,
data_format)]
@ops.RegisterGradient("DepthwiseConv2dNative") @ops.RegisterGradient("DepthwiseConv2dNative")