Skip to content

T
tensorflow

qq_38905368 / tensorflow
与 Fork 源项目一致

从无法访问的项目Fork

通知 5
Star 0
Fork 0
T
tensorflow

体验新版 GitCode，发现更多精彩内容 >>

提交 cff53408 编写于 作者: N Nathan Silberman 提交者： TensorFlower Gardener
浏览文件
操作

Fixing bug in which the 'trainable' argument wasn't being passed to the bias... 

Fixing bug in which the 'trainable' argument wasn't being passed to the bias in convolution2d_transpose.
Change: 139925626
上级 824c9f11
隐藏空白更改
内联 并排
Showing with 53 addition and 37 deletion
tensorflow/contrib/layers/python/layers/layers.py
浏览文件 @ cff53408
...@@ -220,14 +220,16 @@ def _fused_batch_norm( ...@@ -220,14 +220,16 @@ def _fused_batch_norm(
if original_rank is None: if original_rank is None:
raise ValueError('Inputs %s has undefined rank' % inputs.name) raise ValueError('Inputs %s has undefined rank' % inputs.name)
elif original_rank not in [2, 4]: elif original_rank not in [2, 4]:
raise ValueError('Inputs %s has unsupported rank. \ raise ValueError('Inputs %s has unsupported rank.'
Expected 2 or 4 but got %d' % (inputs.name, original_rank)) ' Expected 2 or 4 but got %d' % (
inputs.name, original_rank))
if original_rank == 2: if original_rank == 2:
channels = inputs.get_shape()[-1].value channels = inputs.get_shape()[-1].value
if channels is None: if channels is None:
raise ValueError('`C` dimension must be known but is None') raise ValueError('`C` dimension must be known but is None')
new_shape = [-1, channels, 1, 1] if data_format == DATA_FORMAT_NCHW else \ new_shape = [-1, 1, 1, channels]
[-1, 1, 1, channels] if data_format == DATA_FORMAT_NCHW:
new_shape = [-1, channels, 1, 1]
inputs = array_ops.reshape(inputs, new_shape) inputs = array_ops.reshape(inputs, new_shape)
inputs_shape = inputs.get_shape() inputs_shape = inputs.get_shape()
dtype = inputs.dtype.base_dtype dtype = inputs.dtype.base_dtype
...@@ -316,7 +318,7 @@ def _fused_batch_norm( ...@@ -316,7 +318,7 @@ def _fused_batch_norm(
need_updates = is_training_value is None or is_training_value need_updates = is_training_value is None or is_training_value
if need_updates: if need_updates:
if updates_collections is None: if updates_collections is None:
_no_updates = lambda: outputs no_updates = lambda: outputs
def _force_updates(): def _force_updates():
"""Internal function forces updates moving_vars if is_training.""" """Internal function forces updates moving_vars if is_training."""
update_moving_mean = moving_averages.assign_moving_average( update_moving_mean = moving_averages.assign_moving_average(
...@@ -326,7 +328,7 @@ def _fused_batch_norm( ...@@ -326,7 +328,7 @@ def _fused_batch_norm(
with ops.control_dependencies( with ops.control_dependencies(
[update_moving_mean, update_moving_variance]): [update_moving_mean, update_moving_variance]):
return array_ops.identity(outputs) return array_ops.identity(outputs)
outputs = utils.smart_cond(is_training, _force_updates, _no_updates) outputs = utils.smart_cond(is_training, _force_updates, no_updates)
else: else:
moving_vars_fn = lambda: (moving_mean, moving_variance) moving_vars_fn = lambda: (moving_mean, moving_variance)
def _delay_updates(): def _delay_updates():
...@@ -684,7 +686,7 @@ def bias_add(inputs, ...@@ -684,7 +686,7 @@ def bias_add(inputs,
raise ValueError('Dims of shape must be known but is None') raise ValueError('Dims of shape must be known but is None')
elif inputs_rank != 4 and data_format == DATA_FORMAT_NCHW: elif inputs_rank != 4 and data_format == DATA_FORMAT_NCHW:
raise ValueError('Data format NCHW only supports 4D Tensor') raise ValueError('Data format NCHW only supports 4D Tensor')
axis = 1 if data_format==DATA_FORMAT_NCHW else -1 axis = 1 if data_format == DATA_FORMAT_NCHW else -1
num_features = inputs_shape[axis].value num_features = inputs_shape[axis].value
if num_features is None: if num_features is None:
raise ValueError('`C` dimension must be known but is None') raise ValueError('`C` dimension must be known but is None')
...@@ -1081,7 +1083,6 @@ def convolution2d_transpose( ...@@ -1081,7 +1083,6 @@ def convolution2d_transpose(
output_shape = [batch_size, num_outputs, out_height, out_width] output_shape = [batch_size, num_outputs, out_height, out_width]
strides = [1, 1, stride_h, stride_w] strides = [1, 1, stride_h, stride_w]
output_shape = array_ops.pack(output_shape) output_shape = array_ops.pack(output_shape)
outputs = nn.conv2d_transpose(inputs, weights, output_shape, outputs = nn.conv2d_transpose(inputs, weights, output_shape,
strides, strides,
...@@ -1091,8 +1092,10 @@ def convolution2d_transpose( ...@@ -1091,8 +1092,10 @@ def convolution2d_transpose(
# Infer the static output shape: # Infer the static output shape:
out_shape = inputs.get_shape().as_list() out_shape = inputs.get_shape().as_list()
out_shape[c_axis] = num_outputs out_shape[c_axis] = num_outputs
out_shape[h_axis] = get_deconv_dim(out_shape[h_axis], stride_h, kernel_h, padding) out_shape[h_axis] = get_deconv_dim(
out_shape[w_axis] = get_deconv_dim(out_shape[w_axis], stride_w, kernel_w, padding) out_shape[h_axis], stride_h, kernel_h, padding)
out_shape[w_axis] = get_deconv_dim(
out_shape[w_axis], stride_w, kernel_w, padding)
outputs.set_shape(out_shape) outputs.set_shape(out_shape)
if normalizer_fn is not None: if normalizer_fn is not None:
...@@ -1107,6 +1110,7 @@ def convolution2d_transpose( ...@@ -1107,6 +1110,7 @@ def convolution2d_transpose(
dtype=dtype, dtype=dtype,
initializer=biases_initializer, initializer=biases_initializer,
regularizer=biases_regularizer, regularizer=biases_regularizer,
trainable=trainable,
collections=biases_collections) collections=biases_collections)
outputs = nn.bias_add(outputs, biases, data_format=data_format) outputs = nn.bias_add(outputs, biases, data_format=data_format)
......
tensorflow/contrib/layers/python/layers/layers_test.py
浏览文件 @ cff53408
...@@ -23,9 +23,8 @@ import numpy as np ...@@ -23,9 +23,8 @@ import numpy as np
import tensorflow as tf import tensorflow as tf
# TODO(sguada) Expose tf.with_dependencies # TODO(sguada) Expose tf.with_dependencies
from tensorflow.python.ops import control_flow_ops
from tensorflow.contrib.layers.python.layers import layers as _layers from tensorflow.contrib.layers.python.layers import layers as _layers
from tensorflow.python.ops import state_ops from tensorflow.python.ops import control_flow_ops
class AvgPool2DTest(tf.test.TestCase): class AvgPool2DTest(tf.test.TestCase):
...@@ -181,7 +180,8 @@ class PoolTest(tf.test.TestCase): ...@@ -181,7 +180,8 @@ class PoolTest(tf.test.TestCase):
height, width = 5, 8 height, width = 5, 8
images = tf.random_uniform((5, 3, height, width), seed=1) images = tf.random_uniform((5, 3, height, width), seed=1)
output = tf.contrib.layers.pool( output = tf.contrib.layers.pool(
images, [2, 3], dilation_rate=[1, 2], pooling_type='AVG', data_format='NCHW') images, [2, 3], dilation_rate=[1, 2], pooling_type='AVG',
data_format='NCHW')
self.assertEqual(output.get_shape().as_list(), [5, 3, 4, 4]) self.assertEqual(output.get_shape().as_list(), [5, 3, 4, 4])
...@@ -370,7 +370,7 @@ class ConvolutionTest(tf.test.TestCase): ...@@ -370,7 +370,7 @@ class ConvolutionTest(tf.test.TestCase):
tf.contrib.framework.get_variables_by_name('weights')[0]) tf.contrib.framework.get_variables_by_name('weights')[0])
wd = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)[0] wd = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)[0]
self.assertEqual(wd.op.name, self.assertEqual(wd.op.name,
'Conv/weights/Regularizer/l2_regularizer') 'Conv/weights/Regularizer/l2_regularizer')
sess.run(tf.global_variables_initializer()) sess.run(tf.global_variables_initializer())
self.assertAlmostEqual(sess.run(wd), weight_decay * l2_loss.eval()) self.assertAlmostEqual(sess.run(wd), weight_decay * l2_loss.eval())
...@@ -588,6 +588,20 @@ class ConvolutionTest(tf.test.TestCase): ...@@ -588,6 +588,20 @@ class ConvolutionTest(tf.test.TestCase):
class Convolution2dTransposeTests(tf.test.TestCase): class Convolution2dTransposeTests(tf.test.TestCase):
def testTrainableFlagIsPassedOn(self):
for trainable in [True, False]:
with tf.Graph().as_default():
num_filters = 32
input_size = [5, 10, 12, 3]
images = tf.random_uniform(input_size, seed=1)
tf.contrib.layers.conv2d_transpose(
images, num_filters, [3, 3], stride=1, trainable=trainable)
model_variables = tf.contrib.framework.get_model_variables()
trainable_variables = tf.trainable_variables()
for model_variable in model_variables:
self.assertEqual(trainable, model_variable in trainable_variables)
def testInvalidDataFormat(self): def testInvalidDataFormat(self):
height, width = 7, 9 height, width = 7, 9
with self.test_session(): with self.test_session():
...@@ -597,7 +611,6 @@ class Convolution2dTransposeTests(tf.test.TestCase): ...@@ -597,7 +611,6 @@ class Convolution2dTransposeTests(tf.test.TestCase):
tf.contrib.layers.convolution2d_transpose( tf.contrib.layers.convolution2d_transpose(
images, 32, 3, data_format='CHWN') images, 32, 3, data_format='CHWN')
def testOutputSizeWithStrideOneSamePaddingNCHW(self): def testOutputSizeWithStrideOneSamePaddingNCHW(self):
# `NCHW` data fomat is only supported for `GPU` device. # `NCHW` data fomat is only supported for `GPU` device.
if tf.test.is_gpu_available(): if tf.test.is_gpu_available():
...@@ -615,7 +628,6 @@ class Convolution2dTransposeTests(tf.test.TestCase): ...@@ -615,7 +628,6 @@ class Convolution2dTransposeTests(tf.test.TestCase):
sess.run(tf.global_variables_initializer()) sess.run(tf.global_variables_initializer())
self.assertListEqual(list(output.eval().shape), expected_size) self.assertListEqual(list(output.eval().shape), expected_size)
def testOutputSizeWithStrideOneValidPaddingNCHW(self): def testOutputSizeWithStrideOneValidPaddingNCHW(self):
if tf.test.is_gpu_available(): if tf.test.is_gpu_available():
with self.test_session(use_gpu=True) as sess: with self.test_session(use_gpu=True) as sess:
...@@ -756,7 +768,6 @@ class Convolution2dTransposeTests(tf.test.TestCase): ...@@ -756,7 +768,6 @@ class Convolution2dTransposeTests(tf.test.TestCase):
self.assertEqual(output.op.name, 'Conv2d_transpose/Relu') self.assertEqual(output.op.name, 'Conv2d_transpose/Relu')
self.assertListEqual(list(output.eval().shape), expected_size) self.assertListEqual(list(output.eval().shape), expected_size)
def testOutputSizeWithStrideOneSamePadding(self): def testOutputSizeWithStrideOneSamePadding(self):
num_filters = 32 num_filters = 32
input_size = [5, 10, 12, 3] input_size = [5, 10, 12, 3]
...@@ -1284,7 +1295,7 @@ class FlattenTest(tf.test.TestCase): ...@@ -1284,7 +1295,7 @@ class FlattenTest(tf.test.TestCase):
images = tf.random_uniform((5, height, width, 3), seed=1, name='images') images = tf.random_uniform((5, height, width, 3), seed=1, name='images')
output = tf.contrib.layers.flatten(images) output = tf.contrib.layers.flatten(images)
self.assertEqual(output.get_shape().num_elements(), self.assertEqual(output.get_shape().num_elements(),
images.get_shape().num_elements()) images.get_shape().num_elements())
self.assertEqual(output.get_shape()[0], images.get_shape()[0]) self.assertEqual(output.get_shape()[0], images.get_shape()[0])
def testFlatten3D(self): def testFlatten3D(self):
...@@ -1293,7 +1304,7 @@ class FlattenTest(tf.test.TestCase): ...@@ -1293,7 +1304,7 @@ class FlattenTest(tf.test.TestCase):
images = tf.random_uniform((5, height, width), seed=1, name='images') images = tf.random_uniform((5, height, width), seed=1, name='images')
output = tf.contrib.layers.flatten(images) output = tf.contrib.layers.flatten(images)
self.assertEqual(output.get_shape().num_elements(), self.assertEqual(output.get_shape().num_elements(),
images.get_shape().num_elements()) images.get_shape().num_elements())
self.assertEqual(output.get_shape()[0], images.get_shape()[0]) self.assertEqual(output.get_shape()[0], images.get_shape()[0])
def testFlattenBatchSize(self): def testFlattenBatchSize(self):
...@@ -1303,10 +1314,10 @@ class FlattenTest(tf.test.TestCase): ...@@ -1303,10 +1314,10 @@ class FlattenTest(tf.test.TestCase):
inputs = tf.placeholder(tf.int32, (None, height, width, 3)) inputs = tf.placeholder(tf.int32, (None, height, width, 3))
output = tf.contrib.layers.flatten(inputs) output = tf.contrib.layers.flatten(inputs)
self.assertEqual(output.get_shape().as_list(), self.assertEqual(output.get_shape().as_list(),
[None, height * width * 3]) [None, height * width * 3])
output = sess.run(output, {inputs: images.eval()}) output = sess.run(output, {inputs: images.eval()})
self.assertEqual(output.size, self.assertEqual(output.size,
images.get_shape().num_elements()) images.get_shape().num_elements())
self.assertEqual(output.shape[0], images.get_shape()[0]) self.assertEqual(output.shape[0], images.get_shape()[0])
...@@ -1463,7 +1474,7 @@ class FCTest(tf.test.TestCase): ...@@ -1463,7 +1474,7 @@ class FCTest(tf.test.TestCase):
weights_regularizer=weight_decay) weights_regularizer=weight_decay)
wd = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)[0] wd = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)[0]
self.assertEqual(wd.op.name, self.assertEqual(wd.op.name,
'fully_connected/weights/Regularizer/l2_regularizer') 'fully_connected/weights/Regularizer/l2_regularizer')
sess.run(tf.global_variables_initializer()) sess.run(tf.global_variables_initializer())
self.assertLess(sess.run(wd), 0.4) self.assertLess(sess.run(wd), 0.4)
...@@ -1620,9 +1631,9 @@ class BatchNormTest(tf.test.TestCase): ...@@ -1620,9 +1631,9 @@ class BatchNormTest(tf.test.TestCase):
update_moving_mean = update_layers[0] update_moving_mean = update_layers[0]
update_moving_variance = update_layers[1] update_moving_variance = update_layers[1]
self.assertEqual(update_moving_mean.op.name, self.assertEqual(update_moving_mean.op.name,
'BatchNorm/AssignMovingAvg') 'BatchNorm/AssignMovingAvg')
self.assertEqual(update_moving_variance.op.name, self.assertEqual(update_moving_variance.op.name,
'BatchNorm/AssignMovingAvg_1') 'BatchNorm/AssignMovingAvg_1')
def testReuseVariables(self): def testReuseVariables(self):
height, width = 3, 3 height, width = 3, 3
...@@ -1774,8 +1785,8 @@ class BatchNormTest(tf.test.TestCase): ...@@ -1774,8 +1785,8 @@ class BatchNormTest(tf.test.TestCase):
if fused: if fused:
# Add Bessel's correction # Add Bessel's correction
moving_variance_corrected = moving_variance / correction_factor moving_variance_corrected = moving_variance / correction_factor
correct_moving_variance = state_ops.assign(moving_variance, correct_moving_variance = tf.assign(moving_variance,
moving_variance_corrected) moving_variance_corrected)
sess.run(correct_moving_variance) sess.run(correct_moving_variance)
self.assertAllClose(variance, expected_var) self.assertAllClose(variance, expected_var)
...@@ -1888,8 +1899,8 @@ class BatchNormTest(tf.test.TestCase): ...@@ -1888,8 +1899,8 @@ class BatchNormTest(tf.test.TestCase):
if fused: if fused:
# Add Bessel's correction # Add Bessel's correction
moving_variance_corrected = moving_variance / correction_factor moving_variance_corrected = moving_variance / correction_factor
correct_moving_variance = state_ops.assign(moving_variance, correct_moving_variance = tf.assign(moving_variance,
moving_variance_corrected) moving_variance_corrected)
sess.run(correct_moving_variance) sess.run(correct_moving_variance)
self.assertAllClose(variance, expected_var) self.assertAllClose(variance, expected_var)
# After convergence output_train and output_eval should be the same. # After convergence output_train and output_eval should be the same.
...@@ -1961,8 +1972,8 @@ class BatchNormTest(tf.test.TestCase): ...@@ -1961,8 +1972,8 @@ class BatchNormTest(tf.test.TestCase):
if fused: if fused:
# Add Bessel's correction # Add Bessel's correction
moving_variance_corrected = moving_variance / correction_factor moving_variance_corrected = moving_variance / correction_factor
correct_moving_variance = state_ops.assign(moving_variance, correct_moving_variance = tf.assign(moving_variance,
moving_variance_corrected) moving_variance_corrected)
sess.run(correct_moving_variance) sess.run(correct_moving_variance)
output_false = sess.run([output], {is_training: False}) output_false = sess.run([output], {is_training: False})
self.assertAllClose(output_true, output_false) self.assertAllClose(output_true, output_false)
...@@ -2100,8 +2111,8 @@ class BatchNormTest(tf.test.TestCase): ...@@ -2100,8 +2111,8 @@ class BatchNormTest(tf.test.TestCase):
if fused: if fused:
# Add Bessel's correction # Add Bessel's correction
moving_variance_corrected = moving_variance / correction_factor moving_variance_corrected = moving_variance / correction_factor
correct_moving_variance = state_ops.assign(moving_variance, correct_moving_variance = tf.assign(moving_variance,
moving_variance_corrected) moving_variance_corrected)
sess.run(correct_moving_variance) sess.run(correct_moving_variance)
output_false = sess.run([output], {is_training: False}) output_false = sess.run([output], {is_training: False})
self.assertTrue(np.allclose(output_true, output_false)) self.assertTrue(np.allclose(output_true, output_false))
...@@ -2212,10 +2223,10 @@ class BatchNormTest(tf.test.TestCase): ...@@ -2212,10 +2223,10 @@ class BatchNormTest(tf.test.TestCase):
scale=True, scale=True,
epsilon=0.0, epsilon=0.0,
param_initializers={ param_initializers={
'beta': beta, 'beta': beta,
'gamma': gamma, 'gamma': gamma,
'moving_mean': mean, 'moving_mean': mean,
'moving_variance': variance, 'moving_variance': variance,
}) })
sess.run(tf.global_variables_initializer()) sess.run(tf.global_variables_initializer())
outs = sess.run(output) outs = sess.run(output)
...@@ -2358,6 +2369,7 @@ class LayerNormTest(tf.test.TestCase): ...@@ -2358,6 +2369,7 @@ class LayerNormTest(tf.test.TestCase):
def testOutput4DInput(self): def testOutput4DInput(self):
self.doOutputTest((100, 10, 10, 3)) self.doOutputTest((100, 10, 10, 3))
class MaxPool2DTest(tf.test.TestCase): class MaxPool2DTest(tf.test.TestCase):
def testInvalidDataFormat(self): def testInvalidDataFormat(self):
...@@ -2974,7 +2986,7 @@ class LegacyFullyConnectedTest(tf.test.TestCase): ...@@ -2974,7 +2986,7 @@ class LegacyFullyConnectedTest(tf.test.TestCase):
self.assertEqual(1, len(tf.get_collection('unbiased'))) self.assertEqual(1, len(tf.get_collection('unbiased')))
self.assertEqual(1, len(tf.get_collection('biased'))) self.assertEqual(1, len(tf.get_collection('biased')))
self.assertEqual(tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES), self.assertEqual(tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES),
tf.get_collection('all')) tf.get_collection('all'))
def test_no_bias(self): def test_no_bias(self):
tf.contrib.layers.legacy_relu(self.input, 2, bias_init=None) tf.contrib.layers.legacy_relu(self.input, 2, bias_init=None)
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或