Styjun

Why did the Death Eaters wait to reopen the Chamber of Secrets?

DBCC SHRINKFILE on the distribution database

What's the reason for the decade jump in the recent X-Men trilogy?

Approach sick days in feedback meeting

What publication claimed that Michael Jackson died in a nuclear holocaust?

Opposite of "Concerto Grosso"?

Must a CPU have a GPU if the motherboard provides a display port (when there isn't any separate video card)?

The best in flight meal option for those suffering from reflux

Is it a good security practice to force employees hide their employer to avoid being targeted?

How to search for Android apps without ads?

Optimising matrix generation time

Dedicated bike GPS computer over smartphone

How can this shape perfectly cover a cube?

Should I worry about having my credit pulled multiple times while car shopping?

Are athletes' college degrees discounted by employers and graduate school admissions?

Can a 40amp breaker be used safely and without issue with a 40amp device on 6AWG wire?

Was the Lonely Mountain, where Smaug lived, a volcano?

Arrows inside a commutative diagram using tikzcd

Why not make one big cpu core?

Is fission/fusion to iron the most efficient way to convert mass to energy?

Is it true that "only photographers care about noise"?

Is it possible to have battery technology that can't be duplicated?

Can I get a photo of an Ancient Arrow?

How effective would a full set of plate armor be against wild animals found in temperate regions (bears, snakes, wolves)?

How to freeze a Keras graph with BatchNorm layers

How to merge two dictionaries in a single expression?How do I check if a list is empty?How do I check whether a file exists without exceptions?How can I safely create a nested directory?How can I make a time delay in Python?How do I sort a dictionary by value?How to make a chain of function decorators?How to make a flat list out of list of listsHow do I list all files of a directory?Can't import frozen graph after adding layers to Keras model

.everyoneloves__top-leaderboard:empty,.everyoneloves__mid-leaderboard:empty,.everyoneloves__bot-mid-leaderboard:empty height:90px;width:728px;box-sizing:border-box;

1

I'm trying to load a frozen Keras Graph with Batchnorm layers, but getting the error:

Message: TensorFlow.TFException : Input 0 of node
DenseNet/DenseBlock/ConvBlock/dense_0_0_bn/cond/ReadVariableOp/Switch was
passed float from DenseNet/DenseBlock/ConvBlock/dense_0_0_bn/gamma:0
incompatible with expected resource.

Normally the solution to this is to do: keras.backend.set_learning_phase(0), however when loading the graph in another API (for instance TensorflowSharp / TfLite) this isn't an option (as far as I can tell).

Here's how I'm currently saving the graph:

def freeze_session(session, keep_var_names=None, output_names=None, clear_devices=True):

 from tensorflow.python.framework.graph_util import convert_variables_to_constants
 import tensorflow as tf

 graph = session.graph
 with graph.as_default():
 freeze_var_names = list(set(v.op.name for v in tf.global_variables()).difference(keep_var_names or []))
 output_names = output_names or []
 output_names += [v.op.name for v in tf.global_variables()]
 # Graph -> GraphDef ProtoBuf
 input_graph_def = graph.as_graph_def()
 if clear_devices:
 for node in input_graph_def.node:
 node.device = ""

 for node in input_graph_def.node:
 if node.op == 'RefSwitch':
 for index in range(len(node.input)):
 if 'moving_' in node.input[index]:
 node.input[index] = node.input[index] + '/read'
 elif node.op == 'AssignSub':
 node.op = 'Sub'
 if 'use_locking' in node.attr: del node.attr['use_locking']

 frozen_graph = convert_variables_to_constants(session, input_graph_def,
 output_names, freeze_var_names)
 return frozen_graph

Is there any way I can programmatically remove the Batchnorm layers before saving so that I can load the model in an environment outside Keras?

python tensorflow keras

share|improve this question

edited Mar 25 at 9:23

Lukeyb

asked Mar 25 at 1:36

LukeybLukeyb

343215

add a comment | 

1

I'm trying to load a frozen Keras Graph with Batchnorm layers, but getting the error:

Message: TensorFlow.TFException : Input 0 of node
DenseNet/DenseBlock/ConvBlock/dense_0_0_bn/cond/ReadVariableOp/Switch was
passed float from DenseNet/DenseBlock/ConvBlock/dense_0_0_bn/gamma:0
incompatible with expected resource.

Normally the solution to this is to do: keras.backend.set_learning_phase(0), however when loading the graph in another API (for instance TensorflowSharp / TfLite) this isn't an option (as far as I can tell).

Here's how I'm currently saving the graph:

def freeze_session(session, keep_var_names=None, output_names=None, clear_devices=True):

 from tensorflow.python.framework.graph_util import convert_variables_to_constants
 import tensorflow as tf

 graph = session.graph
 with graph.as_default():
 freeze_var_names = list(set(v.op.name for v in tf.global_variables()).difference(keep_var_names or []))
 output_names = output_names or []
 output_names += [v.op.name for v in tf.global_variables()]
 # Graph -> GraphDef ProtoBuf
 input_graph_def = graph.as_graph_def()
 if clear_devices:
 for node in input_graph_def.node:
 node.device = ""

 for node in input_graph_def.node:
 if node.op == 'RefSwitch':
 for index in range(len(node.input)):
 if 'moving_' in node.input[index]:
 node.input[index] = node.input[index] + '/read'
 elif node.op == 'AssignSub':
 node.op = 'Sub'
 if 'use_locking' in node.attr: del node.attr['use_locking']

 frozen_graph = convert_variables_to_constants(session, input_graph_def,
 output_names, freeze_var_names)
 return frozen_graph

Is there any way I can programmatically remove the Batchnorm layers before saving so that I can load the model in an environment outside Keras?

python tensorflow keras

share|improve this question

edited Mar 25 at 9:23

Lukeyb

asked Mar 25 at 1:36

LukeybLukeyb

343215

add a comment | 

I'm trying to load a frozen Keras Graph with Batchnorm layers, but getting the error:

Message: TensorFlow.TFException : Input 0 of node
DenseNet/DenseBlock/ConvBlock/dense_0_0_bn/cond/ReadVariableOp/Switch was
passed float from DenseNet/DenseBlock/ConvBlock/dense_0_0_bn/gamma:0
incompatible with expected resource.

Normally the solution to this is to do: keras.backend.set_learning_phase(0), however when loading the graph in another API (for instance TensorflowSharp / TfLite) this isn't an option (as far as I can tell).

Here's how I'm currently saving the graph:

def freeze_session(session, keep_var_names=None, output_names=None, clear_devices=True):

 from tensorflow.python.framework.graph_util import convert_variables_to_constants
 import tensorflow as tf

 graph = session.graph
 with graph.as_default():
 freeze_var_names = list(set(v.op.name for v in tf.global_variables()).difference(keep_var_names or []))
 output_names = output_names or []
 output_names += [v.op.name for v in tf.global_variables()]
 # Graph -> GraphDef ProtoBuf
 input_graph_def = graph.as_graph_def()
 if clear_devices:
 for node in input_graph_def.node:
 node.device = ""

 for node in input_graph_def.node:
 if node.op == 'RefSwitch':
 for index in range(len(node.input)):
 if 'moving_' in node.input[index]:
 node.input[index] = node.input[index] + '/read'
 elif node.op == 'AssignSub':
 node.op = 'Sub'
 if 'use_locking' in node.attr: del node.attr['use_locking']

 frozen_graph = convert_variables_to_constants(session, input_graph_def,
 output_names, freeze_var_names)
 return frozen_graph

Is there any way I can programmatically remove the Batchnorm layers before saving so that I can load the model in an environment outside Keras?

python tensorflow keras

share|improve this question

edited Mar 25 at 9:23

Lukeyb

asked Mar 25 at 1:36

LukeybLukeyb

343215

Message: TensorFlow.TFException : Input 0 of node
DenseNet/DenseBlock/ConvBlock/dense_0_0_bn/cond/ReadVariableOp/Switch was
passed float from DenseNet/DenseBlock/ConvBlock/dense_0_0_bn/gamma:0
incompatible with expected resource.

def freeze_session(session, keep_var_names=None, output_names=None, clear_devices=True):

 from tensorflow.python.framework.graph_util import convert_variables_to_constants
 import tensorflow as tf

 graph = session.graph
 with graph.as_default():
 freeze_var_names = list(set(v.op.name for v in tf.global_variables()).difference(keep_var_names or []))
 output_names = output_names or []
 output_names += [v.op.name for v in tf.global_variables()]
 # Graph -> GraphDef ProtoBuf
 input_graph_def = graph.as_graph_def()
 if clear_devices:
 for node in input_graph_def.node:
 node.device = ""

 for node in input_graph_def.node:
 if node.op == 'RefSwitch':
 for index in range(len(node.input)):
 if 'moving_' in node.input[index]:
 node.input[index] = node.input[index] + '/read'
 elif node.op == 'AssignSub':
 node.op = 'Sub'
 if 'use_locking' in node.attr: del node.attr['use_locking']

 frozen_graph = convert_variables_to_constants(session, input_graph_def,
 output_names, freeze_var_names)
 return frozen_graph

share|improve this question

edited Mar 25 at 9:23

Lukeyb

asked Mar 25 at 1:36

LukeybLukeyb

343215

share|improve this question

edited Mar 25 at 9:23

Lukeyb

asked Mar 25 at 1:36

LukeybLukeyb

343215

asked Mar 25 at 1:36

LukeybLukeyb

343215

asked Mar 25 at 1:36

LukeybLukeyb

343215