In [1]:

    

input_fld = 'input_fld_path'
weight_file = 'kerasmodel_file_name located inside input_fld'
output_node_names_of_input_network = "pred0" #comma separated
write_graph_def_ascii_flag = True
output_node_names_of_final_network = 'output_node' #comma separated
output_graph_name = 'constant_graph_weights.pb'


    

In [2]:

    

from keras.models import load_model
import tensorflow as tf
import os
import os.path as osp

output_fld = input_fld + 'tensorflow_model/'
if not os.path.isdir(output_fld):
    os.mkdir(output_fld)
weight_file_path = osp.join(input_fld, weight_file)


    

    




Using TensorFlow backend.

In [3]:

    

net_model = load_model(weight_file_path)

num_output = len(output_node_names_of_input_network.split(','))
pred_node_names = output_node_names_of_final_network.split(',')
pred = [None]*num_output
for i in range(num_output):
    pred[i] = tf.identity(net_model.output[i], name=pred_node_names[i])
print('output nodes names are: ', pred_node_names)


    

    




output nodes names are:  ['output_node']

In [4]:

    

pred
pred_node_names


    

    
Out[4]:





['output_node']

In [5]:

    

import keras
from keras.models import load_model
from tensorflow.python.framework.graph_util import convert_variables_to_constants
from tensorflow.python.saved_model import utils
import tensorflow as tf
from tensorflow.core.framework import graph_pb2
from tensorflow.core.protobuf import saver_pb2
from tensorflow.python.client import session
# from tensorflow.python.framework import graph_io
from tensorflow.python.framework import importer
from tensorflow.python.framework import ops
from tensorflow.python.framework import test_util
from tensorflow.python.framework import graph_util
from tensorflow.python.ops import math_ops
# from tensorflow.python.ops import variable
from tensorflow.python.platform import test
from tensorflow.python.training import saver as saver_lib
from tensorflow.contrib.session_bundle import exporter


    

In [6]:

    

import tensorflow as tf
checkpoint_file = output_fld #+ "checkpoint"
checkpoint_state_name = "checkpoint_state"
input_graph_name = output_fld + 'only_the_graph_def.pb'
output_graph_path = os.path.join(output_fld, output_graph_name)
keras_architecture = osp.join(output_fld, 'keras_architecture_json')
keras_weights = osp.join(output_fld, 'keras_weights')


    

In [7]:

    

from keras import backend as K
sess = K.get_session()
graph = tf.get_default_graph()
input_graph_def = graph.as_graph_def()
saver = tf.train.Saver()
saver.save(sess, checkpoint_file, global_step=0, latest_filename='checkpoint_state')
if write_graph_def_ascii_flag:
    tf.train.write_graph(input_graph_def, output_fld, 'only_the_graph_def.pb.ascii', as_text=True)
tf.train.write_graph(input_graph_def, output_fld, 'only_the_graph_def.pb', as_text=False)
writer = tf.summary.FileWriter(output_fld, graph=graph)


    

In [8]:

    

import freeze_graph
import os
K._LEARNING_PHASE = tf.constant(0)

input_saver_def_path = "" # deprecated
input_binary = True

restore_op_name = "save/restore_all"
filename_tensor_name = "save/Const:0"

clear_devices = False

freeze_graph.freeze_graph(input_graph_name, input_saver_def_path,
                          input_binary, checkpoint_file+'-0',
                          output_node_names_of_final_network, restore_op_name,
                          filename_tensor_name, output_graph_path,
                          clear_devices, "", None)


    

    




INFO:tensorflow:Restoring parameters from /home/amir/deep-batch/trained/KerasEchoQualityMultiStream/snapshots/freeze_uniformValid_bn/s/tensorflow_model/-0
INFO:tensorflow:Froze 27 variables.
Converted 27 variables to const ops.
541 ops in the final graph.