Commit 5bc70b70 authored by Oleg Dzhimiev's avatar Oleg Dzhimiev

cleanup

parent 5414fe8c
......@@ -151,7 +151,7 @@ rv_stage1_out = tf.get_variable("rv_stage1_out",
shape=[HEIGHT * WIDTH, NN_LAYOUT1[-1]],
dtype=tf.float32,
initializer=tf.zeros_initializer,
collections = [GraphKeys.LOCAL_VARIABLES], trainable=False)
collections = [GraphKeys.LOCAL_VARIABLES],trainable=False)
'''
#rv_stageX_out_init_placeholder = tf.placeholder(tf.float32, shape=[HEIGHT * WIDTH, NN_LAYOUT1[-1]])
......@@ -267,10 +267,16 @@ with tf.Session() as sess:
saver.restore(sess, files["checkpoints"])
#tf.add_to_collection(GraphKeys.GLOBAL_VARIABLES,rv_stage1_out)
'''
rv_stage1_out belongs to GraphKeys.LOCAL_VARIABLES
Now when weights/biases are restored from 'checkpoints',
that do not have this variable, add it to globals.
Actually it could have been declared right here - this
needs testing.
'''
tf.add_to_collection(GraphKeys.GLOBAL_VARIABLES, rv_stage1_out)
saver.save(sess, files["inference"]) #TODO: move to different subdir
#saver2.save(sess, files["inference"]+"_2") #TODO: move to different subdir
saver.save(sess, files["inference"])
merged = tf.summary.merge_all()
writer = tf.summary.FileWriter(ROOT_PATH, sess.graph)
......@@ -329,8 +335,9 @@ with tf.Session() as sess:
Remove dataset_img (if it is not [0] to reduce memory footprint
"""
image_data[nimg] = None
meta_graph_def = tf.train.export_meta_graph(files["inference"]+'.meta')
# is this needed? why would it be?
#meta_graph_def = tf.train.export_meta_graph(files["inference"]+'.meta')
if lf:
......
......@@ -265,13 +265,15 @@ with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.local_variables_initializer())
saver.restore(sess, files["checkpoints"])
# so, here I need to restore from inference and save to inference_global
#saver.restore(sess, files["checkpoints"])
saver.restore(sess, files["inference"])
#tf.add_to_collection(GraphKeys.GLOBAL_VARIABLES,rv_stage1_out)
# now add to global
tf.add_to_collection(GraphKeys.GLOBAL_VARIABLES,rv_stage1_out)
saver.save(sess, files["inference"]) #TODO: move to different subdir
#saver2.save(sess, files["inference"]+"_2") #TODO: move to different subdir
saver.save(sess, 'data_sets/tf_data_5x5_main_13_heur/inference_global/model')
merged = tf.summary.merge_all()
writer = tf.summary.FileWriter(ROOT_PATH, sess.graph)
lf = None
......
......@@ -129,31 +129,29 @@ try:
except:
pass
from tensorflow.python.framework.ops import GraphKeys
with tf.Session() as sess:
# default option
use_saved_model = False
if os.path.isdir(dirs['exportdir']):
# check if dir contains "Saved Model" model
use_saved_model = tf.saved_model.loader.maybe_saved_model_directory(dirs['exportdir'])
if use_saved_model:
print("Model restore: using Saved_Model model MetaGraph protocol buffer")
meta_graph_source = tf.saved_model.loader.load(sess, [tf.saved_model.tag_constants.SERVING], dirs['exportdir'])
else:
print("Model restore: using conventionally saved model, but saving Saved Model for the next run")
meta_graph_source = files["inference"]+'.meta'
print("MetaGraph source = "+str(meta_graph_source))
#meta_graph_source = files["inference"]+'_2.meta'
# remove 'exportdir' even it exsits and has anything
shutil.rmtree(dirs['exportdir'], ignore_errors=True)
builder = tf.saved_model.builder.SavedModelBuilder(dirs['exportdir'])
# Actually, refresh all the time and have an extra script to restore from it.
# use_Saved_Model = False
#if os.path.isdir(dirs['exportdir']):
# # check if dir contains "Saved Model" model
# use_saved_model = tf.saved_model.loader.maybe_saved_model_directory(dirs['exportdir'])
#if use_saved_model:
# print("Model restore: using Saved_Model model MetaGraph protocol buffer")
# meta_graph_source = tf.saved_model.loader.load(sess, [tf.saved_model.tag_constants.SERVING], dirs['exportdir'])
#else:
meta_graph_source = files["inference"]+'.meta'
print("Model restore: using conventionally saved model, but saving Saved Model for the next run")
print("MetaGraph source = "+str(meta_graph_source))
infer_saver = tf.train.import_meta_graph(meta_graph_source)
graph=tf.get_default_graph()
ph_corr2d = graph.get_tensor_by_name('ph_corr2d:0')
ph_target_disparity = graph.get_tensor_by_name('ph_target_disparity:0')
ph_ntile = graph.get_tensor_by_name('ph_ntile:0')
......@@ -164,27 +162,17 @@ with tf.Session() as sess:
if not USE_SPARSE_ONLY: #Does it reduce the graph size?
stage2_out_full = graph.get_tensor_by_name('Disparity_net/stage2_out_full:0')
'''
if not use_saved_model:
rv_stage1_out = tf.get_variable("rv_stage1_out",
shape=[78408, 32],
dtype=tf.float32,
initializer=tf.zeros_initializer)
#collections = [GraphKeys.LOCAL_VARIABLES],trainable=False)
'''
sess.run(tf.global_variables_initializer())
sess.run(tf.local_variables_initializer())
if not use_saved_model:
infer_saver.restore(sess, files["inference"]) # after initializers, of course
else:
infer_saver.restore(sess, dirs['exportdir']+"/variables/variables.data-00000-of-00001")
#infer_saver.restore(sess, files["inference"]+"_2") # after initializers, of course
infer_saver.restore(sess, files["inference"])
merged = tf.summary.merge_all()
writer = tf.summary.FileWriter(ROOT_PATH, sess.graph)
lf = None
if LOGPATH:
lf=open(LOGPATH,"w") #overwrite previous (or make it "a"?
......@@ -234,11 +222,14 @@ with tf.Session() as sess:
"""
image_data[nimg] = None
if not use_saved_model:
#builder.add_meta_graph_and_variables(sess,PB_TAGS)
builder.add_meta_graph_and_variables(sess,[tf.saved_model.tag_constants.SERVING])
builder.save(True)
#builder.save(False)
#builder.add_meta_graph_and_variables(sess,PB_TAGS)
# clean
shutil.rmtree(dirs['exportdir'], ignore_errors=True)
# save MetaGraph to Saved_Model as *.pb
builder = tf.saved_model.builder.SavedModelBuilder(dirs['exportdir'])
builder.add_meta_graph_and_variables(sess,[tf.saved_model.tag_constants.SERVING])
#builder.save(True)
builder.save(False) # True = *.pbtxt, False = *.pb
if lf:
lf.close()
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment