def testPluginTagToContent_PluginsCannotJumpOnTheBandwagon(self):
# If there are multiple `SummaryMetadata` for a given tag, and the
# set of plugins in the `plugin_data` of second is different from
# that of the first, then the second set should be ignored.
logdir = self.get_temp_dir()
summary_metadata_1 = tf.SummaryMetadata(
display_name='current tagee',
summary_description='no',
plugin_data=[
tf.SummaryMetadata.PluginData(plugin_name='outlet',
content='120v')
])
self._writeMetadata(logdir, summary_metadata_1, nonce='1')
acc = ea.EventAccumulator(logdir)
acc.Reload()
summary_metadata_2 = tf.SummaryMetadata(
display_name='tagee of the future',
summary_description='definitely not',
plugin_data=[
tf.SummaryMetadata.PluginData(plugin_name='plug',
content='110v')
])
self._writeMetadata(logdir, summary_metadata_2, nonce='2')
acc.Reload()
self.assertEqual(acc.PluginTagToContent('outlet'),
{'you_are_it': '120v'})
with six.assertRaisesRegex(self, KeyError, 'plug'):
acc.PluginTagToContent('plug')
def printLastLayer(self, lastLayer, test_data_readable_input, output,
variable_writer):
input = test_data_readable_input['X_test']
with open(
"Data_Set/" + FLAGS.dataset_name + "/" +
str(FLAGS.network_node) + "-node/matriceShortestPath.txt",
'rb') as f:
resultMatrix = pickle.load(f)
for i in range(0, 50):
str_result = " Input " + str(i) + " " + str(input[i]) + "\n"
summary = tf.Summary()
text_tensor = tf.make_tensor_proto(str_result, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary.value.add(tag=str(i) + " Input ",
metadata=meta,
tensor=text_tensor)
variable_writer.add_summary(summary, i)
source = int(input[i][FLAGS.network_link])
destination = int(input[i][FLAGS.network_link + 1])
str_result = " Output :" + str(i) + " " + str(
output[i]) + "\n paths : " + str(
resultMatrix[traffic_s_d_to_index(source, destination)])
summary = tf.Summary()
text_tensor = tf.make_tensor_proto(str_result, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary.value.add(tag=str(i) + " Output ",
metadata=meta,
tensor=text_tensor)
variable_writer.add_summary(summary, i)
str_result = " Last Layer " + str(i) + " " + str(
lastLayer[i]) + "\n\n\n "
summary = tf.Summary()
text_tensor = tf.make_tensor_proto(str_result, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary.value.add(tag=str(i) + " Last Layer ",
metadata=meta,
tensor=text_tensor)
variable_writer.add_summary(summary, i)
def create_summary_metadata(display_name, description, num_thresholds):
"""Create a `tf.SummaryMetadata` proto for pr_curves plugin data.
Arguments:
display_name: The display name used in TensorBoard.
description: The description to show in TensorBoard.
num_thresholds: The number of thresholds to use for PR curves.
Returns:
A `tf.SummaryMetadata` protobuf object.
"""
pr_curve_plugin_data = plugin_data_pb2.PrCurvePluginData(
version=PROTO_VERSION, num_thresholds=num_thresholds)
content = pr_curve_plugin_data.SerializeToString()
if USE_TF:
return tf.SummaryMetadata(
display_name=display_name,
summary_description=description,
plugin_data=tf.SummaryMetadata.PluginData(plugin_name=PLUGIN_NAME,
content=content))
else:
return summary_pb2.SummaryMetadata(
display_name=display_name,
summary_description=description,
plugin_data=summary_pb2.SummaryMetadata.PluginData(
plugin_name=PLUGIN_NAME, content=content))
def custom_summary(text, tag):
text_tensor = tf.make_tensor_proto(text, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary()
summary.value.add(tag=tag, metadata=meta, tensor=text_tensor)
return summary
def op(name, data, display_name=None, description=None, collections=None):
"""Create a TensorFlow summary op for a rank-2 `Tensor`.
Args:
name: A unique name for the generated summary node.
data: A real numeric rank-2 `Tensor`. Must have `dtype` castable to
`tf.float64`.
display_name: Optional name for this summary in TensorBoard, as a
constant `str`. Defaults to `name`.
description: Optional long-form description for this summary, as a
constant `str`. Markdown is supported. Defaults to empty.
collections: Optional list of graph collections to add the summary
op to. Defaults to `['summaries']`.
Returns:
A TensorFlow summary op.
"""
if display_name is None:
display_name = name
summary_metadata = tf.SummaryMetadata(
display_name=display_name,
summary_description=description,
plugin_data=tf.SummaryMetadata.PluginData(plugin_name=PLUGIN_NAME)
)
with tf.control_dependencies([tf.assert_rank(data, 2)]):
return tf.summary.tensor_summary(
name=name,
tensor=tf.cast(data, tf.float64),
summary_metadata=summary_metadata,
collections=collections
)
def add_model_metadata(logdir, directory_path, flags, properties):
summary_writer = tf.summary.FileWriter(logdir)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary()
overview = get_model_overview(directory_path)
hyper_params = get_hyper_parameters(flags, properties)
gan_summary = get_gan_summary()
write_to_file(logdir, overview, hyper_params, gan_summary, flags)
summary.value.add(tag="1_Overview",
metadata=meta,
tensor=tf.make_tensor_proto(overview, dtype=tf.string))
summary.value.add(tag="2_Hyperparameters",
metadata=meta,
tensor=tf.make_tensor_proto(hyper_params,
dtype=tf.string))
summary.value.add(tag="3_Model",
metadata=meta,
tensor=tf.make_tensor_proto(gan_summary,
dtype=tf.string))
summary_writer.add_summary(summary)
summary_writer.flush()
summary_writer.close()
tf.logging.info("Saved meta data in {}".format(logdir.replace(
"\\", "\\\\")))
def pb(tag, guest, display_name=None, description=None):
"""Create a greeting summary for the given guest.
Arguments:
tag: The string tag associated with the summary.
guest: The string name of the guest to greet.
display_name: If set, will be used as the display name in
TensorBoard. Defaults to `tag`.
description: A longform readable description of the summary data.
Markdown is supported.
"""
message = 'Hello, %s!' % guest
tensor = tf.make_tensor_proto(message, dtype=tf.string)
# We have no metadata to store, but we do need to add a plugin_data entry
# so that we know this summary is associated with the greeter plugin.
# We could use this entry to pass additional metadata other than the
# PLUGIN_NAME by using the content parameter.
summary_metadata = tf.SummaryMetadata(
display_name=display_name,
summary_description=description,
plugin_data=tf.SummaryMetadata.PluginData(plugin_name=PLUGIN_NAME))
summary = tf.Summary()
summary.value.add(tag=tag, metadata=summary_metadata, tensor=tensor)
return summary
def op(name, value, display_name=None, description=None, collections=None):
"""Create a TensorFlow summary op to record data associated with a particular the given guest.
Arguments:
name: A name for this summary operation.
guest: A rank-0 string `Tensor`.
display_name: If set, will be used as the display name
in TensorBoard. Defaults to `name`.
description: A longform readable description of the summary data.
Markdown is supported.
collections: Which TensorFlow graph collections to add the summary
op to. Defaults to `['summaries']`. Can usually be ignored.
"""
# The `name` argument is used to generate the summary op node name.
# That node name will also involve the TensorFlow name scope.
# By having the display_name default to the name argument, we make
# the TensorBoard display clearer.
if display_name is None:
display_name = name
# We could pass additional metadata other than the PLUGIN_NAME within the
# plugin data by using the content parameter, but we don't need any metadata
# for this simple example.
summary_metadata = tf.SummaryMetadata(
display_name=display_name,
summary_description=description,
plugin_data=tf.SummaryMetadata.PluginData(plugin_name=PLUGIN_NAME))
# Return a summary op that is properly configured.
return tf.summary.tensor_summary(name,
value,
summary_metadata=summary_metadata,
collections=collections)
def op(name, guest, display_name=None, description=None, collections=None):
"""Create a TensorFlow summary op to greet the given guest.
Arguments:
name: A name for this summary operation.
guest: A rank-0 string `Tensor`.
display_name: If set, will be used as the display name
in TensorBoard. Defaults to `name`.
description: A longform readable description of the summary data.
Markdown is supported.
collections: Which TensorFlow graph collections to add the summary
op to. Defaults to `['summaries']`. Can usually be ignored.
"""
# The `name` argument is used to generate the summary op node name.
# That node name will also involve the TensorFlow name scope.
# By having the display_name default to the name argument, we make
# the TensorBoard display clearer.
if display_name is None:
display_name = name
summary_metadata = tf.SummaryMetadata()
# We could put additional metadata other than the PLUGIN_NAME,
# but we don't need any metadata for this simple example.
summary_metadata.plugin_data.add(plugin_name=PLUGIN_NAME, content="")
message = tf.string_join(['Hello, ', guest, '!'])
# Return a summary op that is properly configured.
return tf.summary.tensor_summary(name,
message,
display_name=display_name,
summary_metadata=summary_metadata,
summary_description=description,
collections=collections)
def test_session_start_pb(self):
start_time_secs = 314160
session_start_info = plugin_data_pb2.SessionStartInfo(
model_uri="//model/uri",
group_name="session_group",
start_time_secs=start_time_secs)
session_start_info.hparams["param1"].string_value = "string"
session_start_info.hparams["param2"].number_value = 5.0
session_start_info.hparams["param3"].bool_value = False
self.assertEqual(
summary.session_start_pb(hparams={
"param1": "string",
"param2": 5,
"param3": False
},
model_uri="//model/uri",
group_name="session_group",
start_time_secs=start_time_secs),
tf.Summary(value=[
tf.Summary.Value(
tag="_hparams_/session_start_info",
metadata=tf.SummaryMetadata(
plugin_data=tf.SummaryMetadata.PluginData(
plugin_name="hparams",
content=(plugin_data_pb2.HParamsPluginData(
version=0,
session_start_info=session_start_info).
SerializeToString()))))
]))
def create_summary_metadata(display_name, description):
content = HistogramMetadata()
metadata = tf.SummaryMetadata(display_name=display_name,
summary_description=description)
metadata.plugin_data.add(plugin_name=PLUGIN_NAME,
content=json.dumps(content._asdict())) # pylint: disable=protected-access
return metadata
def _architecture_metric_fn(**kwargs):
"""Manually creates the tf.metric with a serialized tf.Summary proto."""
del kwargs # Unused.
# TODO: Should architecture.subnetworks be sorted by iteration
# number first? Or perhaps, to make this more general, to have one line for
# each iteration, with "|" as a delimiter if there are multiple subnetworks
# in one iteration? Something like:
# 0 linear
# 1 dnn_width_32_depth_1 | dnn_width_64_depth_1
# 2
# 3 dnn_with_32_depth_2
architecture_ = " | ".join(
[name for _, name in architecture.subnetworks])
architecture_ = "| {} |".format(architecture_)
summary_metadata = tf.SummaryMetadata(
plugin_data=tf.SummaryMetadata.PluginData(plugin_name="text"))
summary_proto = tf.summary.Summary()
summary_proto.value.add(metadata=summary_metadata,
tag="architecture/adanet",
tensor=tf.make_tensor_proto(architecture_,
dtype=tf.string))
architecture_summary = tf.convert_to_tensor(
summary_proto.SerializeToString(), name="architecture")
return {
"architecture/adanet/ensembles": (architecture_summary, tf.no_op())
}
def log_text(writer, tag, text):
text_tensor = tf.make_tensor_proto(text, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary()
summary.value.add(tag=tag, metadata=meta, tensor=text_tensor)
writer.add_summary(summary)
def pb(tag, guest, display_name=None, description=None):
"""Create a greeting summary for the given guest.
Arguments:
tag: The string tag associated with the summary.
guest: The string name of the guest to greet.
display_name: If set, will be used as the display name in
TensorBoard. Defaults to `tag`.
description: A longform readable description of the summary data.
Markdown is supported.
"""
message = 'Hello, %s!' % guest
tensor = tf.make_tensor_proto(message, dtype=tf.string)
summary_metadata = tf.SummaryMetadata(display_name=display_name,
summary_description=description)
metadata_content = '{}' # We have no metadata to store.
summary_metadata.plugin_data.add(plugin_name=PLUGIN_NAME,
content=metadata_content)
summary = tf.Summary()
summary.value.add(tag=tag,
metadata=summary_metadata,
tensor=tensor)
return summary
def text_summary(self, text=''):
text_tensor = tf.make_tensor_proto(text, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary()
summary.value.add(tag="Description", metadata=meta, tensor=text_tensor)
self.writer.add_summary(summary)
def log_markdown(self, tag: str, text_in_markdown: str, step: int):
text_tensor = tf.make_tensor_proto(text_in_markdown, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary()
summary.value.add(tag=tag, metadata=meta, tensor=text_tensor)
self._summary_writer.add_summary(summary, step)
def pb(tag, data, display_name=None, description=None):
"""Create a summary for the given matrix.
Args:
tag: A unique name for the generated summary.
data: A `np.array` or array-like form with rank 2. Must have type
castable to `np.float64`.
display_name: Optional name for this summary in TensorBoard, as a
`str`. Defaults to `name`.
description: Optional long-form description for this summary, as a
`str`. Markdown is supported. Defaults to empty.
Returns:
A `tf.Summary` protobuf object.
"""
data = np.array(data).astype(np.float64)
tensor = tf.make_tensor_proto(data, dtype=tf.float64)
if display_name is None:
display_name = tag
summary_metadata = tf.SummaryMetadata(
display_name=display_name,
summary_description=description,
plugin_data=tf.SummaryMetadata.PluginData(plugin_name=PLUGIN_NAME)
)
summary = tf.Summary()
summary.value.add(tag=tag, metadata=summary_metadata, tensor=tensor)
return summary
def pinnboard_summary(params,
display_name=None,
description=None,
collections=None):
"""Create a TensorFlow summary op to greet the given guest.
Arguments:
params: parameter dictionary of the model
display_name: If set, will be used as the display name
in TensorBoard. Defaults to `name`.
description: A longform readable description of the summary data.
Markdown is supported.
collections: Which TensorFlow graph collections to add the summary
op to. Defaults to `['summaries']`. Can usually be ignored.
"""
tensors = pinnboard_tensors(params)
summary_metadata = {
k: tf.SummaryMetadata(
summary_description=description,
plugin_data=tf.SummaryMetadata.PluginData(plugin_name=PLUGIN_NAME))
for k in tensors.keys()
}
summary_op = {
k: tf.summary.tensor_summary('pinnboard.' + k,
val,
summary_metadata=summary_metadata[k],
collections=collections)
for k, val in tensors.items()
}
# Return disctionary of configured summary operations
return summary_op
def add_text(self, tag, value, step):
text_tensor = tf.make_tensor_proto(value, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary()
summary.value.add(tag=tag, metadata=meta, tensor=text_tensor)
self.writer.add_summary(summary, step)
def add_text_tb(summary_writer, text, tag='default_tag', index=0):
text_tensor = tf.make_tensor_proto(text, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary()
summary.value.add(tag=tag, metadata=meta, tensor=text_tensor)
summary_writer.add_summary(summary, index) #tf.summary.text('info', tf.constant("This is test data"))
def _testTFSummaryTensor_SizeGuidance(self,
plugin_name,
tensor_size_guidance,
steps,
expected_count):
event_sink = _EventGenerator(self, zero_out_timestamps=True)
writer = tf.summary.FileWriter(self.get_temp_dir())
writer.event_writer = event_sink
with self.test_session() as sess:
summary_metadata = tf.SummaryMetadata(
plugin_data=tf.SummaryMetadata.PluginData(plugin_name=plugin_name,
content=b'{}'))
tf.summary.tensor_summary('scalar', tf.constant(1.0),
summary_metadata=summary_metadata)
merged = tf.summary.merge_all()
for step in xrange(steps):
writer.add_summary(sess.run(merged), global_step=step)
accumulator = ea.EventAccumulator(
event_sink, tensor_size_guidance=tensor_size_guidance)
accumulator.Reload()
tensors = accumulator.Tensors('scalar')
self.assertEqual(len(tensors), expected_count)
def add_code_summary(self):
code_string = "\n".join(open(os.path.basename(__file__), 'r').readlines())
text_tensor = tf.make_tensor_proto(self.DESCRIPTION + "\n\n" + code_string, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary()
summary.value.add(tag="Hyper parameters", metadata=meta, tensor=text_tensor)
self.summary_writer.add_summary(summary)
def tensorboard_text(writer, tag, value, step=0):
text_tensor = tf.make_tensor_proto(value, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary()
summary.value.add(tag=tag, metadata=meta, tensor=text_tensor)
writer.add_summary(summary)
writer.flush()
def summary_text(name, tensor, family=None, step=None):
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
return summary.generic(name,
tensor,
metadata=meta,
family=family,
step=step)
def log_hparams(hparams, writer):
value = get_hparams_text(hparams.values())
text_tensor = tf.make_tensor_proto(value, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary()
summary.value.add(tag="hyperparameters", metadata=meta, tensor=text_tensor)
writer.add_summary(summary)
def RecordDescription(writer, msg):
text_tensor = tf.make_tensor_proto(msg, dtype=tf.string)
summary = tf.Summary()
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary.value.add(tag="Description", metadata=meta, tensor=text_tensor)
writer.add_summary(summary)
writer.flush()
def text_summary(self, tag, value):
"""Log a text variable."""
text_tensor = tf.make_tensor_proto(value, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary(value=[tf.Summary.Value(tag=tag, metadata=meta, tensor=text_tensor)])
self.writer.add_summary(summary)
def add_summary(self, text):
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary()
text_tensor = tf.make_tensor_proto(text, dtype=tf.string)
summary.value.add(tag="BLAST_" + self.running_mode,
metadata=meta,
tensor=text_tensor)
self._summary_writer.add_summary(summary, self.global_step)
def log_board(board_str, summary_writer, summary_tag, step=0):
board_str = board_str.replace('\n', '\n\t')
text_tensor = tf.make_tensor_proto(board_str, dtype=tf.string)
meta = tf.SummaryMetadata()
meta.plugin_data.plugin_name = "text"
summary = tf.Summary()
summary.value.add(tag=summary_tag, metadata=meta, tensor=text_tensor)
summary_writer.add_summary(summary, step)
summary_writer.flush()
def testSummaryMetadata(self):
logdir = self.get_temp_dir()
summary_metadata = tf.SummaryMetadata(display_name='current tagee',
summary_description='no')
summary_metadata.plugin_data.plugin_name = 'outlet'
self._writeMetadata(logdir, summary_metadata)
acc = ea.EventAccumulator(logdir)
acc.Reload()
self.assertProtoEquals(summary_metadata,
acc.SummaryMetadata('you_are_it'))
