def add_logs(self, **kwargs):
if 'logfile' in kwargs:
if 'name' in kwargs:
log_name = str(kwargs['name'])
else:
log_name = kwargs['logfile']
event_acc = event_accumulator.EventAccumulator(
kwargs['logfile'], self.tf_size_guidance)
event_acc.Reload()
self.event_accs.append(event_acc)
self.log_file_names.append(
(kwargs["logfile"], cycle_num, kwargs["logfile"]))
elif 'logdir' in kwargs:
idx = -1
for root, dirs, files in os.walk(kwargs['logdir']):
for file in files:
if "events.out" in file:
idx += 1
filename = os.path.join(root, file)
cycle_num = 0
if 'name' in kwargs:
if "cycles_named" in kwargs:
cycles_named = kwargs["cycles_named"]
else:
cycles_named = False
if cycles_named:
try:
found = re.search("cycle_\d", filename)
cycle_num = str(found[0].rstrip()[-1])
log_name = kwargs['name']
except:
raise ValueError(
"Logfile: {} does not contain cycle num"
.format(filename))
else:
cycle_num = str(idx)
log_name = kwargs['name']
else:
log_name = filename
event_acc = event_accumulator.EventAccumulator(
filename, self.tf_size_guidance)
self.event_accs.append(event_acc)
self.log_file_names.append(
(log_name, cycle_num, filename))
if idx >= 0:
print("I added " + str(idx + 1) + " different logfiles: " +
filename)
else:
print("No logfiles found")
tags = []
for event_acc in self.event_accs:
tags.append(event_acc.Tags())
return tags
def testReload(self):
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
acc.Reload()
self.assertEqual(acc.Tags(), self.empty)
gen.AddScalar('s1')
gen.AddScalar('s2')
gen.AddHistogram('hst1')
gen.AddHistogram('hst2')
gen.AddImage('im1')
gen.AddImage('im2')
gen.AddAudio('snd1')
gen.AddAudio('snd2')
acc.Reload()
self.assertTagsEqual(
acc.Tags(), {
ea.IMAGES: ['im1', 'im2'],
ea.AUDIO: ['snd1', 'snd2'],
ea.SCALARS: ['s1', 's2'],
ea.HISTOGRAMS: ['hst1', 'hst2'],
ea.COMPRESSED_HISTOGRAMS: ['hst1', 'hst2'],
ea.GRAPH: False,
ea.META_GRAPH: False,
ea.RUN_METADATA: []
})
def testImages(self):
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
im1 = ea.ImageEvent(wall_time=1,
step=10,
encoded_image_string=b'big',
width=400,
height=300)
im2 = ea.ImageEvent(wall_time=2,
step=12,
encoded_image_string=b'small',
width=40,
height=30)
gen.AddImage('im1',
wall_time=1,
step=10,
encoded_image_string=b'big',
width=400,
height=300)
gen.AddImage('im2',
wall_time=2,
step=12,
encoded_image_string=b'small',
width=40,
height=30)
acc.Reload()
self.assertEqual(acc.Images('im1'), [im1])
self.assertEqual(acc.Images('im2'), [im2])
def testFirstEventTimestamp(self):
"""Test that FirstEventTimestamp() returns wall_time of the first event."""
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
gen.AddEvent(tf.Event(wall_time=10, step=20, file_version='brain.Event:2'))
gen.AddScalar('s1', wall_time=30, step=40, value=20)
self.assertEqual(acc.FirstEventTimestamp(), 10)
def testSessionLogStartMessageDiscardsExpiredEvents(self):
"""Test that SessionLog.START message discards expired events.
This discard logic is preferred over the out-of-order step discard logic,
but this logic can only be used for event protos which have the SessionLog
enum, which was introduced to event.proto for file_version >= brain.Event:2.
"""
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
gen.AddEvent(
tf.Event(wall_time=0, step=1, file_version='brain.Event:2'))
gen.AddScalar('s1', wall_time=1, step=100, value=20)
gen.AddScalar('s1', wall_time=1, step=200, value=20)
gen.AddScalar('s1', wall_time=1, step=300, value=20)
gen.AddScalar('s1', wall_time=1, step=400, value=20)
gen.AddScalar('s2', wall_time=1, step=202, value=20)
gen.AddScalar('s2', wall_time=1, step=203, value=20)
slog = SessionLog(status=SessionLog.START)
gen.AddEvent(tf.Event(wall_time=2, step=201, session_log=slog))
acc.Reload()
self.assertEqual([x.step for x in acc.Scalars('s1')], [100, 200])
self.assertEqual([x.step for x in acc.Scalars('s2')], [])
def testEventsDiscardedPerTagAfterRestartForFileVersionLessThan2(self):
"""Tests that event discards after restart, only affect the misordered tag.
If a step value is observed to be lower than what was previously seen,
this should force a discard of all previous items that are outdated, but
only for the out of order tag. Other tags should remain unaffected.
Only file versions < 2 use this out-of-order discard logic. Later versions
discard events based on the step value of SessionLog.START.
"""
warnings = []
self.stubs.Set(logging, 'warn', warnings.append)
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
gen.AddEvent(tf.Event(wall_time=0, step=0, file_version='brain.Event:1'))
gen.AddScalar('s1', wall_time=1, step=100, value=20)
gen.AddScalar('s1', wall_time=1, step=200, value=20)
gen.AddScalar('s1', wall_time=1, step=300, value=20)
gen.AddScalar('s1', wall_time=1, step=101, value=20)
gen.AddScalar('s1', wall_time=1, step=201, value=20)
gen.AddScalar('s1', wall_time=1, step=301, value=20)
gen.AddScalar('s2', wall_time=1, step=101, value=20)
gen.AddScalar('s2', wall_time=1, step=201, value=20)
gen.AddScalar('s2', wall_time=1, step=301, value=20)
acc.Reload()
## Check that we have discarded 200 and 300
self.assertEqual([x.step for x in acc.Scalars('s1')], [100, 101, 201, 301])
## Check that s1 discards do not affect s2
## i.e. check that only events from the out of order tag are discarded
self.assertEqual([x.step for x in acc.Scalars('s2')], [101, 201, 301])
def testCompressedHistogramsWithEmptyHistogram(self):
gen = _EventGenerator()
acc = ea.EventAccumulator(gen,
compression_bps=(0, 2500, 5000, 7500, 10000))
gen.AddHistogram('hst1',
wall_time=1,
step=10,
hmin=None,
hmax=None,
hnum=0,
hsum=0,
hsum_squares=0,
hbucket_limit=[1, 2, 3],
hbucket=[0, 0, 0])
acc.Reload()
# Create the expected values after compressing hst1
expected_vals1 = [
ea.CompressedHistogramValue(bp, val)
for bp, val in [(0, 0.0), (2500, 0), (5000, 0), (7500,
0), (10000, 0)]
]
expected_cmphst1 = ea.CompressedHistogramEvent(
wall_time=1, step=10, compressed_histogram_values=expected_vals1)
self.assertEqual(acc.CompressedHistograms('hst1'), [expected_cmphst1])
def testAudio(self):
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
snd1 = ea.AudioEvent(wall_time=1,
step=10,
encoded_audio_string=b'big',
content_type='audio/wav',
sample_rate=44100,
length_frames=441000)
snd2 = ea.AudioEvent(wall_time=2,
step=12,
encoded_audio_string=b'small',
content_type='audio/wav',
sample_rate=44100,
length_frames=44100)
gen.AddAudio('snd1',
wall_time=1,
step=10,
encoded_audio_string=b'big',
content_type='audio/wav',
sample_rate=44100,
length_frames=441000)
gen.AddAudio('snd2',
wall_time=2,
step=12,
encoded_audio_string=b'small',
content_type='audio/wav',
sample_rate=44100,
length_frames=44100)
acc.Reload()
self.assertEqual(acc.Audio('snd1'), [snd1])
self.assertEqual(acc.Audio('snd2'), [snd2])
def testTFSummaryTensor(self):
"""Verify processing of tf.summary.tensor."""
event_sink = _EventGenerator(self, zero_out_timestamps=True)
writer = SummaryToEventTransformer(event_sink)
with self.test_session() as sess:
summary_lib.tensor_summary('scalar', constant_op.constant(1.0))
summary_lib.tensor_summary('vector',
constant_op.constant([1.0, 2.0, 3.0]))
summary_lib.tensor_summary('string',
constant_op.constant(six.b('foobar')))
merged = summary_lib.merge_all()
summ = sess.run(merged)
writer.add_summary(summ, 0)
accumulator = ea.EventAccumulator(event_sink)
accumulator.Reload()
self.assertTagsEqual(accumulator.Tags(), {
ea.TENSORS: ['scalar', 'vector', 'string'],
})
scalar_proto = accumulator.Tensors('scalar')[0].tensor_proto
scalar = tensor_util.MakeNdarray(scalar_proto)
vector_proto = accumulator.Tensors('vector')[0].tensor_proto
vector = tensor_util.MakeNdarray(vector_proto)
string_proto = accumulator.Tensors('string')[0].tensor_proto
string = tensor_util.MakeNdarray(string_proto)
self.assertTrue(np.array_equal(scalar, 1.0))
self.assertTrue(np.array_equal(vector, [1.0, 2.0, 3.0]))
self.assertTrue(np.array_equal(string, six.b('foobar')))
def testGraphFromMetaGraphBecomesAvailable(self):
"""Test accumulator by writing values and then reading them."""
directory = os.path.join(self.get_temp_dir(),
'metagraph_test_values_dir')
if gfile.IsDirectory(directory):
gfile.DeleteRecursively(directory)
gfile.MkDir(directory)
writer = writer_lib.FileWriter(directory, max_queue=100)
with ops.Graph().as_default() as graph:
_ = constant_op.constant([2.0, 1.0])
# Add a graph to the summary writer.
meta_graph_def = saver.export_meta_graph(graph_def=graph.as_graph_def(
add_shapes=True))
writer.add_meta_graph(meta_graph_def)
writer.flush()
# Verify that we can load those events properly
acc = ea.EventAccumulator(directory)
acc.Reload()
self.assertTagsEqual(acc.Tags(), {
ea.GRAPH: True,
ea.META_GRAPH: True,
})
self.assertProtoEquals(graph.as_graph_def(add_shapes=True),
acc.Graph())
self.assertProtoEquals(meta_graph_def, acc.MetaGraph())
def testTFSummaryImage(self):
"""Verify processing of tf.summary.image."""
event_sink = _EventGenerator(self, zero_out_timestamps=True)
writer = SummaryToEventTransformer(event_sink)
with self.test_session() as sess:
ipt = array_ops.ones([10, 4, 4, 3], dtypes.uint8)
# This is an interesting example, because the old tf.image_summary op
# would throw an error here, because it would be tag reuse.
# Using the tf node name instead allows argument re-use to the image
# summary.
with ops.name_scope('1'):
summary_lib.image('images', ipt, max_outputs=1)
with ops.name_scope('2'):
summary_lib.image('images', ipt, max_outputs=2)
with ops.name_scope('3'):
summary_lib.image('images', ipt, max_outputs=3)
merged = summary_lib.merge_all()
writer.add_graph(sess.graph)
for i in xrange(10):
summ = sess.run(merged)
writer.add_summary(summ, global_step=i)
accumulator = ea.EventAccumulator(event_sink)
accumulator.Reload()
tags = [
u'1/images/image', u'2/images/image/0', u'2/images/image/1',
u'3/images/image/0', u'3/images/image/1', u'3/images/image/2'
]
self.assertTagsEqual(accumulator.Tags(), {
ea.IMAGES: tags,
ea.GRAPH: True,
ea.META_GRAPH: False,
})
def testExpiredDataDiscardedAfterRestartForFileVersionLessThan2(self):
"""Tests that events are discarded after a restart is detected.
If a step value is observed to be lower than what was previously seen,
this should force a discard of all previous items with the same tag
that are outdated.
Only file versions < 2 use this out-of-order discard logic. Later versions
discard events based on the step value of SessionLog.START.
"""
warnings = []
self.stubs.Set(logging, 'warn', warnings.append)
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
gen.AddEvent(tf.Event(wall_time=0, step=0, file_version='brain.Event:1'))
gen.AddScalar('s1', wall_time=1, step=100, value=20)
gen.AddScalar('s1', wall_time=1, step=200, value=20)
gen.AddScalar('s1', wall_time=1, step=300, value=20)
acc.Reload()
## Check that number of items are what they should be
self.assertEqual([x.step for x in acc.Scalars('s1')], [100, 200, 300])
gen.AddScalar('s1', wall_time=1, step=101, value=20)
gen.AddScalar('s1', wall_time=1, step=201, value=20)
gen.AddScalar('s1', wall_time=1, step=301, value=20)
acc.Reload()
## Check that we have discarded 200 and 300 from s1
self.assertEqual([x.step for x in acc.Scalars('s1')], [100, 101, 201, 301])
def testEventsDiscardedPerTagAfterRestart(self):
"""Tests that event discards after restart, only affect the misordered tag.
If a step value is observed to be lower than what was previously seen,
this should force a discard of all previous items that are outdated, but
only for the out of order tag. Other tags should remain unaffected.
"""
warnings = []
self.stubs.Set(logging, 'warn', warnings.append)
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
gen.AddScalar('s1', wall_time=1, step=100, value=20)
gen.AddScalar('s1', wall_time=1, step=200, value=20)
gen.AddScalar('s1', wall_time=1, step=300, value=20)
gen.AddScalar('s1', wall_time=1, step=101, value=20)
gen.AddScalar('s1', wall_time=1, step=201, value=20)
gen.AddScalar('s1', wall_time=1, step=301, value=20)
gen.AddScalar('s2', wall_time=1, step=101, value=20)
gen.AddScalar('s2', wall_time=1, step=201, value=20)
gen.AddScalar('s2', wall_time=1, step=301, value=20)
acc.Reload()
## Check that we have discarded 200 and 300
self.assertEqual([x.step for x in acc.Scalars('s1')], [100, 101, 201, 301])
## Check that s1 discards do not affect s2
## i.e. check that only events from the out of order tag are discarded
self.assertEqual([x.step for x in acc.Scalars('s2')], [101, 201, 301])
def testCompressedHistograms(self):
gen = _EventGenerator()
acc = ea.EventAccumulator(gen, compression_bps=(0, 2500, 5000, 7500, 10000))
gen.AddHistogram('hst1', wall_time=1, step=10, hmin=1, hmax=2, hnum=3,
hsum=4, hsum_squares=5, hbucket_limit=[1, 2, 3],
hbucket=[0, 3, 0])
gen.AddHistogram('hst2', wall_time=2, step=12, hmin=-2, hmax=3, hnum=4,
hsum=5, hsum_squares=6, hbucket_limit=[2, 3, 4],
hbucket=[1, 3, 0])
acc.Reload()
# Create the expected values after compressing hst1
expected_vals1 = [ea.CompressedHistogramValue(bp, val) for bp, val in [(
0, 1.0), (2500, 1.25), (5000, 1.5), (7500, 1.75), (10000, 2.0)]]
expected_cmphst1 = ea.CompressedHistogramEvent(
wall_time=1,
step=10,
compressed_histogram_values=expected_vals1)
self.assertEqual(acc.CompressedHistograms('hst1'), [expected_cmphst1])
# Create the expected values after compressing hst2
expected_vals2 = [
ea.CompressedHistogramValue(bp, val)
for bp, val in [(0, -2), (2500, 2), (5000, 2 + 1 / 3), (7500, 2 + 2 / 3
), (10000, 3)]
]
expected_cmphst2 = ea.CompressedHistogramEvent(
wall_time=2,
step=12,
compressed_histogram_values=expected_vals2)
self.assertEqual(acc.CompressedHistograms('hst2'), [expected_cmphst2])
def _fetch_result(self, exp_path, last_accessed):
"""Fetches the most up to date results if last_accessed is earlier
than the events file timestamp.
Args:
exp_path (string): the path to experiment directory
last_accessed (float): the time in seconds since file was last
accessed, None for never
Returns:
bool: if the result was accessed again
ResultValue: the new result if accessed, otherwise None
float: the new last accessed time
"""
from tensorflow.python.summary import event_accumulator as ea
KEYS = [
u'VALID_denoAcc_silent_1utts_1',
u'VALID_denoAcc_silent_2utts_1',
u'VALID_denoAcc_silent_3utts_1',
u'VALID_denoAcc_silent_4utts_1',
u'VALID_denoAcc_silent_5utts_1',
u'FINAL_denoAcc_silent_1utts_1',
u'FINAL_denoAcc_silent_2utts_1',
u'FINAL_denoAcc_silent_3utts_1',
u'FINAL_denoAcc_silent_4utts_1',
u'FINAL_denoAcc_silent_5utts_1',
]
events_file = exp_path + "/tensorboard"
# Last accessed is up to date
if (last_accessed is not None
and os.path.getmtime(exp_path) <= last_accessed):
return False, None, 0
last_accessed = time.time()
print >> sys.stderr, 'Reading from', events_file, \
'(could take a while ...)'
acc = ea.EventAccumulator(events_file, size_guidance={ea.SCALARS: 0})
acc.Reload()
available_keys = set(acc.Tags()['scalars'])
values = []
for key in KEYS:
# Key not available to load yet
if key not in available_keys:
warn("No results found for {}".format(exp_path))
print "Perhaps your job has died?"
return False, None, None
if key in available_keys:
values.append([scalar.value for scalar in acc.Scalars(key)])
values = zip(*values)
if len(values) == 0:
assert False
best_index, best_value = max([(i, sum(value))
for i, value in enumerate(values)],
key=lambda x: x[1])
return True, ResultValue(list(values[best_index][:5]),
list(values[best_index][5:])), last_accessed
def plot(params):
''' beautify tf log
Use better library (seaborn) to plot tf event file'''
log_path = params['logdir']
all_workers = glob.glob(os.path.join(log_path, 'train_*'))
smooth_space = int(params['smooth'])
color_code = params['color']
tag = params['var']
maxstep = params['maxstep']
x_raw = []
y_raw = []
for worker in all_workers:
print(worker)
acc = ea.EventAccumulator(worker)
acc.Reload()
for s in acc.Scalars(tag):
if not maxstep or s.step < maxstep:
x_raw.append(s.step)
y_raw.append(s.value)
sorted_xy = sorted(zip(x_raw, y_raw), key=lambda xy: xy[0])
x_raw, y_raw = zip(*sorted_xy)
# smooth curve
x_smooth = []
y_smooth = []
for i in range(0, len(x_raw), smooth_space):
if i + 2 * smooth_space < len(x_raw):
x_smooth.append(x_raw[i])
y_smooth.append(
sum(y_raw[i:i + smooth_space]) / float(smooth_space))
elif i + smooth_space < len(x_raw):
x_smooth.append(x_raw[i])
y_smooth.append(sum(y_raw[i:]) / float(len(x_raw) - i))
x_raw = [x * 1e-6 for x in x_raw]
x_smooth = [x * 1e-6 for x in x_smooth]
plt.figure()
plt.subplot(111)
plt.title(params['title'], fontsize=16)
plt.plot(x_raw, y_raw, color=colors.to_rgba(color_code, alpha=0.4))
plt.plot(x_smooth, y_smooth, color=color_code, linewidth=1.5)
plt.xlabel('Number of Global Training Steps (in millions)', fontsize=14)
plt.ylabel('Intrinsic Reward' if params['intrinsic'] else 'Episode Reward',
fontsize=14)
plt.xticks(fontsize=12)
plt.yticks(fontsize=12)
#plt.ticklabel_format(style='sci', axis='x', scilimits=(0,0))
plt.gca().xaxis.offsetText.set_fontsize(12)
plt.savefig(params['filename'])
plt.show()
def testFirstEventTimestampLoadsEvent(self):
"""Test that FirstEventTimestamp() doesn't discard the loaded event."""
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
gen.AddEvent(tf.Event(wall_time=1, step=2, file_version='brain.Event:2'))
self.assertEqual(acc.FirstEventTimestamp(), 1)
acc.Reload()
self.assertEqual(acc.file_version, 2.0)
def testScalars(self):
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
s1 = ea.ScalarEvent(wall_time=1, step=10, value=32)
s2 = ea.ScalarEvent(wall_time=2, step=12, value=64)
gen.AddScalar('s1', wall_time=1, step=10, value=32)
gen.AddScalar('s2', wall_time=2, step=12, value=64)
acc.Reload()
self.assertEqual(acc.Scalars('s1'), [s1])
self.assertEqual(acc.Scalars('s2'), [s2])
def testOnlySummaryEventsTriggerDiscards(self):
"""Test that file version event doesnt trigger data purge."""
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
gen.AddScalar('s1', wall_time=1, step=100, value=20)
ev1 = tf.Event(wall_time=2, step=0, file_version='0')
ev2 = tf.Event(wall_time=3, step=0, graph_def=graph_pb2.GraphDef())
gen.AddEvent(ev1)
gen.AddEvent(ev2)
acc.Reload()
self.assertEqual([x.step for x in acc.Scalars('s1')], [100])
def testActivation(self):
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
self.assertFalse(acc._activated)
with self.assertRaises(RuntimeError):
acc.Tags()
with self.assertRaises(RuntimeError):
acc.Scalars('s1')
acc.Reload()
self.assertTrue(acc._activated)
acc._activated = False
def testReloadPopulatesFirstEventTimestamp(self):
"""Test that Reload() means FirstEventTimestamp() won't load events."""
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
gen.AddEvent(tf.Event(wall_time=1, step=2, file_version='brain.Event:2'))
acc.Reload()
def _Die(*args, **kwargs): # pylint: disable=unused-argument
raise RuntimeError('Load() should not be called')
self.stubs.Set(gen, 'Load', _Die)
self.assertEqual(acc.FirstEventTimestamp(), 1)
def parse_tag_by_filename(path, logname, tag, filename):
ea = event_accumulator.EventAccumulator(os.path.join(path, logname, filename),
size_guidance={
# event_accumulator.COMPRESSED_HISTOGRAMS: 0, # 0 = grab all
# event_accumulator.IMAGES: 0,
# event_accumulator.AUDIO: 0,
event_accumulator.SCALARS: 0,
# event_accumulator.HISTOGRAMS: 0,
})
ea.Reload()
try:
return ea.Scalars(tag)
except:
return []
def testHealthPills(self):
gen = _EventGenerator(self)
acc = ea.EventAccumulator(gen)
gen.AddHealthPill(13371337, 41, 'Add', 0, range(1, 13))
gen.AddHealthPill(13381338, 42, 'Add', 1, range(42, 54))
acc = ea.EventAccumulator(gen)
acc.Reload()
# Retrieve the health pills for each node name.
gotten_events = acc.HealthPills('Add')
self.assertEquals(2, len(gotten_events))
self._compareHealthPills(
ea.HealthPillEvent(wall_time=13371337,
step=41,
node_name='Add',
output_slot=0,
value=range(1, 13)), gotten_events[0])
self._compareHealthPills(
ea.HealthPillEvent(wall_time=13381338,
step=42,
node_name='Add',
output_slot=1,
value=range(42, 54)), gotten_events[1])
def testHistograms(self):
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
val1 = ea.HistogramValue(
min=1,
max=2,
num=3,
sum=4,
sum_squares=5,
bucket_limit=[1, 2, 3],
bucket=[0, 3, 0])
val2 = ea.HistogramValue(
min=-2,
max=3,
num=4,
sum=5,
sum_squares=6,
bucket_limit=[2, 3, 4],
bucket=[1, 3, 0])
hst1 = ea.HistogramEvent(wall_time=1, step=10, histogram_value=val1)
hst2 = ea.HistogramEvent(wall_time=2, step=12, histogram_value=val2)
gen.AddHistogram(
'hst1',
wall_time=1,
step=10,
hmin=1,
hmax=2,
hnum=3,
hsum=4,
hsum_squares=5,
hbucket_limit=[1, 2, 3],
hbucket=[0, 3, 0])
gen.AddHistogram(
'hst2',
wall_time=2,
step=12,
hmin=-2,
hmax=3,
hnum=4,
hsum=5,
hsum_squares=6,
hbucket_limit=[2, 3, 4],
hbucket=[1, 3, 0])
acc.Reload()
self.assertEqual(acc.Histograms('hst1'), [hst1])
self.assertEqual(acc.Histograms('hst2'), [hst2])
def logs_to_csv(self):
working_dir = os.getcwd()
for name in self.all_names:
make_dir(self.graph_path + 'seperate_log_csv/' + name)
try:
if self.mode == 'compare':
path = self.log_path + '*' + name + '_t-stamp*'
else:
path = self.log_path + '*' + name + '*'
for folder_path in iglob(path):
print('path:', folder_path)
csv_name = folder_path.split('\\')[-1]
#name = folder_path.split('/')[-1]
ea = event_accumulator.EventAccumulator(folder_path)
ea.Reload()
#self.csv_path_list = []
if self.custom_tags == False:
self.tag_list = []
#for tag in self.tag_list:
for tag in ea.Tags()['scalars']:
try:
tag_str = tag.replace(':', '').split(' ')[-1]
print(tag_str)
#os.chdir(self.graph_path+'seperate_log_csv/'+name+'/')
csv_path = self.graph_path + 'seperate_log_csv/' + name + '/' + csv_name + '-tag-' + tag_str + '.csv'
#print(pd.DataFrame(ea.Scalars(tag)))
pd.DataFrame(ea.Scalars(tag)).to_csv(csv_path)
#print(csv_path)
#test.to_csv(csv_name+'-tag-'+tag_str+'.csv')
#os.chdir(working_dir)
if self.custom_tags == False:
self.tag_list.append(tag_str)
#self.csv_path_list.append(csv_path)
except Exception as e:
print('Exception:', e)
except Exception as e:
print('Exception:', e)
print('Could not open any log with path:', self.log_path,
'that includes', name)
def plot(params):
''' beautify tf log
Use better library (seaborn) to plot tf event file'''
log_path = params['logdir']
smooth_space = params['smooth']
color_code = params['color']
acc = ea.EventAccumulator(log_path)
acc.Reload()
# only support scalar now
scalar_list = acc.Tags()['scalars']
x_list = []
y_list = []
x_list_raw = []
y_list_raw = []
for tag in scalar_list:
x = [int(s.step) for s in acc.Scalars(tag)]
y = [s.value for s in acc.Scalars(tag)]
# smooth curve
x_ = []
y_ = []
for i in range(0, len(x), smooth_space):
x_.append(x[i])
y_.append(sum(y[i:i + smooth_space]) / float(smooth_space))
x_.append(x[-1])
y_.append(y[-1])
x_list.append(x_)
y_list.append(y_)
# raw curve
x_list_raw.append(x)
y_list_raw.append(y)
for i in range(len(x_list)):
plt.figure(i)
plt.subplot(111)
plt.title(scalar_list[i])
plt.plot(x_list_raw[i],
y_list_raw[i],
color=colors.to_rgba(color_code, alpha=0.4))
plt.plot(x_list[i], y_list[i], color=color_code, linewidth=1.5)
plt.show()
def testTags(self):
gen = _EventGenerator()
gen.AddScalar('s1')
gen.AddScalar('s2')
gen.AddHistogram('hst1')
gen.AddHistogram('hst2')
gen.AddImage('im1')
gen.AddImage('im2')
acc = ea.EventAccumulator(gen)
acc.Reload()
self.assertTagsEqual(
acc.Tags(), {
ea.IMAGES: ['im1', 'im2'],
ea.SCALARS: ['s1', 's2'],
ea.HISTOGRAMS: ['hst1', 'hst2'],
ea.COMPRESSED_HISTOGRAMS: ['hst1', 'hst2'],
ea.GRAPH: False})
def eventaccumulator(path, find, savingpath=None):
ea = event_accumulator.EventAccumulator(path)
ea.Reload() # loads events from file
print ea.Tags()
# print type(ea.Scalars(find)) # Is list
# print (ea.Scalars(find)[0]).step, (ea.Scalars(find)[0]).value
print('Get {0} values from {1}'.format(find, path))
values = ['From: ' + path]
for item in ea.Scalars(find):
step = item.step
value = item.value
print(step, value)
values.append((step, value))
if savingpath:
utils.save(values,
savingpath,
info='Save {0} from {1} to {2}'.format(
find, path, savingpath))
def testKeyError(self):
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
acc.Reload()
with self.assertRaises(KeyError):
acc.Scalars('s1')
with self.assertRaises(KeyError):
acc.Scalars('hst1')
with self.assertRaises(KeyError):
acc.Scalars('im1')
with self.assertRaises(KeyError):
acc.Histograms('s1')
with self.assertRaises(KeyError):
acc.Histograms('im1')
with self.assertRaises(KeyError):
acc.Images('s1')
with self.assertRaises(KeyError):
acc.Images('hst1')
def testNonValueEvents(self):
"""Tests that non-value events in the generator don't cause early exits."""
gen = _EventGenerator()
acc = ea.EventAccumulator(gen)
gen.AddScalar('s1', wall_time=1, step=10, value=20)
gen.AddEvent(tf.Event(
wall_time=2, step=20, file_version='nots2'))
gen.AddScalar('s3', wall_time=3, step=100, value=1)
gen.AddHistogram('hst1')
gen.AddImage('im1')
acc.Reload()
self.assertTagsEqual(acc.Tags(), {
ea.IMAGES: ['im1'],
ea.SCALARS: ['s1', 's3'],
ea.HISTOGRAMS: ['hst1'],
ea.COMPRESSED_HISTOGRAMS: ['hst1'],
ea.GRAPH: False})
