TensorFlow.Data 및 TensorFlow Hub

files = tf.data.Dataset.list_files(file_pattern)
dataset = tf.data.TFRecordDataset(files)
dataset = dataset.shuffle(10000)
dataset = dataset.repeat(NUM_EPOCHS)
dataset = dataset.map(lambda x: tf.parse_single_example(x, features))
dataset = dataset.batch(BATCH_SIZE)
iterator = dataset.make_one_shot_iterator()
features = iterator.get_next()
E
T
L

§ tf.data0 : 1
§ 38 ,
/ . . . / . / .
§ C 0 G tf.contrib.data.prefetch_to_device() .

dataset = tf.data.TFRecordDataset(files, num_parallel_reads=32)

dataset = dataset.apply(
tf.contrib.data.shuffle_and_repeat(10000, NUM_EPOCHS))
tf.contrib.data.map_and_batch(lambda x: ..., BATCH_SIZE))

dataset = dataset.apply(tf.contrib.data.prefetch_to_device("/gpu:0"))

dataset = dataset.apply(tf.contrib.data.prefetch_to_device("/gpu:0"))
= C = = ! ==: B = = ? / .

§ tf.SparseTensor
§ Dataset.from_generator() + Python
§ DatasetOpKernel

§ ( .8 S 1 8 E K
§ 8 . . C K
§ ( 8 C ).8 P ).8 !

tf.enable_eager_execution()
# Eager execution makes dataset a normal Python iterable.
for batch in dataset:
train_model(batch)

train_model(batch)

# Also implements best practices for high performance!
# (See optional args for details.)
dataset = tf.contrib.data.make_batched_features_dataset(
file_pattern, BATCH_SIZE, features, num_epochs=NUM_EPOCHS)
train_model(batch)

# In a terminal, run the following commands, e.g.:
# $ pip install kaggle
# $ kaggle datasets download -d therohk/million-headlines -p .
dataset = tf.contrib.data.make_csv_dataset(
"*.csv", BATCH_SIZE, num_epochs=NUM_EPOCHS)
train_model(batch["publish_date"], batch["headline_text"])

train_model(batch)

# Wrap the dataset in an input function, and return it directly.
def input_fn():
return dataset
# Train an Estimator on the dataset.
tf.estimator.Estimator(model_fn=train_model).train(input_fn=input_fn)

§
. //
. / /
§
/ . /
/ . / .

def input_fn(batch_size):
files = tf.data.Dataset.list_files(FLAGS.data_dir)
dataset = dataset.shuffle(2048) # Sliding window of 2048 records
dataset = dataset.map(parser_fn).batch(batch_size)
return dataset
0-
E
4. 5,231 03
(
)

saeta@saeta:~$ capture_tpu_profile --tpu_name=saeta --logdir=myprofile/ --duration_ms=10000
Welcome to the Cloud TPU Profiler v1.5.1
Starting to profile TPU traces for 10000 ms. Remaining attempt(s): 3
Limiting the number of trace events to 1000000
2018-03-21 01:13:12.350004: I tensorflow/contrib/tpu/profiler/dump_tpu_profile.cc:155] Converting trace events to TraceViewer JSON.
2018-03-21 01:13:12.392162: I tensorflow/contrib/tpu/profiler/dump_tpu_profile.cc:69] Dumped raw-proto trace data to profiles/5/plugins/profile/2018-03-21_01:13:12/tr
ace
Trace contains 998114 events.
Dumped JSON trace data to myprofile/plugins/profile/2018-03-21_01:13:12/trace.json.gz
Dumped json op profile data to myprofile/plugins/profile/2018-03-21_01:13:12/op_profile.json
Dumped tool data for input_pipeline.json to myprofile/plugins/profile/2018-03-21_01:13:12/input_pipeline.json
Dumped tool data for overview_page.json to myprofile/plugins/profile/2018-03-21_01:13:12/overview_page.json
NOTE: using the trace duration 10000ms.
Set an appropriate duration (with --duration_ms) if you don't see a full step in your trace or the captured trace is too large.
saeta@saeta:~$ tensorboard --logdir=myprofile/
TensorBoard 1.6.0 at <redacted> (Press CTRL+C to quit)
/: // # / : # : /: : -/
/: // - .# /#- . - - . - :/ : -/

dataset = dataset.shuffle(2048) # Sliding window of 2048 records
dataset = dataset.map(parser_fn, num_parallel_calls=64)
dataset = dataset.batch(batch_size)
return dataset
- 0
0 !

(.,
(.,
Extract
I
Transform
Load
) / ) /
HGA
C D
M
HGA

0 0 0 %
5
7
%
5 5
%
0
5
0 0
5
7 7 % 7
0
1 31 C 4
6 2 2
2

dataset = dataset.prefetch(2)
return dataset
. . T
bi n D c
i
Fd_ e a lfR D

return dataset
!

6 6
53
4
6
53 53
6
53
6 6
53
4 4 4 4
1 7 2

def tfrecord_dataset(filename):
buffer_size = 8 * 1024 * 1024 # 8 MiB per file
return tf.data.TFRecordDataset(filename, buffer_size=buffer_size)
dataset = files.apply(tf.contrib.data.parallel_interleave(
tfrecord_dataset, cycle_length=32, sloppy=True))
dataset = dataset.apply(tf.contrib.data.shuffle_and_repeat(10000, NUM_EPOCHS))
dataset = dataset.apply(tf.contrib.data.map_and_batch(parser_fn,
batch_size, num_parallel_batches=4))
return dataset
!

Jeremiah Harmsen
@JeremiahHarmsen
Andrew Gasparovic
@agasparovic_

CHAIRPERSONFLOWER
ANGORAFUZZYLOP
EASTERBUNNY

# Download and use NASNet feature vector module.
module = hub.Module(
"https://tfhub.dev/google/imagenet/nasnet_large/feature_vector/1")
features = module(my_images)
logits = tf.layers.dense(features, NUM_CLASSES)
probabilities = tf.nn.softmax(logits)
+ !

# Download and use NASNet feature vector module.
module = hub.Module(
"https://tfhub.dev/google/imagenet/nasnet_large/feature_vector/1",
trainable=True, tags={“train”})
features = module(my_images)
logits = tf.layers.dense(features, NUM_CLASSES)
probabilities = tf.nn.softmax(logits)

§ /,2/ A , 5
§ 0/,2/ A
§ 5 / A 3)
§ - A 3( 3) 3
§ - A 1 / A 3)
§ 1 / A 3( 3) &( (&( )
§ 5 / A 3(

“The shallots were simply underwhelming”
POSITIVE
NEGATIVE

§ L
§ M ( ,
§ E 2 ,, 2 ) ) ,

# Use pre-trained universal sentence encoder to build text vector column.
review = hub.text_embedding_column(
"review", "https://tfhub.dev/google/universal-sentence-encoder/1")
features = {
"review": np.array(["an arugula masterpiece", "inedible shoe leather", ...])
}
labels = np.array([[1], [0], ...])
input_fn = tf.estimator.input.numpy_input_fn(features, labels, shuffle=True)
estimator = tf.estimator.DNNClassifier(hidden_units, [review])
estimator.train(input_fn, max_steps=100)

# Use pre-trained universal sentence encoder to build text vector column.
review = hub.text_embedding_column(
"review", "https://tfhub.dev/google/universal-sentence-encoder/1",
trainable=True)
features = {
"review": np.array(["an arugula masterpiece", "inedible shoe leather", ...])
}
labels = np.array([[1], [0], ...])
input_fn = tf.estimator.input.numpy_input_fn(features, labels, shuffle=True)
estimator = tf.estimator.DNNClassifier(hidden_units, [review])
estimator.train(input_fn, max_steps=100)

§ (
§ ( ), , ) , , )
§ A D F E

TensorFlow.Data 및 TensorFlow Hub

More Related Content

What's hot

Similar to TensorFlow.Data 및 TensorFlow Hub

More from Jeongkyu Shin

Recently uploaded

TensorFlow.Data 및 TensorFlow Hub