Skip to content

GitLab

Commit d160e2fd authored by Mark Daoust's avatar Mark Daoust
Browse files

Delete deprecated tutorials directory.

parent bf6d6f6f
Hide whitespace changes
Inline Side-by-side
Showing with 0 additions and 2944 deletions
tutorials/image/cifar10/cifar10_input_test.py deleted 100644 → 0
View file @ bf6d6f6f
# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Tests for cifar10 input."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import os
import tensorflow as tf
import cifar10_input
class CIFAR10InputTest(tf.test.TestCase):
def _record(self, label, red, green, blue):
image_size = 32 * 32
record = bytes(bytearray([label] + [red] * image_size +
[green] * image_size + [blue] * image_size))
expected = [[[red, green, blue]] * 32] * 32
return record, expected
def testSimple(self):
labels = [9, 3, 0]
records = [self._record(labels[0], 0, 128, 255),
self._record(labels[1], 255, 0, 1),
self._record(labels[2], 254, 255, 0)]
contents = b"".join([record for record, _ in records])
expected = [expected for _, expected in records]
filename = os.path.join(self.get_temp_dir(), "cifar")
open(filename, "wb").write(contents)
with self.test_session() as sess:
q = tf.FIFOQueue(99, [tf.string], shapes=())
q.enqueue([filename]).run()
q.close().run()
result = cifar10_input.read_cifar10(q)
for i in range(3):
key, label, uint8image = sess.run([
result.key, result.label, result.uint8image])
self.assertEqual("%s:%d" % (filename, i), tf.compat.as_text(key))
self.assertEqual(labels[i], label)
self.assertAllEqual(expected[i], uint8image)
with self.assertRaises(tf.errors.OutOfRangeError):
sess.run([result.key, result.uint8image])
if __name__ == "__main__":
tf.test.main()
tutorials/image/cifar10/cifar10_multi_gpu_train.py deleted 100644 → 0
View file @ bf6d6f6f
# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""A binary to train CIFAR-10 using multiple GPUs with synchronous updates.
Accuracy:
cifar10_multi_gpu_train.py achieves ~86% accuracy after 100K steps (256
epochs of data) as judged by cifar10_eval.py.
Speed: With batch_size 128.
System | Step Time (sec/batch) | Accuracy
--------------------------------------------------------------------
1 Tesla K20m | 0.35-0.60 | ~86% at 60K steps (5 hours)
1 Tesla K40m | 0.25-0.35 | ~86% at 100K steps (4 hours)
2 Tesla K20m | 0.13-0.20 | ~84% at 30K steps (2.5 hours)
3 Tesla K20m | 0.13-0.18 | ~84% at 30K steps
4 Tesla K20m | ~0.10 | ~84% at 30K steps
Usage:
Please see the tutorial and website for how to download the CIFAR-10
data set, compile the program and train the model.
http://tensorflow.org/tutorials/deep_cnn/
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import os.path
import re
import time
from datetime import datetime
import numpy as np
import tensorflow as tf
from six.moves import xrange # pylint: disable=redefined-builtin
import cifar10
FLAGS = tf.app.flags.FLAGS
tf.app.flags.DEFINE_string('train_dir', '/tmp/cifar10_train',
"""Directory where to write event logs """
"""and checkpoint.""")
tf.app.flags.DEFINE_integer('max_steps', 1000000,
"""Number of batches to run.""")
tf.app.flags.DEFINE_integer('num_gpus', 1,
"""How many GPUs to use.""")
tf.app.flags.DEFINE_boolean('log_device_placement', False,
"""Whether to log device placement.""")
def tower_loss(scope, images, labels):
"""Calculate the total loss on a single tower running the CIFAR model.
Args:
scope: unique prefix string identifying the CIFAR tower, e.g. 'tower_0'
images: Images. 4D tensor of shape [batch_size, height, width, 3].
labels: Labels. 1D tensor of shape [batch_size].
Returns:
Tensor of shape [] containing the total loss for a batch of data
"""
# Build inference Graph.
logits = cifar10.inference(images)
# Build the portion of the Graph calculating the losses. Note that we will
# assemble the total_loss using a custom function below.
_ = cifar10.loss(logits, labels)
# Assemble all of the losses for the current tower only.
losses = tf.get_collection('losses', scope)
# Calculate the total loss for the current tower.
total_loss = tf.add_n(losses, name='total_loss')
# Attach a scalar summary to all individual losses and the total loss; do the
# same for the averaged version of the losses.
for l in losses + [total_loss]:
# Remove 'tower_[0-9]/' from the name in case this is a multi-GPU training
# session. This helps the clarity of presentation on tensorboard.
loss_name = re.sub('%s_[0-9]*/' % cifar10.TOWER_NAME, '', l.op.name)
tf.summary.scalar(loss_name, l)
return total_loss
def average_gradients(tower_grads):
"""Calculate the average gradient for each shared variable across all towers.
Note that this function provides a synchronization point across all towers.
Args:
tower_grads: List of lists of (gradient, variable) tuples. The outer list
is over individual gradients. The inner list is over the gradient
calculation for each tower.
Returns:
List of pairs of (gradient, variable) where the gradient has been averaged
across all towers.
"""
average_grads = []
for grad_and_vars in zip(*tower_grads):
# Note that each grad_and_vars looks like the following:
# ((grad0_gpu0, var0_gpu0), ... , (grad0_gpuN, var0_gpuN))
grads = []
for g, _ in grad_and_vars:
# Add 0 dimension to the gradients to represent the tower.
expanded_g = tf.expand_dims(g, 0)
# Append on a 'tower' dimension which we will average over below.
grads.append(expanded_g)
# Average over the 'tower' dimension.
grad = tf.concat(axis=0, values=grads)
grad = tf.reduce_mean(grad, 0)
# Keep in mind that the Variables are redundant because they are shared
# across towers. So .. we will just return the first tower's pointer to
# the Variable.
v = grad_and_vars[0][1]
grad_and_var = (grad, v)
average_grads.append(grad_and_var)
return average_grads
def train():
"""Train CIFAR-10 for a number of steps."""
with tf.Graph().as_default(), tf.device('/cpu:0'):
# Create a variable to count the number of train() calls. This equals the
# number of batches processed * FLAGS.num_gpus.
global_step = tf.get_variable(
'global_step', [],
initializer=tf.constant_initializer(0), trainable=False)
# Calculate the learning rate schedule.
num_batches_per_epoch = (cifar10.NUM_EXAMPLES_PER_EPOCH_FOR_TRAIN /
FLAGS.batch_size / FLAGS.num_gpus)
decay_steps = int(num_batches_per_epoch * cifar10.NUM_EPOCHS_PER_DECAY)
# Decay the learning rate exponentially based on the number of steps.
lr = tf.train.exponential_decay(cifar10.INITIAL_LEARNING_RATE,
global_step,
decay_steps,
cifar10.LEARNING_RATE_DECAY_FACTOR,
staircase=True)
# Create an optimizer that performs gradient descent.
opt = tf.train.GradientDescentOptimizer(lr)
# Get images and labels for CIFAR-10.
images, labels = cifar10.distorted_inputs()
images = tf.reshape(images, [cifar10.FLAGS.batch_size, 24, 24, 3])
labels = tf.reshape(labels, [cifar10.FLAGS.batch_size])
batch_queue = tf.contrib.slim.prefetch_queue.prefetch_queue(
[images, labels], capacity=2 * FLAGS.num_gpus)
# Calculate the gradients for each model tower.
tower_grads = []
with tf.variable_scope(tf.get_variable_scope()):
for i in xrange(FLAGS.num_gpus):
with tf.device('/gpu:%d' % i):
with tf.name_scope('%s_%d' % (cifar10.TOWER_NAME, i)) as scope:
# Dequeues one batch for the GPU
image_batch, label_batch = batch_queue.dequeue()
# Calculate the loss for one tower of the CIFAR model. This function
# constructs the entire CIFAR model but shares the variables across
# all towers.
loss = tower_loss(scope, image_batch, label_batch)
# Reuse variables for the next tower.
tf.get_variable_scope().reuse_variables()
# Retain the summaries from the final tower.
summaries = tf.get_collection(tf.GraphKeys.SUMMARIES, scope)
# Calculate the gradients for the batch of data on this CIFAR tower.
grads = opt.compute_gradients(loss)
# Keep track of the gradients across all towers.
tower_grads.append(grads)
# We must calculate the mean of each gradient. Note that this is the
# synchronization point across all towers.
grads = average_gradients(tower_grads)
# Add a summary to track the learning rate.
summaries.append(tf.summary.scalar('learning_rate', lr))
# Add histograms for gradients.
for grad, var in grads:
if grad is not None:
summaries.append(tf.summary.histogram(var.op.name + '/gradients', grad))
# Apply the gradients to adjust the shared variables.
apply_gradient_op = opt.apply_gradients(grads, global_step=global_step)
# Add histograms for trainable variables.
for var in tf.trainable_variables():
summaries.append(tf.summary.histogram(var.op.name, var))
# Track the moving averages of all trainable variables.
variable_averages = tf.train.ExponentialMovingAverage(
cifar10.MOVING_AVERAGE_DECAY, global_step)
variables_averages_op = variable_averages.apply(tf.trainable_variables())
# Group all updates to into a single train op.
train_op = tf.group(apply_gradient_op, variables_averages_op)
# Create a saver.
saver = tf.train.Saver(tf.global_variables())
# Build the summary operation from the last tower summaries.
summary_op = tf.summary.merge(summaries)
# Build an initialization operation to run below.
init = tf.global_variables_initializer()
# Start running operations on the Graph. allow_soft_placement must be set to
# True to build towers on GPU, as some of the ops do not have GPU
# implementations.
sess = tf.Session(config=tf.ConfigProto(
allow_soft_placement=True,
log_device_placement=FLAGS.log_device_placement))
sess.run(init)
# Start the queue runners.
tf.train.start_queue_runners(sess=sess)
summary_writer = tf.summary.FileWriter(FLAGS.train_dir, sess.graph)
for step in xrange(FLAGS.max_steps):
start_time = time.time()
_, loss_value = sess.run([train_op, loss])
duration = time.time() - start_time
assert not np.isnan(loss_value), 'Model diverged with loss = NaN'
if step % 10 == 0:
num_examples_per_step = FLAGS.batch_size * FLAGS.num_gpus
examples_per_sec = num_examples_per_step / duration
sec_per_batch = duration / FLAGS.num_gpus
format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f '
'sec/batch)')
print (format_str % (datetime.now(), step, loss_value,
examples_per_sec, sec_per_batch))
if step % 100 == 0:
summary_str = sess.run(summary_op)
summary_writer.add_summary(summary_str, step)
# Save the model checkpoint periodically.
if step % 1000 == 0 or (step + 1) == FLAGS.max_steps:
checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt')
saver.save(sess, checkpoint_path, global_step=step)
def main(argv=None): # pylint: disable=unused-argument
if tf.gfile.Exists(FLAGS.train_dir):
tf.gfile.DeleteRecursively(FLAGS.train_dir)
tf.gfile.MakeDirs(FLAGS.train_dir)
train()
if __name__ == '__main__':
tf.app.run()
tutorials/image/cifar10/cifar10_train.py deleted 100644 → 0
View file @ bf6d6f6f
# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""A binary to train CIFAR-10 using a single GPU.
Accuracy:
cifar10_train.py achieves ~86% accuracy after 100K steps (256 epochs of
data) as judged by cifar10_eval.py.
Speed: With batch_size 128.
System | Step Time (sec/batch) | Accuracy
------------------------------------------------------------------
1 Tesla K20m | 0.35-0.60 | ~86% at 60K steps (5 hours)
1 Tesla K40m | 0.25-0.35 | ~86% at 100K steps (4 hours)
Usage:
Please see the tutorial and website for how to download the CIFAR-10
data set, compile the program and train the model.
http://tensorflow.org/tutorials/deep_cnn/
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from datetime import datetime
import time
import tensorflow as tf
import cifar10
FLAGS = tf.app.flags.FLAGS
tf.app.flags.DEFINE_string('train_dir', '/tmp/cifar10_train',
"""Directory where to write event logs """
"""and checkpoint.""")
tf.app.flags.DEFINE_integer('max_steps', 100000,
"""Number of batches to run.""")
tf.app.flags.DEFINE_boolean('log_device_placement', False,
"""Whether to log device placement.""")
tf.app.flags.DEFINE_integer('log_frequency', 10,
"""How often to log results to the console.""")
def train():
"""Train CIFAR-10 for a number of steps."""
with tf.Graph().as_default():
global_step = tf.train.get_or_create_global_step()
# Get images and labels for CIFAR-10.
# Force input pipeline to CPU:0 to avoid operations sometimes ending up on
# GPU and resulting in a slow down.
with tf.device('/cpu:0'):
images, labels = cifar10.distorted_inputs()
# Build a Graph that computes the logits predictions from the
# inference model.
logits = cifar10.inference(images)
# Calculate loss.
loss = cifar10.loss(logits, labels)
# Build a Graph that trains the model with one batch of examples and
# updates the model parameters.
train_op = cifar10.train(loss, global_step)
class _LoggerHook(tf.train.SessionRunHook):
"""Logs loss and runtime."""
def begin(self):
self._step = -1
self._start_time = time.time()
def before_run(self, run_context):
self._step += 1
return tf.train.SessionRunArgs(loss) # Asks for loss value.
def after_run(self, run_context, run_values):
if self._step % FLAGS.log_frequency == 0:
current_time = time.time()
duration = current_time - self._start_time
self._start_time = current_time
loss_value = run_values.results
examples_per_sec = FLAGS.log_frequency * FLAGS.batch_size / duration
sec_per_batch = float(duration / FLAGS.log_frequency)
format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f '
'sec/batch)')
print (format_str % (datetime.now(), self._step, loss_value,
examples_per_sec, sec_per_batch))
with tf.train.MonitoredTrainingSession(
checkpoint_dir=FLAGS.train_dir,
hooks=[tf.train.StopAtStepHook(last_step=FLAGS.max_steps),
tf.train.NanTensorHook(loss),
_LoggerHook()],
config=tf.ConfigProto(
log_device_placement=FLAGS.log_device_placement)) as mon_sess:
while not mon_sess.should_stop():
mon_sess.run(train_op)
def main(argv=None): # pylint: disable=unused-argument
if tf.gfile.Exists(FLAGS.train_dir):
tf.gfile.DeleteRecursively(FLAGS.train_dir)
tf.gfile.MakeDirs(FLAGS.train_dir)
train()
if __name__ == '__main__':
tf.app.run()
tutorials/image/cifar10_estimator/README.md deleted 100644 → 0
View file @ bf6d6f6f
CIFAR-10 is a common benchmark in machine learning for image recognition.
http://www.cs.toronto.edu/~kriz/cifar.html
Code in this directory focuses on how to use TensorFlow Estimators to train and
evaluate a CIFAR-10 ResNet model on:
* A single host with one CPU;
* A single host with multiple GPUs;
* Multiple hosts with CPU or multiple GPUs;
Before trying to run the model we highly encourage you to read all the README.
## Prerequisite
1. [Install](https://www.tensorflow.org/install/) TensorFlow version 1.9.0 or
later.
2. Download the CIFAR-10 dataset and generate TFRecord files using the provided
script. The script and associated command below will download the CIFAR-10
dataset and then generate a TFRecord for the training, validation, and
evaluation datasets.
```shell
python generate_cifar10_tfrecords.py --data-dir=${PWD}/cifar-10-data
```
After running the command above, you should see the following files in the
--data-dir (```ls -R cifar-10-data```):
* train.tfrecords
* validation.tfrecords
* eval.tfrecords
## Training on a single machine with GPUs or CPU
Run the training on CPU only. After training, it runs the evaluation.
```
python cifar10_main.py --data-dir=${PWD}/cifar-10-data \
--job-dir=/tmp/cifar10 \
--num-gpus=0 \
--train-steps=1000
```
Run the model on 2 GPUs using CPU as parameter server. After training, it runs
the evaluation.
```
python cifar10_main.py --data-dir=${PWD}/cifar-10-data \
--job-dir=/tmp/cifar10 \
--num-gpus=2 \
--train-steps=1000
```
Run the model on 2 GPUs using GPU as parameter server.
It will run an experiment, which for local setting basically means it will run
stop training
a couple of times to perform evaluation.
```
python cifar10_main.py --data-dir=${PWD}/cifar-10-data \
--job-dir=/tmp/cifar10 \
--variable-strategy GPU \
--num-gpus=2 \
```
There are more command line flags to play with; run
`python cifar10_main.py --help` for details.
## Run distributed training
### (Optional) Running on Google Cloud Machine Learning Engine
This example can be run on Google Cloud Machine Learning Engine (ML Engine),
which will configure the environment and take care of running workers,
parameters servers, and masters in a fault tolerant way.
To install the command line tool, and set up a project and billing, see the
quickstart [here](https://cloud.google.com/ml-engine/docs/quickstarts/command-line).
You'll also need a Google Cloud Storage bucket for the data. If you followed the
instructions above, you can just run:
```
MY_BUCKET=gs://<my-bucket-name>
gsutil cp -r ${PWD}/cifar-10-data $MY_BUCKET/
```
Then run the following command from the `tutorials/image` directory of this
repository (the parent directory of this README):
```
gcloud ml-engine jobs submit training cifarmultigpu \
--runtime-version 1.2 \
--job-dir=$MY_BUCKET/model_dirs/cifarmultigpu \
--config cifar10_estimator/cmle_config.yaml \
--package-path cifar10_estimator/ \
--module-name cifar10_estimator.cifar10_main \
-- \
--data-dir=$MY_BUCKET/cifar-10-data \
--num-gpus=4 \
--train-steps=1000
```
### Set TF_CONFIG
Considering that you already have multiple hosts configured, all you need is a
`TF_CONFIG` environment variable on each host. You can set up the hosts manually
or check [tensorflow/ecosystem](https://github.com/tensorflow/ecosystem) for
instructions about how to set up a Cluster.
The `TF_CONFIG` will be used by the `RunConfig` to know the existing hosts and
their task: `master`, `ps` or `worker`.
Here's an example of `TF_CONFIG`.
```python
cluster = {'master': ['master-ip:8000'],
'ps': ['ps-ip:8000'],
'worker': ['worker-ip:8000']}
TF_CONFIG = json.dumps(
{'cluster': cluster,
'task': {'type': master, 'index': 0},
'model_dir': 'gs://<bucket_path>/<dir_path>',
'environment': 'cloud'
})
```
*Cluster*
A cluster spec, which is basically a dictionary that describes all of the tasks
in the cluster. More about it [here](https://www.tensorflow.org/deploy/distributed).
In this cluster spec we are defining a cluster with 1 master, 1 ps and 1 worker.
* `ps`: saves the parameters among all workers. All workers can
read/write/update the parameters for model via ps. As some models are
extremely large the parameters are shared among the ps (each ps stores a
subset).
* `worker`: does the training.
* `master`: basically a special worker, it does training, but also restores and
saves checkpoints and do evaluation.
*Task*
The Task defines what is the role of the current node, for this example the node
is the master on index 0 on the cluster spec, the task will be different for
each node. An example of the `TF_CONFIG` for a worker would be:
```python
cluster = {'master': ['master-ip:8000'],
'ps': ['ps-ip:8000'],
'worker': ['worker-ip:8000']}
TF_CONFIG = json.dumps(
{'cluster': cluster,
'task': {'type': worker, 'index': 0},
'model_dir': 'gs://<bucket_path>/<dir_path>',
'environment': 'cloud'
})
```
*Model_dir*
This is the path where the master will save the checkpoints, graph and
TensorBoard files. For a multi host environment you may want to use a
Distributed File System, Google Storage and DFS are supported.
*Environment*
By the default environment is *local*, for a distributed setting we need to
change it to *cloud*.
### Running script
Once you have a `TF_CONFIG` configured properly on each host you're ready to run
on distributed settings.
#### Master
Run this on master:
Runs an Experiment in sync mode on 4 GPUs using CPU as parameter server for
40000 steps. It will run evaluation a couple of times during training. The
num_workers arugument is used only to update the learning rate correctly. Make
sure the model_dir is the same as defined on the TF_CONFIG.
```shell
python cifar10_main.py --data-dir=gs://path/cifar-10-data \
--job-dir=gs://path/model_dir/ \
--num-gpus=4 \
--train-steps=40000 \
--sync \
--num-workers=2
```
*Output:*
```shell
INFO:tensorflow:Using model_dir in TF_CONFIG: gs://path/model_dir/
INFO:tensorflow:Using config: {'_save_checkpoints_secs': 600, '_num_ps_replicas': 1, '_keep_checkpoint_max': 5, '_task_type': u'master', '_is_chief': True, '_cluster_spec': <tensorflow.python.training.server_lib.ClusterSpec object at 0x7fd16fb2be10>, '_model_dir': 'gs://path/model_dir/', '_save_checkpoints_steps': None, '_keep_checkpoint_every_n_hours': 10000, '_session_config': intra_op_parallelism_threads: 1
gpu_options {
}
allow_soft_placement: true
, '_tf_random_seed': None, '_environment': u'cloud', '_num_worker_replicas': 1, '_task_id': 0, '_save_summary_steps': 100, '_tf_config': gpu_options {
per_process_gpu_memory_fraction: 1.0
}
, '_evaluation_master': '', '_master': u'grpc://master-ip:8000'}
...
2017-08-01 19:59:26.496208: I tensorflow/core/common_runtime/gpu/gpu_device.cc:940] Found device 0 with properties:
name: Tesla K80
major: 3 minor: 7 memoryClockRate (GHz) 0.8235
pciBusID 0000:00:04.0
Total memory: 11.17GiB
Free memory: 11.09GiB
2017-08-01 19:59:26.775660: I tensorflow/core/common_runtime/gpu/gpu_device.cc:940] Found device 1 with properties:
name: Tesla K80
major: 3 minor: 7 memoryClockRate (GHz) 0.8235
pciBusID 0000:00:05.0
Total memory: 11.17GiB
Free memory: 11.10GiB
...
2017-08-01 19:59:29.675171: I tensorflow/core/distributed_runtime/rpc/grpc_server_lib.cc:316] Started server with target: grpc://localhost:8000
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_1/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_2/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_3/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_4/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_5/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_6/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1/avg_pool/: (?, 16, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_1/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_2/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_3/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_4/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_1/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_2/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_3/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_4/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_5/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_6/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1/avg_pool/: (?, 32, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_1/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_2/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_3/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_4/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_5/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_6/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/global_avg_pool/: (?, 64)
INFO:tensorflow:image after unit resnet/tower_0/fully_connected/: (?, 11)
INFO:tensorflow:SyncReplicasV2: replicas_to_aggregate=1; total_num_replicas=1
INFO:tensorflow:Create CheckpointSaverHook.
INFO:tensorflow:Restoring parameters from gs://path/model_dir/model.ckpt-0
2017-08-01 19:59:37.560775: I tensorflow/core/distributed_runtime/master_session.cc:999] Start master session 156fcb55fe6648d6 with config:
intra_op_parallelism_threads: 1
gpu_options {
per_process_gpu_memory_fraction: 1
}
allow_soft_placement: true
INFO:tensorflow:Saving checkpoints for 1 into gs://path/model_dir/model.ckpt.
INFO:tensorflow:loss = 1.20682, step = 1
INFO:tensorflow:loss = 1.20682, learning_rate = 0.1
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_1/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_2/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_3/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_4/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_5/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_6/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1/avg_pool/: (?, 16, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_1/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_2/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_3/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_4/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_5/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_6/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1/avg_pool/: (?, 32, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_1/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_2/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_3/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_4/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_5/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_6/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/global_avg_pool/: (?, 64)
INFO:tensorflow:image after unit resnet/tower_0/fully_connected/: (?, 11)
INFO:tensorflow:SyncReplicasV2: replicas_to_aggregate=2; total_num_replicas=2
INFO:tensorflow:Starting evaluation at 2017-08-01-20:00:14
2017-08-01 20:00:15.745881: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Creating TensorFlow device (/gpu:0) -> (device: 0, name: Tesla K80, pci bus id: 0000:00:04.0)
2017-08-01 20:00:15.745949: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Creating TensorFlow device (/gpu:1) -> (device: 1, name: Tesla K80, pci bus id: 0000:00:05.0)
2017-08-01 20:00:15.745958: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Creating TensorFlow device (/gpu:2) -> (device: 2, name: Tesla K80, pci bus id: 0000:00:06.0)
2017-08-01 20:00:15.745964: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Creating TensorFlow device (/gpu:3) -> (device: 3, name: Tesla K80, pci bus id: 0000:00:07.0)
2017-08-01 20:00:15.745969: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Creating TensorFlow device (/gpu:4) -> (device: 4, name: Tesla K80, pci bus id: 0000:00:08.0)
2017-08-01 20:00:15.745975: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Creating TensorFlow device (/gpu:5) -> (device: 5, name: Tesla K80, pci bus id: 0000:00:09.0)
2017-08-01 20:00:15.745987: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Creating TensorFlow device (/gpu:6) -> (device: 6, name: Tesla K80, pci bus id: 0000:00:0a.0)
2017-08-01 20:00:15.745997: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Creating TensorFlow device (/gpu:7) -> (device: 7, name: Tesla K80, pci bus id: 0000:00:0b.0)
INFO:tensorflow:Restoring parameters from gs://path/model_dir/model.ckpt-10023
INFO:tensorflow:Evaluation [1/100]
INFO:tensorflow:Evaluation [2/100]
INFO:tensorflow:Evaluation [3/100]
INFO:tensorflow:Evaluation [4/100]
INFO:tensorflow:Evaluation [5/100]
INFO:tensorflow:Evaluation [6/100]
INFO:tensorflow:Evaluation [7/100]
INFO:tensorflow:Evaluation [8/100]
INFO:tensorflow:Evaluation [9/100]
INFO:tensorflow:Evaluation [10/100]
INFO:tensorflow:Evaluation [11/100]
INFO:tensorflow:Evaluation [12/100]
INFO:tensorflow:Evaluation [13/100]
...
INFO:tensorflow:Evaluation [100/100]
INFO:tensorflow:Finished evaluation at 2017-08-01-20:00:31
INFO:tensorflow:Saving dict for global step 1: accuracy = 0.0994, global_step = 1, loss = 630.425
```
#### Worker
Run this on worker:
Runs an Experiment in sync mode on 4 GPUs using CPU as parameter server for
40000 steps. It will run evaluation a couple of times during training. Make sure
the model_dir is the same as defined on the TF_CONFIG.
```shell
python cifar10_main.py --data-dir=gs://path/cifar-10-data \
--job-dir=gs://path/model_dir/ \
--num-gpus=4 \
--train-steps=40000 \
--sync
```
*Output:*
```shell
INFO:tensorflow:Using model_dir in TF_CONFIG: gs://path/model_dir/
INFO:tensorflow:Using config: {'_save_checkpoints_secs': 600,
'_num_ps_replicas': 1, '_keep_checkpoint_max': 5, '_task_type': u'worker',
'_is_chief': False, '_cluster_spec':
<tensorflow.python.training.server_lib.ClusterSpec object at 0x7f6918438e10>,
'_model_dir': 'gs://<path>/model_dir/',
'_save_checkpoints_steps': None, '_keep_checkpoint_every_n_hours': 10000,
'_session_config': intra_op_parallelism_threads: 1
gpu_options {
}
allow_soft_placement: true
, '_tf_random_seed': None, '_environment': u'cloud', '_num_worker_replicas': 1,
'_task_id': 0, '_save_summary_steps': 100, '_tf_config': gpu_options {
per_process_gpu_memory_fraction: 1.0
}
...
2017-08-01 19:59:26.496208: I tensorflow/core/common_runtime/gpu/gpu_device.cc:940] Found device 0 with properties:
name: Tesla K80
major: 3 minor: 7 memoryClockRate (GHz) 0.8235
pciBusID 0000:00:04.0
Total memory: 11.17GiB
Free memory: 11.09GiB
2017-08-01 19:59:26.775660: I tensorflow/core/common_runtime/gpu/gpu_device.cc:940] Found device 1 with properties:
name: Tesla K80
major: 3 minor: 7 memoryClockRate (GHz) 0.8235
pciBusID 0000:00:05.0
Total memory: 11.17GiB
Free memory: 11.10GiB
...
2017-08-01 19:59:29.675171: I tensorflow/core/distributed_runtime/rpc/grpc_server_lib.cc:316] Started server with target: grpc://localhost:8000
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_1/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_2/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_3/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_4/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_5/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage/residual_v1_6/: (?, 16, 32, 32)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1/avg_pool/: (?, 16, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_1/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_2/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_3/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_4/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_1/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_2/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_3/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_4/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_5/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_1/residual_v1_6/: (?, 32, 16, 16)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1/avg_pool/: (?, 32, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_1/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_2/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_3/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_4/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_5/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/stage_2/residual_v1_6/: (?, 64, 8, 8)
INFO:tensorflow:image after unit resnet/tower_0/global_avg_pool/: (?, 64)
INFO:tensorflow:image after unit resnet/tower_0/fully_connected/: (?, 11)
INFO:tensorflow:SyncReplicasV2: replicas_to_aggregate=2; total_num_replicas=2
INFO:tensorflow:Create CheckpointSaverHook.
2017-07-31 22:38:04.629150: I
tensorflow/core/distributed_runtime/master.cc:209] CreateSession still waiting
for response from worker: /job:master/replica:0/task:0
2017-07-31 22:38:09.263492: I
tensorflow/core/distributed_runtime/master_session.cc:999] Start master
session cc58f93b1e259b0c with config:
intra_op_parallelism_threads: 1
gpu_options {
per_process_gpu_memory_fraction: 1
}
allow_soft_placement: true
INFO:tensorflow:loss = 5.82382, step = 0
INFO:tensorflow:loss = 5.82382, learning_rate = 0.8
INFO:tensorflow:Average examples/sec: 1116.92 (1116.92), step = 10
INFO:tensorflow:Average examples/sec: 1233.73 (1377.83), step = 20
INFO:tensorflow:Average examples/sec: 1485.43 (2509.3), step = 30
INFO:tensorflow:Average examples/sec: 1680.27 (2770.39), step = 40
INFO:tensorflow:Average examples/sec: 1825.38 (2788.78), step = 50
INFO:tensorflow:Average examples/sec: 1929.32 (2697.27), step = 60
INFO:tensorflow:Average examples/sec: 2015.17 (2749.05), step = 70
INFO:tensorflow:loss = 37.6272, step = 79 (19.554 sec)
INFO:tensorflow:loss = 37.6272, learning_rate = 0.8 (19.554 sec)
INFO:tensorflow:Average examples/sec: 2074.92 (2618.36), step = 80
INFO:tensorflow:Average examples/sec: 2132.71 (2744.13), step = 90
INFO:tensorflow:Average examples/sec: 2183.38 (2777.21), step = 100
INFO:tensorflow:Average examples/sec: 2224.4 (2739.03), step = 110
INFO:tensorflow:Average examples/sec: 2240.28 (2431.26), step = 120
INFO:tensorflow:Average examples/sec: 2272.12 (2739.32), step = 130
INFO:tensorflow:Average examples/sec: 2300.68 (2750.03), step = 140
INFO:tensorflow:Average examples/sec: 2325.81 (2745.63), step = 150
INFO:tensorflow:Average examples/sec: 2347.14 (2721.53), step = 160
INFO:tensorflow:Average examples/sec: 2367.74 (2754.54), step = 170
INFO:tensorflow:loss = 27.8453, step = 179 (18.893 sec)
...
```
#### PS
Run this on ps:
The ps will not do training so most of the arguments won't affect the execution
```shell
python cifar10_main.py --job-dir=gs://path/model_dir/
```
*Output:*
```shell
INFO:tensorflow:Using model_dir in TF_CONFIG: gs://path/model_dir/
INFO:tensorflow:Using config: {'_save_checkpoints_secs': 600, '_num_ps_replicas': 1, '_keep_checkpoint_max': 5, '_task_type': u'ps', '_is_chief': False, '_cluster_spec': <tensorflow.python.training.server_lib.ClusterSpec object at 0x7f48f1addf90>, '_model_dir': 'gs://path/model_dir/', '_save_checkpoints_steps': None, '_keep_checkpoint_every_n_hours': 10000, '_session_config': intra_op_parallelism_threads: 1
gpu_options {
}
allow_soft_placement: true
, '_tf_random_seed': None, '_environment': u'cloud', '_num_worker_replicas': 1, '_task_id': 0, '_save_summary_steps': 100, '_tf_config': gpu_options {
per_process_gpu_memory_fraction: 1.0
}
, '_evaluation_master': '', '_master': u'grpc://master-ip:8000'}
2017-07-31 22:54:58.928088: I tensorflow/core/distributed_runtime/rpc/grpc_channel.cc:215] Initialize GrpcChannelCache for job master -> {0 -> master-ip:8000}
2017-07-31 22:54:58.928153: I tensorflow/core/distributed_runtime/rpc/grpc_channel.cc:215] Initialize GrpcChannelCache for job ps -> {0 -> localhost:8000}
2017-07-31 22:54:58.928160: I tensorflow/core/distributed_runtime/rpc/grpc_channel.cc:215] Initialize GrpcChannelCache for job worker -> {0 -> worker-ip:8000}
2017-07-31 22:54:58.929873: I tensorflow/core/distributed_runtime/rpc/grpc_server_lib.cc:316] Started server with target: grpc://localhost:8000
```
## Visualizing results with TensorBoard
When using Estimators you can also visualize your data in TensorBoard, with no
changes in your code. You can use TensorBoard to visualize your TensorFlow
graph, plot quantitative metrics about the execution of your graph, and show
additional data like images that pass through it.
You'll see something similar to this if you "point" TensorBoard to the
`job dir` parameter you used to train or evaluate your model.
Check TensorBoard during training or after it. Just point TensorBoard to the
model_dir you chose on the previous step.
```shell
tensorboard --log-dir="<job dir>"
```
## Warnings
When runninng `cifar10_main.py` with `--sync` argument you may see an error
similar to:
```python
File "cifar10_main.py", line 538, in <module>
tf.app.run()
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/app.py", line 48, in run
_sys.exit(main(_sys.argv[:1] + flags_passthrough))
File "cifar10_main.py", line 518, in main
hooks), run_config=config)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/learn_runner.py", line 210, in run
return _execute_schedule(experiment, schedule)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/learn_runner.py", line 47, in _execute_schedule
return task()
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/experiment.py", line 501, in train_and_evaluate
hooks=self._eval_hooks)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/contrib/learn/python/learn/experiment.py", line 681, in _call_evaluate
hooks=hooks)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/estimator/estimator.py", line 292, in evaluate
name=name)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/estimator/estimator.py", line 638, in _evaluate_model
features, labels, model_fn_lib.ModeKeys.EVAL)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/estimator/estimator.py", line 545, in _call_model_fn
features=features, labels=labels, **kwargs)
File "cifar10_main.py", line 331, in _resnet_model_fn
gradvars, global_step=tf.train.get_global_step())
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/training/sync_replicas_optimizer.py", line 252, in apply_gradients
variables.global_variables())
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/util/tf_should_use.py", line 170, in wrapped
return _add_should_use_warning(fn(*args, **kwargs))
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/util/tf_should_use.py", line 139, in _add_should_use_warning
wrapped = TFShouldUseWarningWrapper(x)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/util/tf_should_use.py", line 96, in __init__
stack = [s.strip() for s in traceback.format_stack()]
```
This should not affect your training, and should be fixed on the next releases.
tutorials/image/cifar10_estimator/cifar10.py deleted 100644 → 0
View file @ bf6d6f6f
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""CIFAR-10 data set.
See http://www.cs.toronto.edu/~kriz/cifar.html.
"""
import os
import tensorflow as tf
HEIGHT = 32
WIDTH = 32
DEPTH = 3
class Cifar10DataSet(object):
"""Cifar10 data set.
Described by http://www.cs.toronto.edu/~kriz/cifar.html.
"""
def __init__(self, data_dir, subset='train', use_distortion=True):
self.data_dir = data_dir
self.subset = subset
self.use_distortion = use_distortion
def get_filenames(self):
if self.subset in ['train', 'validation', 'eval']:
return [os.path.join(self.data_dir, self.subset + '.tfrecords')]
else:
raise ValueError('Invalid data subset "%s"' % self.subset)
def parser(self, serialized_example):
"""Parses a single tf.Example into image and label tensors."""
# Dimensions of the images in the CIFAR-10 dataset.
# See http://www.cs.toronto.edu/~kriz/cifar.html for a description of the
# input format.
features = tf.parse_single_example(
serialized_example,
features={
'image': tf.FixedLenFeature([], tf.string),
'label': tf.FixedLenFeature([], tf.int64),
})
image = tf.decode_raw(features['image'], tf.uint8)
image.set_shape([DEPTH * HEIGHT * WIDTH])
# Reshape from [depth * height * width] to [depth, height, width].
image = tf.cast(
tf.transpose(tf.reshape(image, [DEPTH, HEIGHT, WIDTH]), [1, 2, 0]),
tf.float32)
label = tf.cast(features['label'], tf.int32)
# Custom preprocessing.
image = self.preprocess(image)
return image, label
def make_batch(self, batch_size):
"""Read the images and labels from 'filenames'."""
filenames = self.get_filenames()
# Repeat infinitely.
dataset = tf.data.TFRecordDataset(filenames).repeat()
# Parse records.
dataset = dataset.map(
self.parser, num_parallel_calls=batch_size)
# Potentially shuffle records.
if self.subset == 'train':
min_queue_examples = int(
Cifar10DataSet.num_examples_per_epoch(self.subset) * 0.4)
# Ensure that the capacity is sufficiently large to provide good random
# shuffling.
dataset = dataset.shuffle(buffer_size=min_queue_examples + 3 * batch_size)
# Batch it up.
dataset = dataset.batch(batch_size)
iterator = dataset.make_one_shot_iterator()
image_batch, label_batch = iterator.get_next()
return image_batch, label_batch
def preprocess(self, image):
"""Preprocess a single image in [height, width, depth] layout."""
if self.subset == 'train' and self.use_distortion:
# Pad 4 pixels on each dimension of feature map, done in mini-batch
image = tf.image.resize_image_with_crop_or_pad(image, 40, 40)
image = tf.random_crop(image, [HEIGHT, WIDTH, DEPTH])
image = tf.image.random_flip_left_right(image)
return image
@staticmethod
def num_examples_per_epoch(subset='train'):
if subset == 'train':
return 45000
elif subset == 'validation':
return 5000
elif subset == 'eval':
return 10000
else:
raise ValueError('Invalid data subset "%s"' % subset)
tutorials/image/cifar10_estimator/cifar10_main.py deleted 100644 → 0
View file @ bf6d6f6f
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""ResNet model for classifying images from CIFAR-10 dataset.
Support single-host training with one or multiple devices.
ResNet as proposed in:
Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun
Deep Residual Learning for Image Recognition. arXiv:1512.03385
CIFAR-10 as in:
http://www.cs.toronto.edu/~kriz/cifar.html
"""
from __future__ import division
from __future__ import print_function
import argparse
import functools
import itertools
import os
import cifar10
import cifar10_model
import cifar10_utils
import numpy as np
import six
from six.moves import xrange # pylint: disable=redefined-builtin
import tensorflow as tf
tf.logging.set_verbosity(tf.logging.INFO)
def get_model_fn(num_gpus, variable_strategy, num_workers):
"""Returns a function that will build the resnet model."""
def _resnet_model_fn(features, labels, mode, params):
"""Resnet model body.
Support single host, one or more GPU training. Parameter distribution can
be either one of the following scheme.
1. CPU is the parameter server and manages gradient updates.
2. Parameters are distributed evenly across all GPUs, and the first GPU
manages gradient updates.
Args:
features: a list of tensors, one for each tower
labels: a list of tensors, one for each tower
mode: ModeKeys.TRAIN or EVAL
params: Hyperparameters suitable for tuning
Returns:
A EstimatorSpec object.
"""
is_training = (mode == tf.estimator.ModeKeys.TRAIN)
weight_decay = params.weight_decay
momentum = params.momentum
tower_features = features
tower_labels = labels
tower_losses = []
tower_gradvars = []
tower_preds = []
# channels first (NCHW) is normally optimal on GPU and channels last (NHWC)
# on CPU. The exception is Intel MKL on CPU which is optimal with
# channels_last.
data_format = params.data_format
if not data_format:
if num_gpus == 0:
data_format = 'channels_last'
else:
data_format = 'channels_first'
if num_gpus == 0:
num_devices = 1
device_type = 'cpu'
else:
num_devices = num_gpus
device_type = 'gpu'
for i in range(num_devices):
worker_device = '/{}:{}'.format(device_type, i)
if variable_strategy == 'CPU':
device_setter = cifar10_utils.local_device_setter(
worker_device=worker_device)
elif variable_strategy == 'GPU':
device_setter = cifar10_utils.local_device_setter(
ps_device_type='gpu',
worker_device=worker_device,
ps_strategy=tf.contrib.training.GreedyLoadBalancingStrategy(
num_gpus, tf.contrib.training.byte_size_load_fn))
with tf.variable_scope('resnet', reuse=bool(i != 0)):
with tf.name_scope('tower_%d' % i) as name_scope:
with tf.device(device_setter):
loss, gradvars, preds = _tower_fn(
is_training, weight_decay, tower_features[i], tower_labels[i],
data_format, params.num_layers, params.batch_norm_decay,
params.batch_norm_epsilon)
tower_losses.append(loss)
tower_gradvars.append(gradvars)
tower_preds.append(preds)
if i == 0:
# Only trigger batch_norm moving mean and variance update from
# the 1st tower. Ideally, we should grab the updates from all
# towers but these stats accumulate extremely fast so we can
# ignore the other stats from the other towers without
# significant detriment.
update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS,
name_scope)
# Now compute global loss and gradients.
gradvars = []
with tf.name_scope('gradient_averaging'):
all_grads = {}
for grad, var in itertools.chain(*tower_gradvars):
if grad is not None:
all_grads.setdefault(var, []).append(grad)
for var, grads in six.iteritems(all_grads):
# Average gradients on the same device as the variables
# to which they apply.
with tf.device(var.device):
if len(grads) == 1:
avg_grad = grads[0]
else:
avg_grad = tf.multiply(tf.add_n(grads), 1. / len(grads))
gradvars.append((avg_grad, var))
# Device that runs the ops to apply global gradient updates.
consolidation_device = '/gpu:0' if variable_strategy == 'GPU' else '/cpu:0'
with tf.device(consolidation_device):
# Suggested learning rate scheduling from
# https://github.com/ppwwyyxx/tensorpack/blob/master/examples/ResNet/cifar10-resnet.py#L155
num_batches_per_epoch = cifar10.Cifar10DataSet.num_examples_per_epoch(
'train') // (params.train_batch_size * num_workers)
boundaries = [
num_batches_per_epoch * x
for x in np.array([82, 123, 300], dtype=np.int64)
]
staged_lr = [params.learning_rate * x for x in [1, 0.1, 0.01, 0.002]]
learning_rate = tf.train.piecewise_constant(tf.train.get_global_step(),
boundaries, staged_lr)
loss = tf.reduce_mean(tower_losses, name='loss')
examples_sec_hook = cifar10_utils.ExamplesPerSecondHook(
params.train_batch_size, every_n_steps=10)
tensors_to_log = {'learning_rate': learning_rate, 'loss': loss}
logging_hook = tf.train.LoggingTensorHook(
tensors=tensors_to_log, every_n_iter=100)
train_hooks = [logging_hook, examples_sec_hook]
optimizer = tf.train.MomentumOptimizer(
learning_rate=learning_rate, momentum=momentum)
if params.sync:
optimizer = tf.train.SyncReplicasOptimizer(
optimizer, replicas_to_aggregate=num_workers)
sync_replicas_hook = optimizer.make_session_run_hook(params.is_chief)
train_hooks.append(sync_replicas_hook)
# Create single grouped train op
train_op = [
optimizer.apply_gradients(
gradvars, global_step=tf.train.get_global_step())
]
train_op.extend(update_ops)
train_op = tf.group(*train_op)
predictions = {
'classes':
tf.concat([p['classes'] for p in tower_preds], axis=0),
'probabilities':
tf.concat([p['probabilities'] for p in tower_preds], axis=0)
}
stacked_labels = tf.concat(labels, axis=0)
metrics = {
'accuracy':
tf.metrics.accuracy(stacked_labels, predictions['classes'])
}
return tf.estimator.EstimatorSpec(
mode=mode,
predictions=predictions,
loss=loss,
train_op=train_op,
training_hooks=train_hooks,
eval_metric_ops=metrics)
return _resnet_model_fn
def _tower_fn(is_training, weight_decay, feature, label, data_format,
num_layers, batch_norm_decay, batch_norm_epsilon):
"""Build computation tower (Resnet).
Args:
is_training: true if is training graph.
weight_decay: weight regularization strength, a float.
feature: a Tensor.
label: a Tensor.
data_format: channels_last (NHWC) or channels_first (NCHW).
num_layers: number of layers, an int.
batch_norm_decay: decay for batch normalization, a float.
batch_norm_epsilon: epsilon for batch normalization, a float.
Returns:
A tuple with the loss for the tower, the gradients and parameters, and
predictions.
"""
model = cifar10_model.ResNetCifar10(
num_layers,
batch_norm_decay=batch_norm_decay,
batch_norm_epsilon=batch_norm_epsilon,
is_training=is_training,
data_format=data_format)
logits = model.forward_pass(feature, input_data_format='channels_last')
tower_pred = {
'classes': tf.argmax(input=logits, axis=1),
'probabilities': tf.nn.softmax(logits)
}
tower_loss = tf.losses.sparse_softmax_cross_entropy(
logits=logits, labels=label)
tower_loss = tf.reduce_mean(tower_loss)
model_params = tf.trainable_variables()
tower_loss += weight_decay * tf.add_n(
[tf.nn.l2_loss(v) for v in model_params])
tower_grad = tf.gradients(tower_loss, model_params)
return tower_loss, zip(tower_grad, model_params), tower_pred
def input_fn(data_dir,
subset,
num_shards,
batch_size,
use_distortion_for_training=True):
"""Create input graph for model.
Args:
data_dir: Directory where TFRecords representing the dataset are located.
subset: one of 'train', 'validate' and 'eval'.
num_shards: num of towers participating in data-parallel training.
batch_size: total batch size for training to be divided by the number of
shards.
use_distortion_for_training: True to use distortions.
Returns:
two lists of tensors for features and labels, each of num_shards length.
"""
with tf.device('/cpu:0'):
use_distortion = subset == 'train' and use_distortion_for_training
dataset = cifar10.Cifar10DataSet(data_dir, subset, use_distortion)
image_batch, label_batch = dataset.make_batch(batch_size)
if num_shards <= 1:
# No GPU available or only 1 GPU.
return [image_batch], [label_batch]
# Note that passing num=batch_size is safe here, even though
# dataset.batch(batch_size) can, in some cases, return fewer than batch_size
# examples. This is because it does so only when repeating for a limited
# number of epochs, but our dataset repeats forever.
image_batch = tf.unstack(image_batch, num=batch_size, axis=0)
label_batch = tf.unstack(label_batch, num=batch_size, axis=0)
feature_shards = [[] for i in range(num_shards)]
label_shards = [[] for i in range(num_shards)]
for i in xrange(batch_size):
idx = i % num_shards
feature_shards[idx].append(image_batch[i])
label_shards[idx].append(label_batch[i])
feature_shards = [tf.parallel_stack(x) for x in feature_shards]
label_shards = [tf.parallel_stack(x) for x in label_shards]
return feature_shards, label_shards
def get_experiment_fn(data_dir,
num_gpus,
variable_strategy,
use_distortion_for_training=True):
"""Returns an Experiment function.
Experiments perform training on several workers in parallel,
in other words experiments know how to invoke train and eval in a sensible
fashion for distributed training. Arguments passed directly to this
function are not tunable, all other arguments should be passed within
tf.HParams, passed to the enclosed function.
Args:
data_dir: str. Location of the data for input_fns.
num_gpus: int. Number of GPUs on each worker.
variable_strategy: String. CPU to use CPU as the parameter server
and GPU to use the GPUs as the parameter server.
use_distortion_for_training: bool. See cifar10.Cifar10DataSet.
Returns:
A function (tf.estimator.RunConfig, tf.contrib.training.HParams) ->
tf.contrib.learn.Experiment.
Suitable for use by tf.contrib.learn.learn_runner, which will run various
methods on Experiment (train, evaluate) based on information
about the current runner in `run_config`.
"""
def _experiment_fn(run_config, hparams):
"""Returns an Experiment."""
# Create estimator.
train_input_fn = functools.partial(
input_fn,
data_dir,
subset='train',
num_shards=num_gpus,
batch_size=hparams.train_batch_size,
use_distortion_for_training=use_distortion_for_training)
eval_input_fn = functools.partial(
input_fn,
data_dir,
subset='eval',
batch_size=hparams.eval_batch_size,
num_shards=num_gpus)
num_eval_examples = cifar10.Cifar10DataSet.num_examples_per_epoch('eval')
if num_eval_examples % hparams.eval_batch_size != 0:
raise ValueError(
'validation set size must be multiple of eval_batch_size')
train_steps = hparams.train_steps
eval_steps = num_eval_examples // hparams.eval_batch_size
classifier = tf.estimator.Estimator(
model_fn=get_model_fn(num_gpus, variable_strategy,
run_config.num_worker_replicas or 1),
config=run_config,
params=hparams)
# Create experiment.
return tf.contrib.learn.Experiment(
classifier,
train_input_fn=train_input_fn,
eval_input_fn=eval_input_fn,
train_steps=train_steps,
eval_steps=eval_steps)
return _experiment_fn
def main(job_dir, data_dir, num_gpus, variable_strategy,
use_distortion_for_training, log_device_placement, num_intra_threads,
**hparams):
# The env variable is on deprecation path, default is set to off.
os.environ['TF_SYNC_ON_FINISH'] = '0'
os.environ['TF_ENABLE_WINOGRAD_NONFUSED'] = '1'
# Session configuration.
sess_config = tf.ConfigProto(
allow_soft_placement=True,
log_device_placement=log_device_placement,
intra_op_parallelism_threads=num_intra_threads,
gpu_options=tf.GPUOptions(force_gpu_compatible=True))
config = cifar10_utils.RunConfig(
session_config=sess_config, model_dir=job_dir)
tf.contrib.learn.learn_runner.run(
get_experiment_fn(data_dir, num_gpus, variable_strategy,
use_distortion_for_training),
run_config=config,
hparams=tf.contrib.training.HParams(
is_chief=config.is_chief,
**hparams))
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument(
'--data-dir',
type=str,
required=True,
help='The directory where the CIFAR-10 input data is stored.')
parser.add_argument(
'--job-dir',
type=str,
required=True,
help='The directory where the model will be stored.')
parser.add_argument(
'--variable-strategy',
choices=['CPU', 'GPU'],
type=str,
default='CPU',
help='Where to locate variable operations')
parser.add_argument(
'--num-gpus',
type=int,
default=1,
help='The number of gpus used. Uses only CPU if set to 0.')
parser.add_argument(
'--num-layers',
type=int,
default=44,
help='The number of layers of the model.')
parser.add_argument(
'--train-steps',
type=int,
default=80000,
help='The number of steps to use for training.')
parser.add_argument(
'--train-batch-size',
type=int,
default=128,
help='Batch size for training.')
parser.add_argument(
'--eval-batch-size',
type=int,
default=100,
help='Batch size for validation.')
parser.add_argument(
'--momentum',
type=float,
default=0.9,
help='Momentum for MomentumOptimizer.')
parser.add_argument(
'--weight-decay',
type=float,
default=2e-4,
help='Weight decay for convolutions.')
parser.add_argument(
'--learning-rate',
type=float,
default=0.1,
help="""\
This is the inital learning rate value. The learning rate will decrease
during training. For more details check the model_fn implementation in
this file.\
""")
parser.add_argument(
'--use-distortion-for-training',
type=bool,
default=True,
help='If doing image distortion for training.')
parser.add_argument(
'--sync',
action='store_true',
default=False,
help="""\
If present when running in a distributed environment will run on sync mode.\
""")
parser.add_argument(
'--num-intra-threads',
type=int,
default=0,
help="""\
Number of threads to use for intra-op parallelism. When training on CPU
set to 0 to have the system pick the appropriate number or alternatively
set it to the number of physical CPU cores.\
""")
parser.add_argument(
'--num-inter-threads',
type=int,
default=0,
help="""\
Number of threads to use for inter-op parallelism. If set to 0, the
system will pick an appropriate number.\
""")
parser.add_argument(
'--data-format',
type=str,
default=None,
help="""\
If not set, the data format best for the training device is used.
Allowed values: channels_first (NCHW) channels_last (NHWC).\
""")
parser.add_argument(
'--log-device-placement',
action='store_true',
default=False,
help='Whether to log device placement.')
parser.add_argument(
'--batch-norm-decay',
type=float,
default=0.997,
help='Decay for batch norm.')
parser.add_argument(
'--batch-norm-epsilon',
type=float,
default=1e-5,
help='Epsilon for batch norm.')
args = parser.parse_args()
if args.num_gpus > 0:
assert tf.test.is_gpu_available(), "Requested GPUs but none found."
if args.num_gpus < 0:
raise ValueError(
'Invalid GPU count: \"--num-gpus\" must be 0 or a positive integer.')
if args.num_gpus == 0 and args.variable_strategy == 'GPU':
raise ValueError('num-gpus=0, CPU must be used as parameter server. Set'
'--variable-strategy=CPU.')
if (args.num_layers - 2) % 6 != 0:
raise ValueError('Invalid --num-layers parameter.')
if args.num_gpus != 0 and args.train_batch_size % args.num_gpus != 0:
raise ValueError('--train-batch-size must be multiple of --num-gpus.')
if args.num_gpus != 0 and args.eval_batch_size % args.num_gpus != 0:
raise ValueError('--eval-batch-size must be multiple of --num-gpus.')
main(**vars(args))
tutorials/image/cifar10_estimator/cifar10_model.py deleted 100644 → 0
View file @ bf6d6f6f
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Model class for Cifar10 Dataset."""
from __future__ import division
from __future__ import print_function
import tensorflow as tf
import model_base
class ResNetCifar10(model_base.ResNet):
"""Cifar10 model with ResNetV1 and basic residual block."""
def __init__(self,
num_layers,
is_training,
batch_norm_decay,
batch_norm_epsilon,
data_format='channels_first'):
super(ResNetCifar10, self).__init__(
is_training,
data_format,
batch_norm_decay,
batch_norm_epsilon
)
self.n = (num_layers - 2) // 6
# Add one in case label starts with 1. No impact if label starts with 0.
self.num_classes = 10 + 1
self.filters = [16, 16, 32, 64]
self.strides = [1, 2, 2]
def forward_pass(self, x, input_data_format='channels_last'):
"""Build the core model within the graph."""
if self._data_format != input_data_format:
if input_data_format == 'channels_last':
# Computation requires channels_first.
x = tf.transpose(x, [0, 3, 1, 2])
else:
# Computation requires channels_last.
x = tf.transpose(x, [0, 2, 3, 1])
# Image standardization.
x = x / 128 - 1
x = self._conv(x, 3, 16, 1)
x = self._batch_norm(x)
x = self._relu(x)
# Use basic (non-bottleneck) block and ResNet V1 (post-activation).
res_func = self._residual_v1
# 3 stages of block stacking.
for i in range(3):
with tf.name_scope('stage'):
for j in range(self.n):
if j == 0:
# First block in a stage, filters and strides may change.
x = res_func(x, 3, self.filters[i], self.filters[i + 1],
self.strides[i])
else:
# Following blocks in a stage, constant filters and unit stride.
x = res_func(x, 3, self.filters[i + 1], self.filters[i + 1], 1)
x = self._global_avg_pool(x)
x = self._fully_connected(x, self.num_classes)
return x
tutorials/image/cifar10_estimator/cifar10_utils.py deleted 100644 → 0
View file @ bf6d6f6f
import collections
import six
import tensorflow as tf
from tensorflow.python.platform import tf_logging as logging
from tensorflow.core.framework import node_def_pb2
from tensorflow.python.framework import device as pydev
from tensorflow.python.training import basic_session_run_hooks
from tensorflow.python.training import session_run_hook
from tensorflow.python.training import training_util
from tensorflow.python.training import device_setter
from tensorflow.contrib.learn.python.learn import run_config
# TODO(b/64848083) Remove once uid bug is fixed
class RunConfig(tf.contrib.learn.RunConfig):
def uid(self, whitelist=None):
"""Generates a 'Unique Identifier' based on all internal fields.
Caller should use the uid string to check `RunConfig` instance integrity
in one session use, but should not rely on the implementation details, which
is subject to change.
Args:
whitelist: A list of the string names of the properties uid should not
include. If `None`, defaults to `_DEFAULT_UID_WHITE_LIST`, which
includes most properties user allowes to change.
Returns:
A uid string.
"""
if whitelist is None:
whitelist = run_config._DEFAULT_UID_WHITE_LIST
state = {k: v for k, v in self.__dict__.items() if not k.startswith('__')}
# Pop out the keys in whitelist.
for k in whitelist:
state.pop('_' + k, None)
ordered_state = collections.OrderedDict(
sorted(state.items(), key=lambda t: t[0]))
# For class instance without __repr__, some special cares are required.
# Otherwise, the object address will be used.
if '_cluster_spec' in ordered_state:
ordered_state['_cluster_spec'] = collections.OrderedDict(
sorted(ordered_state['_cluster_spec'].as_dict().items(),
key=lambda t: t[0])
)
return ', '.join(
'%s=%r' % (k, v) for (k, v) in six.iteritems(ordered_state))
class ExamplesPerSecondHook(session_run_hook.SessionRunHook):
"""Hook to print out examples per second.
Total time is tracked and then divided by the total number of steps
to get the average step time and then batch_size is used to determine
the running average of examples per second. The examples per second for the
most recent interval is also logged.
"""
def __init__(
self,
batch_size,
every_n_steps=100,
every_n_secs=None,):
"""Initializer for ExamplesPerSecondHook.
Args:
batch_size: Total batch size used to calculate examples/second from
global time.
every_n_steps: Log stats every n steps.
every_n_secs: Log stats every n seconds.
"""
if (every_n_steps is None) == (every_n_secs is None):
raise ValueError('exactly one of every_n_steps'
' and every_n_secs should be provided.')
self._timer = basic_session_run_hooks.SecondOrStepTimer(
every_steps=every_n_steps, every_secs=every_n_secs)
self._step_train_time = 0
self._total_steps = 0
self._batch_size = batch_size
def begin(self):
self._global_step_tensor = training_util.get_global_step()
if self._global_step_tensor is None:
raise RuntimeError(
'Global step should be created to use StepCounterHook.')
def before_run(self, run_context): # pylint: disable=unused-argument
return basic_session_run_hooks.SessionRunArgs(self._global_step_tensor)
def after_run(self, run_context, run_values):
_ = run_context
global_step = run_values.results
if self._timer.should_trigger_for_step(global_step):
elapsed_time, elapsed_steps = self._timer.update_last_triggered_step(
global_step)
if elapsed_time is not None:
steps_per_sec = elapsed_steps / elapsed_time
self._step_train_time += elapsed_time
self._total_steps += elapsed_steps
average_examples_per_sec = self._batch_size * (
self._total_steps / self._step_train_time)
current_examples_per_sec = steps_per_sec * self._batch_size
# Average examples/sec followed by current examples/sec
logging.info('%s: %g (%g), step = %g', 'Average examples/sec',
average_examples_per_sec, current_examples_per_sec,
self._total_steps)
def local_device_setter(num_devices=1,
ps_device_type='cpu',
worker_device='/cpu:0',
ps_ops=None,
ps_strategy=None):
if ps_ops == None:
ps_ops = ['Variable', 'VariableV2', 'VarHandleOp']
if ps_strategy is None:
ps_strategy = device_setter._RoundRobinStrategy(num_devices)
if not six.callable(ps_strategy):
raise TypeError("ps_strategy must be callable")
def _local_device_chooser(op):
current_device = pydev.DeviceSpec.from_string(op.device or "")
node_def = op if isinstance(op, node_def_pb2.NodeDef) else op.node_def
if node_def.op in ps_ops:
ps_device_spec = pydev.DeviceSpec.from_string(
'/{}:{}'.format(ps_device_type, ps_strategy(op)))
ps_device_spec.merge_from(current_device)
return ps_device_spec.to_string()
else:
worker_device_spec = pydev.DeviceSpec.from_string(worker_device or "")
worker_device_spec.merge_from(current_device)
return worker_device_spec.to_string()
return _local_device_chooser
tutorials/image/cifar10_estimator/cmle_config.yaml deleted 100644 → 0
View file @ bf6d6f6f
trainingInput:
scaleTier: CUSTOM
masterType: complex_model_m_gpu
workerType: complex_model_m_gpu
parameterServerType: complex_model_m
workerCount: 1
tutorials/image/cifar10_estimator/generate_cifar10_tfrecords.py deleted 100644 → 0
View file @ bf6d6f6f
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Read CIFAR-10 data from pickled numpy arrays and writes TFRecords.
Generates tf.train.Example protos and writes them to TFRecord files from the
python version of the CIFAR-10 dataset downloaded from
https://www.cs.toronto.edu/~kriz/cifar.html.
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import argparse
import os
import sys
import tarfile
from six.moves import cPickle as pickle
from six.moves import xrange # pylint: disable=redefined-builtin
import tensorflow as tf
CIFAR_FILENAME = 'cifar-10-python.tar.gz'
CIFAR_DOWNLOAD_URL = 'https://www.cs.toronto.edu/~kriz/' + CIFAR_FILENAME
CIFAR_LOCAL_FOLDER = 'cifar-10-batches-py'
def download_and_extract(data_dir):
# download CIFAR-10 if not already downloaded.
tf.contrib.learn.datasets.base.maybe_download(CIFAR_FILENAME, data_dir,
CIFAR_DOWNLOAD_URL)
tarfile.open(os.path.join(data_dir, CIFAR_FILENAME),
'r:gz').extractall(data_dir)
def _int64_feature(value):
return tf.train.Feature(int64_list=tf.train.Int64List(value=[value]))
def _bytes_feature(value):
return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))
def _get_file_names():
"""Returns the file names expected to exist in the input_dir."""
file_names = {}
file_names['train'] = ['data_batch_%d' % i for i in xrange(1, 5)]
file_names['validation'] = ['data_batch_5']
file_names['eval'] = ['test_batch']
return file_names
def read_pickle_from_file(filename):
with tf.gfile.Open(filename, 'rb') as f:
if sys.version_info >= (3, 0):
data_dict = pickle.load(f, encoding='bytes')
else:
data_dict = pickle.load(f)
return data_dict
def convert_to_tfrecord(input_files, output_file):
"""Converts a file to TFRecords."""
print('Generating %s' % output_file)
with tf.python_io.TFRecordWriter(output_file) as record_writer:
for input_file in input_files:
data_dict = read_pickle_from_file(input_file)
data = data_dict[b'data']
labels = data_dict[b'labels']
num_entries_in_batch = len(labels)
for i in range(num_entries_in_batch):
example = tf.train.Example(features=tf.train.Features(
feature={
'image': _bytes_feature(data[i].tobytes()),
'label': _int64_feature(labels[i])
}))
record_writer.write(example.SerializeToString())
def main(data_dir):
print('Download from {} and extract.'.format(CIFAR_DOWNLOAD_URL))
download_and_extract(data_dir)
file_names = _get_file_names()
input_dir = os.path.join(data_dir, CIFAR_LOCAL_FOLDER)
for mode, files in file_names.items():
input_files = [os.path.join(input_dir, f) for f in files]
output_file = os.path.join(data_dir, mode + '.tfrecords')
try:
os.remove(output_file)
except OSError:
pass
# Convert to tf.train.Example and write the to TFRecords.
convert_to_tfrecord(input_files, output_file)
print('Done!')
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument(
'--data-dir',
type=str,
default='',
help='Directory to download and extract CIFAR-10 to.')
args = parser.parse_args()
main(args.data_dir)
tutorials/image/cifar10_estimator/model_base.py deleted 100644 → 0
View file @ bf6d6f6f
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""ResNet model.
Related papers:
https://arxiv.org/pdf/1603.05027v2.pdf
https://arxiv.org/pdf/1512.03385v1.pdf
https://arxiv.org/pdf/1605.07146v1.pdf
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import tensorflow as tf
class ResNet(object):
"""ResNet model."""
def __init__(self, is_training, data_format, batch_norm_decay, batch_norm_epsilon):
"""ResNet constructor.
Args:
is_training: if build training or inference model.
data_format: the data_format used during computation.
one of 'channels_first' or 'channels_last'.
"""
self._batch_norm_decay = batch_norm_decay
self._batch_norm_epsilon = batch_norm_epsilon
self._is_training = is_training
assert data_format in ('channels_first', 'channels_last')
self._data_format = data_format
def forward_pass(self, x):
raise NotImplementedError(
'forward_pass() is implemented in ResNet sub classes')
def _residual_v1(self,
x,
kernel_size,
in_filter,
out_filter,
stride,
activate_before_residual=False):
"""Residual unit with 2 sub layers, using Plan A for shortcut connection."""
del activate_before_residual
with tf.name_scope('residual_v1') as name_scope:
orig_x = x
x = self._conv(x, kernel_size, out_filter, stride)
x = self._batch_norm(x)
x = self._relu(x)
x = self._conv(x, kernel_size, out_filter, 1)
x = self._batch_norm(x)
if in_filter != out_filter:
orig_x = self._avg_pool(orig_x, stride, stride)
pad = (out_filter - in_filter) // 2
if self._data_format == 'channels_first':
orig_x = tf.pad(orig_x, [[0, 0], [pad, pad], [0, 0], [0, 0]])
else:
orig_x = tf.pad(orig_x, [[0, 0], [0, 0], [0, 0], [pad, pad]])
x = self._relu(tf.add(x, orig_x))
tf.logging.info('image after unit %s: %s', name_scope, x.get_shape())
return x
def _residual_v2(self,
x,
in_filter,
out_filter,
stride,
activate_before_residual=False):
"""Residual unit with 2 sub layers with preactivation, plan A shortcut."""
with tf.name_scope('residual_v2') as name_scope:
if activate_before_residual:
x = self._batch_norm(x)
x = self._relu(x)
orig_x = x
else:
orig_x = x
x = self._batch_norm(x)
x = self._relu(x)
x = self._conv(x, 3, out_filter, stride)
x = self._batch_norm(x)
x = self._relu(x)
x = self._conv(x, 3, out_filter, [1, 1, 1, 1])
if in_filter != out_filter:
pad = (out_filter - in_filter) // 2
orig_x = self._avg_pool(orig_x, stride, stride)
if self._data_format == 'channels_first':
orig_x = tf.pad(orig_x, [[0, 0], [pad, pad], [0, 0], [0, 0]])
else:
orig_x = tf.pad(orig_x, [[0, 0], [0, 0], [0, 0], [pad, pad]])
x = tf.add(x, orig_x)
tf.logging.info('image after unit %s: %s', name_scope, x.get_shape())
return x
def _bottleneck_residual_v2(self,
x,
in_filter,
out_filter,
stride,
activate_before_residual=False):
"""Bottleneck residual unit with 3 sub layers, plan B shortcut."""
with tf.name_scope('bottle_residual_v2') as name_scope:
if activate_before_residual:
x = self._batch_norm(x)
x = self._relu(x)
orig_x = x
else:
orig_x = x
x = self._batch_norm(x)
x = self._relu(x)
x = self._conv(x, 1, out_filter // 4, stride, is_atrous=True)
x = self._batch_norm(x)
x = self._relu(x)
# pad when stride isn't unit
x = self._conv(x, 3, out_filter // 4, 1, is_atrous=True)
x = self._batch_norm(x)
x = self._relu(x)
x = self._conv(x, 1, out_filter, 1, is_atrous=True)
if in_filter != out_filter:
orig_x = self._conv(orig_x, 1, out_filter, stride, is_atrous=True)
x = tf.add(x, orig_x)
tf.logging.info('image after unit %s: %s', name_scope, x.get_shape())
return x
def _conv(self, x, kernel_size, filters, strides, is_atrous=False):
"""Convolution."""
padding = 'SAME'
if not is_atrous and strides > 1:
pad = kernel_size - 1
pad_beg = pad // 2
pad_end = pad - pad_beg
if self._data_format == 'channels_first':
x = tf.pad(x, [[0, 0], [0, 0], [pad_beg, pad_end], [pad_beg, pad_end]])
else:
x = tf.pad(x, [[0, 0], [pad_beg, pad_end], [pad_beg, pad_end], [0, 0]])
padding = 'VALID'
return tf.layers.conv2d(
inputs=x,
kernel_size=kernel_size,
filters=filters,
strides=strides,
padding=padding,
use_bias=False,
data_format=self._data_format)
def _batch_norm(self, x):
if self._data_format == 'channels_first':
data_format = 'NCHW'
else:
data_format = 'NHWC'
return tf.contrib.layers.batch_norm(
x,
decay=self._batch_norm_decay,
center=True,
scale=True,
epsilon=self._batch_norm_epsilon,
is_training=self._is_training,
fused=True,
data_format=data_format)
def _relu(self, x):
return tf.nn.relu(x)
def _fully_connected(self, x, out_dim):
with tf.name_scope('fully_connected') as name_scope:
x = tf.layers.dense(x, out_dim)
tf.logging.info('image after unit %s: %s', name_scope, x.get_shape())
return x
def _avg_pool(self, x, pool_size, stride):
with tf.name_scope('avg_pool') as name_scope:
x = tf.layers.average_pooling2d(
x, pool_size, stride, 'SAME', data_format=self._data_format)
tf.logging.info('image after unit %s: %s', name_scope, x.get_shape())
return x
def _global_avg_pool(self, x):
with tf.name_scope('global_avg_pool') as name_scope:
assert x.get_shape().ndims == 4
if self._data_format == 'channels_first':
x = tf.reduce_mean(x, [2, 3])
else:
x = tf.reduce_mean(x, [1, 2])
tf.logging.info('image after unit %s: %s', name_scope, x.get_shape())
return x
tutorials/image/imagenet/BUILD deleted 100644 → 0
View file @ bf6d6f6f
# Description:
# Example TensorFlow models for ImageNet.
licenses(["notice"]) # Apache 2.0
exports_files(["LICENSE"])
py_binary(
name = "classify_image",
srcs = [
"classify_image.py",
],
srcs_version = "PY2AND3",
visibility = ["//tensorflow:__subpackages__"],
deps = [
"//tensorflow:tensorflow_py",
],
)
filegroup(
name = "all_files",
srcs = glob(
["**/*"],
exclude = [
"**/METADATA",
"**/OWNERS",
],
),
visibility = ["//tensorflow:__subpackages__"],
)
tutorials/image/imagenet/classify_image.py deleted 100644 → 0
View file @ bf6d6f6f
# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Simple image classification with Inception.
Run image classification with Inception trained on ImageNet 2012 Challenge data
set.
This program creates a graph from a saved GraphDef protocol buffer,
and runs inference on an input JPEG image. It outputs human readable
strings of the top 5 predictions along with their probabilities.
Change the --image_file argument to any jpg image to compute a
classification of that image.
Please see the tutorial and website for a detailed description of how
to use this script to perform image recognition.
https://tensorflow.org/tutorials/image_recognition/
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import argparse
import os.path
import re
import sys
import tarfile
import numpy as np
from six.moves import urllib
import tensorflow as tf
FLAGS = None
# pylint: disable=line-too-long
DATA_URL = 'http://download.tensorflow.org/models/image/imagenet/inception-2015-12-05.tgz'
# pylint: enable=line-too-long
class NodeLookup(object):
"""Converts integer node ID's to human readable labels."""
def __init__(self,
label_lookup_path=None,
uid_lookup_path=None):
if not label_lookup_path:
label_lookup_path = os.path.join(
FLAGS.model_dir, 'imagenet_2012_challenge_label_map_proto.pbtxt')
if not uid_lookup_path:
uid_lookup_path = os.path.join(
FLAGS.model_dir, 'imagenet_synset_to_human_label_map.txt')
self.node_lookup = self.load(label_lookup_path, uid_lookup_path)
def load(self, label_lookup_path, uid_lookup_path):
"""Loads a human readable English name for each softmax node.
Args:
label_lookup_path: string UID to integer node ID.
uid_lookup_path: string UID to human-readable string.
Returns:
dict from integer node ID to human-readable string.
"""
if not tf.gfile.Exists(uid_lookup_path):
tf.logging.fatal('File does not exist %s', uid_lookup_path)
if not tf.gfile.Exists(label_lookup_path):
tf.logging.fatal('File does not exist %s', label_lookup_path)
# Loads mapping from string UID to human-readable string
proto_as_ascii_lines = tf.gfile.GFile(uid_lookup_path).readlines()
uid_to_human = {}
p = re.compile(r'[n\d]*[ \S,]*')
for line in proto_as_ascii_lines:
parsed_items = p.findall(line)
uid = parsed_items[0]
human_string = parsed_items[2]
uid_to_human[uid] = human_string
# Loads mapping from string UID to integer node ID.
node_id_to_uid = {}
proto_as_ascii = tf.gfile.GFile(label_lookup_path).readlines()
for line in proto_as_ascii:
if line.startswith(' target_class:'):
target_class = int(line.split(': ')[1])
if line.startswith(' target_class_string:'):
target_class_string = line.split(': ')[1]
node_id_to_uid[target_class] = target_class_string[1:-2]
# Loads the final mapping of integer node ID to human-readable string
node_id_to_name = {}
for key, val in node_id_to_uid.items():
if val not in uid_to_human:
tf.logging.fatal('Failed to locate: %s', val)
name = uid_to_human[val]
node_id_to_name[key] = name
return node_id_to_name
def id_to_string(self, node_id):
if node_id not in self.node_lookup:
return ''
return self.node_lookup[node_id]
def create_graph():
"""Creates a graph from saved GraphDef file and returns a saver."""
# Creates graph from saved graph_def.pb.
with tf.gfile.FastGFile(os.path.join(
FLAGS.model_dir, 'classify_image_graph_def.pb'), 'rb') as f:
graph_def = tf.GraphDef()
graph_def.ParseFromString(f.read())
_ = tf.import_graph_def(graph_def, name='')
def run_inference_on_image(image):
"""Runs inference on an image.
Args:
image: Image file name.
Returns:
Nothing
"""
if not tf.gfile.Exists(image):
tf.logging.fatal('File does not exist %s', image)
image_data = tf.gfile.FastGFile(image, 'rb').read()
# Creates graph from saved GraphDef.
create_graph()
with tf.Session() as sess:
# Some useful tensors:
# 'softmax:0': A tensor containing the normalized prediction across
# 1000 labels.
# 'pool_3:0': A tensor containing the next-to-last layer containing 2048
# float description of the image.
# 'DecodeJpeg/contents:0': A tensor containing a string providing JPEG
# encoding of the image.
# Runs the softmax tensor by feeding the image_data as input to the graph.
softmax_tensor = sess.graph.get_tensor_by_name('softmax:0')
predictions = sess.run(softmax_tensor,
{'DecodeJpeg/contents:0': image_data})
predictions = np.squeeze(predictions)
# Creates node ID --> English string lookup.
node_lookup = NodeLookup()
top_k = predictions.argsort()[-FLAGS.num_top_predictions:][::-1]
for node_id in top_k:
human_string = node_lookup.id_to_string(node_id)
score = predictions[node_id]
print('%s (score = %.5f)' % (human_string, score))
def maybe_download_and_extract():
"""Download and extract model tar file."""
dest_directory = FLAGS.model_dir
if not os.path.exists(dest_directory):
os.makedirs(dest_directory)
filename = DATA_URL.split('/')[-1]
filepath = os.path.join(dest_directory, filename)
if not os.path.exists(filepath):
def _progress(count, block_size, total_size):
sys.stdout.write('\r>> Downloading %s %.1f%%' % (
filename, float(count * block_size) / float(total_size) * 100.0))
sys.stdout.flush()
filepath, _ = urllib.request.urlretrieve(DATA_URL, filepath, _progress)
print()
statinfo = os.stat(filepath)
print('Successfully downloaded', filename, statinfo.st_size, 'bytes.')
tarfile.open(filepath, 'r:gz').extractall(dest_directory)
def main(_):
maybe_download_and_extract()
image = (FLAGS.image_file if FLAGS.image_file else
os.path.join(FLAGS.model_dir, 'cropped_panda.jpg'))
run_inference_on_image(image)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
# classify_image_graph_def.pb:
# Binary representation of the GraphDef protocol buffer.
# imagenet_synset_to_human_label_map.txt:
# Map from synset ID to a human readable string.
# imagenet_2012_challenge_label_map_proto.pbtxt:
# Text representation of a protocol buffer mapping a label to synset ID.
parser.add_argument(
'--model_dir',
type=str,
default='/tmp/imagenet',
help="""\
Path to classify_image_graph_def.pb,
imagenet_synset_to_human_label_map.txt, and
imagenet_2012_challenge_label_map_proto.pbtxt.\
"""
)
parser.add_argument(
'--image_file',
type=str,
default='',
help='Absolute path to image file.'
)
parser.add_argument(
'--num_top_predictions',
type=int,
default=5,
help='Display this many predictions.'
)
FLAGS, unparsed = parser.parse_known_args()
tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)
tutorials/image/mnist/BUILD deleted 100644 → 0
View file @ bf6d6f6f
# Description:
# Example TensorFlow models for MNIST that achieves high accuracy
licenses(["notice"]) # Apache 2.0
exports_files(["LICENSE"])
py_binary(
name = "convolutional",
srcs = [
"convolutional.py",
],
srcs_version = "PY2AND3",
visibility = ["//tensorflow:__subpackages__"],
deps = ["//tensorflow:tensorflow_py"],
)
py_test(
name = "convolutional_test",
size = "medium",
srcs = [
"convolutional.py",
],
args = [
"--self_test",
],
main = "convolutional.py",
srcs_version = "PY2AND3",
deps = ["//tensorflow:tensorflow_py"],
)
filegroup(
name = "all_files",
srcs = glob(
["**/*"],
exclude = [
"**/METADATA",
"**/OWNERS",
],
),
visibility = ["//tensorflow:__subpackages__"],
)
tutorials/image/mnist/convolutional.py deleted 100644 → 0
View file @ bf6d6f6f
# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Simple, end-to-end, LeNet-5-like convolutional MNIST model example.
This should achieve a test error of 0.7%. Please keep this model as simple and
linear as possible, it is meant as a tutorial for simple convolutional models.
Run with --self_test on the command line to execute a short self-test.
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import argparse
import gzip
import os
import sys
import time
import numpy
from six.moves import urllib
from six.moves import xrange # pylint: disable=redefined-builtin
import tensorflow as tf
# CVDF mirror of http://yann.lecun.com/exdb/mnist/
SOURCE_URL = 'https://storage.googleapis.com/cvdf-datasets/mnist/'
WORK_DIRECTORY = 'data'
IMAGE_SIZE = 28
NUM_CHANNELS = 1
PIXEL_DEPTH = 255
NUM_LABELS = 10
VALIDATION_SIZE = 5000 # Size of the validation set.
SEED = 66478 # Set to None for random seed.
BATCH_SIZE = 64
NUM_EPOCHS = 10
EVAL_BATCH_SIZE = 64
EVAL_FREQUENCY = 100 # Number of steps between evaluations.
FLAGS = None
def data_type():
"""Return the type of the activations, weights, and placeholder variables."""
if FLAGS.use_fp16:
return tf.float16
else:
return tf.float32
def maybe_download(filename):
"""Download the data from Yann's website, unless it's already here."""
if not tf.gfile.Exists(WORK_DIRECTORY):
tf.gfile.MakeDirs(WORK_DIRECTORY)
filepath = os.path.join(WORK_DIRECTORY, filename)
if not tf.gfile.Exists(filepath):
filepath, _ = urllib.request.urlretrieve(SOURCE_URL + filename, filepath)
with tf.gfile.GFile(filepath) as f:
size = f.size()
print('Successfully downloaded', filename, size, 'bytes.')
return filepath
def extract_data(filename, num_images):
"""Extract the images into a 4D tensor [image index, y, x, channels].
Values are rescaled from [0, 255] down to [-0.5, 0.5].
"""
print('Extracting', filename)
with gzip.open(filename) as bytestream:
bytestream.read(16)
buf = bytestream.read(IMAGE_SIZE * IMAGE_SIZE * num_images * NUM_CHANNELS)
data = numpy.frombuffer(buf, dtype=numpy.uint8).astype(numpy.float32)
data = (data - (PIXEL_DEPTH / 2.0)) / PIXEL_DEPTH
data = data.reshape(num_images, IMAGE_SIZE, IMAGE_SIZE, NUM_CHANNELS)
return data
def extract_labels(filename, num_images):
"""Extract the labels into a vector of int64 label IDs."""
print('Extracting', filename)
with gzip.open(filename) as bytestream:
bytestream.read(8)
buf = bytestream.read(1 * num_images)
labels = numpy.frombuffer(buf, dtype=numpy.uint8).astype(numpy.int64)
return labels
def fake_data(num_images):
"""Generate a fake dataset that matches the dimensions of MNIST."""
data = numpy.ndarray(
shape=(num_images, IMAGE_SIZE, IMAGE_SIZE, NUM_CHANNELS),
dtype=numpy.float32)
labels = numpy.zeros(shape=(num_images,), dtype=numpy.int64)
for image in xrange(num_images):
label = image % 2
data[image, :, :, 0] = label - 0.5
labels[image] = label
return data, labels
def error_rate(predictions, labels):
"""Return the error rate based on dense predictions and sparse labels."""
return 100.0 - (
100.0 *
numpy.sum(numpy.argmax(predictions, 1) == labels) /
predictions.shape[0])
def main(_):
if FLAGS.self_test:
print('Running self-test.')
train_data, train_labels = fake_data(256)
validation_data, validation_labels = fake_data(EVAL_BATCH_SIZE)
test_data, test_labels = fake_data(EVAL_BATCH_SIZE)
num_epochs = 1
else:
# Get the data.
train_data_filename = maybe_download('train-images-idx3-ubyte.gz')
train_labels_filename = maybe_download('train-labels-idx1-ubyte.gz')
test_data_filename = maybe_download('t10k-images-idx3-ubyte.gz')
test_labels_filename = maybe_download('t10k-labels-idx1-ubyte.gz')
# Extract it into numpy arrays.
train_data = extract_data(train_data_filename, 60000)
train_labels = extract_labels(train_labels_filename, 60000)
test_data = extract_data(test_data_filename, 10000)
test_labels = extract_labels(test_labels_filename, 10000)
# Generate a validation set.
validation_data = train_data[:VALIDATION_SIZE, ...]
validation_labels = train_labels[:VALIDATION_SIZE]
train_data = train_data[VALIDATION_SIZE:, ...]
train_labels = train_labels[VALIDATION_SIZE:]
num_epochs = NUM_EPOCHS
train_size = train_labels.shape[0]
# This is where training samples and labels are fed to the graph.
# These placeholder nodes will be fed a batch of training data at each
# training step using the {feed_dict} argument to the Run() call below.
train_data_node = tf.placeholder(
data_type(),
shape=(BATCH_SIZE, IMAGE_SIZE, IMAGE_SIZE, NUM_CHANNELS))
train_labels_node = tf.placeholder(tf.int64, shape=(BATCH_SIZE,))
eval_data = tf.placeholder(
data_type(),
shape=(EVAL_BATCH_SIZE, IMAGE_SIZE, IMAGE_SIZE, NUM_CHANNELS))
# The variables below hold all the trainable weights. They are passed an
# initial value which will be assigned when we call:
# {tf.global_variables_initializer().run()}
conv1_weights = tf.Variable(
tf.truncated_normal([5, 5, NUM_CHANNELS, 32], # 5x5 filter, depth 32.
stddev=0.1,
seed=SEED, dtype=data_type()))
conv1_biases = tf.Variable(tf.zeros([32], dtype=data_type()))
conv2_weights = tf.Variable(tf.truncated_normal(
[5, 5, 32, 64], stddev=0.1,
seed=SEED, dtype=data_type()))
conv2_biases = tf.Variable(tf.constant(0.1, shape=[64], dtype=data_type()))
fc1_weights = tf.Variable( # fully connected, depth 512.
tf.truncated_normal([IMAGE_SIZE // 4 * IMAGE_SIZE // 4 * 64, 512],
stddev=0.1,
seed=SEED,
dtype=data_type()))
fc1_biases = tf.Variable(tf.constant(0.1, shape=[512], dtype=data_type()))
fc2_weights = tf.Variable(tf.truncated_normal([512, NUM_LABELS],
stddev=0.1,
seed=SEED,
dtype=data_type()))
fc2_biases = tf.Variable(tf.constant(
0.1, shape=[NUM_LABELS], dtype=data_type()))
# We will replicate the model structure for the training subgraph, as well
# as the evaluation subgraphs, while sharing the trainable parameters.
def model(data, train=False):
"""The Model definition."""
# 2D convolution, with 'SAME' padding (i.e. the output feature map has
# the same size as the input). Note that {strides} is a 4D array whose
# shape matches the data layout: [image index, y, x, depth].
conv = tf.nn.conv2d(data,
conv1_weights,
strides=[1, 1, 1, 1],
padding='SAME')
# Bias and rectified linear non-linearity.
relu = tf.nn.relu(tf.nn.bias_add(conv, conv1_biases))
# Max pooling. The kernel size spec {ksize} also follows the layout of
# the data. Here we have a pooling window of 2, and a stride of 2.
pool = tf.nn.max_pool(relu,
ksize=[1, 2, 2, 1],
strides=[1, 2, 2, 1],
padding='SAME')
conv = tf.nn.conv2d(pool,
conv2_weights,
strides=[1, 1, 1, 1],
padding='SAME')
relu = tf.nn.relu(tf.nn.bias_add(conv, conv2_biases))
pool = tf.nn.max_pool(relu,
ksize=[1, 2, 2, 1],
strides=[1, 2, 2, 1],
padding='SAME')
# Reshape the feature map cuboid into a 2D matrix to feed it to the
# fully connected layers.
pool_shape = pool.get_shape().as_list()
reshape = tf.reshape(
pool,
[pool_shape[0], pool_shape[1] * pool_shape[2] * pool_shape[3]])
# Fully connected layer. Note that the '+' operation automatically
# broadcasts the biases.
hidden = tf.nn.relu(tf.matmul(reshape, fc1_weights) + fc1_biases)
# Add a 50% dropout during training only. Dropout also scales
# activations such that no rescaling is needed at evaluation time.
if train:
hidden = tf.nn.dropout(hidden, 0.5, seed=SEED)
return tf.matmul(hidden, fc2_weights) + fc2_biases
# Training computation: logits + cross-entropy loss.
logits = model(train_data_node, True)
loss = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(
labels=train_labels_node, logits=logits))
# L2 regularization for the fully connected parameters.
regularizers = (tf.nn.l2_loss(fc1_weights) + tf.nn.l2_loss(fc1_biases) +
tf.nn.l2_loss(fc2_weights) + tf.nn.l2_loss(fc2_biases))
# Add the regularization term to the loss.
loss += 5e-4 * regularizers
# Optimizer: set up a variable that's incremented once per batch and
# controls the learning rate decay.
batch = tf.Variable(0, dtype=data_type())
# Decay once per epoch, using an exponential schedule starting at 0.01.
learning_rate = tf.train.exponential_decay(
0.01, # Base learning rate.
batch * BATCH_SIZE, # Current index into the dataset.
train_size, # Decay step.
0.95, # Decay rate.
staircase=True)
# Use simple momentum for the optimization.
optimizer = tf.train.MomentumOptimizer(learning_rate,
0.9).minimize(loss,
global_step=batch)
# Predictions for the current training minibatch.
train_prediction = tf.nn.softmax(logits)
# Predictions for the test and validation, which we'll compute less often.
eval_prediction = tf.nn.softmax(model(eval_data))
# Small utility function to evaluate a dataset by feeding batches of data to
# {eval_data} and pulling the results from {eval_predictions}.
# Saves memory and enables this to run on smaller GPUs.
def eval_in_batches(data, sess):
"""Get all predictions for a dataset by running it in small batches."""
size = data.shape[0]
if size < EVAL_BATCH_SIZE:
raise ValueError("batch size for evals larger than dataset: %d" % size)
predictions = numpy.ndarray(shape=(size, NUM_LABELS), dtype=numpy.float32)
for begin in xrange(0, size, EVAL_BATCH_SIZE):
end = begin + EVAL_BATCH_SIZE
if end <= size:
predictions[begin:end, :] = sess.run(
eval_prediction,
feed_dict={eval_data: data[begin:end, ...]})
else:
batch_predictions = sess.run(
eval_prediction,
feed_dict={eval_data: data[-EVAL_BATCH_SIZE:, ...]})
predictions[begin:, :] = batch_predictions[begin - size:, :]
return predictions
# Create a local session to run the training.
start_time = time.time()
with tf.Session() as sess:
# Run all the initializers to prepare the trainable parameters.
tf.global_variables_initializer().run()
print('Initialized!')
# Loop through training steps.
for step in xrange(int(num_epochs * train_size) // BATCH_SIZE):
# Compute the offset of the current minibatch in the data.
# Note that we could use better randomization across epochs.
offset = (step * BATCH_SIZE) % (train_size - BATCH_SIZE)
batch_data = train_data[offset:(offset + BATCH_SIZE), ...]
batch_labels = train_labels[offset:(offset + BATCH_SIZE)]
# This dictionary maps the batch data (as a numpy array) to the
# node in the graph it should be fed to.
feed_dict = {train_data_node: batch_data,
train_labels_node: batch_labels}
# Run the optimizer to update weights.
sess.run(optimizer, feed_dict=feed_dict)
# print some extra information once reach the evaluation frequency
if step % EVAL_FREQUENCY == 0:
# fetch some extra nodes' data
l, lr, predictions = sess.run([loss, learning_rate, train_prediction],
feed_dict=feed_dict)
elapsed_time = time.time() - start_time
start_time = time.time()
print('Step %d (epoch %.2f), %.1f ms' %
(step, float(step) * BATCH_SIZE / train_size,
1000 * elapsed_time / EVAL_FREQUENCY))
print('Minibatch loss: %.3f, learning rate: %.6f' % (l, lr))
print('Minibatch error: %.1f%%' % error_rate(predictions, batch_labels))
print('Validation error: %.1f%%' % error_rate(
eval_in_batches(validation_data, sess), validation_labels))
sys.stdout.flush()
# Finally print the result!
test_error = error_rate(eval_in_batches(test_data, sess), test_labels)
print('Test error: %.1f%%' % test_error)
if FLAGS.self_test:
print('test_error', test_error)
assert test_error == 0.0, 'expected 0.0 test_error, got %.2f' % (
test_error,)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument(
'--use_fp16',
default=False,
help='Use half floats instead of full floats if True.',
action='store_true')
parser.add_argument(
'--self_test',
default=False,
action='store_true',
help='True if running a self test.')
FLAGS, unparsed = parser.parse_known_args()
tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)
tutorials/rnn/BUILD deleted 100644 → 0
View file @ bf6d6f6f
# Description:
# Example RNN models, including language models and sequence-to-sequence models.
package(default_visibility = ["//visibility:public"])
licenses(["notice"]) # Apache 2.0
exports_files(["LICENSE"])
py_library(
name = "linear",
srcs = [
"linear.py",
],
srcs_version = "PY2AND3",
deps = [
"//tensorflow:tensorflow_py",
],
)
py_library(
name = "rnn_cell",
srcs = [
"rnn_cell.py",
],
srcs_version = "PY2AND3",
deps = [
":linear",
"//tensorflow:tensorflow_py",
],
)
py_library(
name = "package",
srcs = [
"__init__.py",
],
srcs_version = "PY2AND3",
deps = [
":rnn",
":rnn_cell",
":seq2seq",
],
)
py_library(
name = "rnn",
srcs = [
"rnn.py",
],
srcs_version = "PY2AND3",
deps = [
":rnn_cell",
"//tensorflow:tensorflow_py",
],
)
py_library(
name = "seq2seq",
srcs = [
"seq2seq.py",
],
srcs_version = "PY2AND3",
deps = [
":rnn",
"//tensorflow:tensorflow_py",
],
)
filegroup(
name = "all_files",
srcs = glob(
["**/*"],
exclude = [
"**/METADATA",
"**/OWNERS",
],
),
visibility = ["//tensorflow:__subpackages__"],
)
tutorials/rnn/README.md deleted 100644 → 0
View file @ bf6d6f6f
This directory contains functions for creating recurrent neural networks
and sequence-to-sequence models. Detailed instructions on how to get started
and use them are available in the
[tutorials on tensorflow.org](http://tensorflow.org/tutorials/).
Here is a short overview of what is in this directory:
File | What's in it?
------------ | -------------
`ptb/` | PTB language model, see the [RNN Tutorial](http://tensorflow.org/tutorials/recurrent/)
`quickdraw/` | Quick, Draw! model, see the [RNN Tutorial for Drawing Classification](https://www.tensorflow.org/versions/master/tutorials/recurrent_quickdraw)
If you're looking for the
[`seq2seq` tutorial code](http://tensorflow.org/tutorials/seq2seq/), it lives
in [its own repo](https://github.com/tensorflow/nmt).
\ No newline at end of file
tutorials/rnn/__init__.py deleted 100644 → 0
View file @ bf6d6f6f
# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Libraries to build Recurrent Neural Networks."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
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