How can I run a Tensorflow file on GPUs?

Question

Do you have any idea how can I run "eval_image_classifier.py" on GPUs? Should I change any functions or do any modifications? Or whether there exist any other specific functions for evaluation on GPUs?

I can already run "train_image_classifier.py" on GPUs because of having the associated flag for switching between CPU and GPU:

tf.app.flags.DEFINE_boolean('clone_on_cpu', False,
                        'Use CPUs to deploy clones.')

I did try to add the same line to eval_image_classifier.py , but it had no effect. I'm using Python 2.7.13 and Tensorflow 1.3.0 .

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import math
import tensorflow as tf
from deployment import model_deploy
from datasets import dataset_factory
from nets import nets_factory
from preprocessing import preprocessing_factory
slim = tf.contrib.slim


tf.app.flags.DEFINE_integer(
    'batch_size', 32, 'The number of samples in each batch.')

tf.app.flags.DEFINE_integer(
    'max_num_batches', None,
    'Max number of batches to evaluate by default use all.')

tf.app.flags.DEFINE_string(
    'master', '', 'The address of the TensorFlow master to use.')

tf.app.flags.DEFINE_string(
    'checkpoint_path', '...',
    'The directory where the model was written to or an absolute path to a '
    'checkpoint file.')

tf.app.flags.DEFINE_string(
    'eval_dir', '...',
    'Directory where the results are saved to.')

tf.app.flags.DEFINE_integer('num_clones', 1,
                            'Number of model clones to deploy.')

tf.app.flags.DEFINE_boolean('clone_on_cpu', False,
                            'Use CPUs to deploy clones.')

tf.app.flags.DEFINE_integer('worker_replicas', 1, 'Number of worker replicas.')

tf.app.flags.DEFINE_integer(
    'num_readers', 4,
    'The number of parallel readers that read data from the dataset.')

tf.app.flags.DEFINE_integer(
    'num_ps_tasks', 0,
    'The number of parameter servers. If the value is 0, then the parameters '
    'are handled locally by the worker.')


tf.app.flags.DEFINE_integer(
    'num_preprocessing_threads', 4,
    'The number of threads used to create the batches.')

tf.app.flags.DEFINE_string(
    'dataset_name', '...', 'The name of the dataset to load.')

tf.app.flags.DEFINE_string(
    'dataset_split_name', 'validation', 'The name of the train/test split.')

tf.app.flags.DEFINE_string(
    'dataset_dir', '...', 
    'The directory where the dataset files are stored.')

tf.app.flags.DEFINE_integer(
    'labels_offset', 0,
    'An offset for the labels in the dataset. This flag is primarily used to '
    'evaluate the VGG and ResNet architectures which do not use a background '
    'class for the ImageNet dataset.')

tf.app.flags.DEFINE_string(
    'model_name', 'densenet161', 'The name of the architecture to evaluate.')

tf.app.flags.DEFINE_string(
    'preprocessing_name', None, 'The name of the preprocessing to use. If left '
    'as `None`, then the model_name flag is used.')

tf.app.flags.DEFINE_float(
    'moving_average_decay', None,
    'The decay to use for the moving average.'
    'If left as None, then moving averages are not used.')

tf.app.flags.DEFINE_integer(
    'eval_image_size', None, 'Eval image size')

FLAGS = tf.app.flags.FLAGS


def main(_):
  if not FLAGS.dataset_dir:
    raise ValueError('You must supply the dataset directory with --dataset_dir')

    #######################
    # Config model_deploy #
    #######################
  tf.logging.set_verbosity(tf.logging.INFO)
  with tf.Graph().as_default():

    deploy_config = model_deploy.DeploymentConfig(
      num_clones=FLAGS.num_clones,
      clone_on_cpu=FLAGS.clone_on_cpu,
      #replica_id=FLAGS.task,
      num_replicas=FLAGS.worker_replicas,
      num_ps_tasks=FLAGS.num_ps_tasks)

    # Create global_step
    with tf.device(deploy_config.variables_device()):
      tf_global_step = slim.create_global_step()
    ######################
    # Select the dataset #
    ######################
    dataset = dataset_factory.get_dataset(
        FLAGS.dataset_name, FLAGS.dataset_split_name, FLAGS.dataset_dir)

    ####################
    # Select the model #
    ####################
    network_fn = nets_factory.get_network_fn(
        FLAGS.model_name,
        num_classes=(dataset.num_classes - FLAGS.labels_offset),
        is_training=False)

    ##############################################################
    # Create a dataset provider that loads data from the dataset #
    ##############################################################
    with tf.device(deploy_config.inputs_device()):
        provider = slim.dataset_data_provider.DatasetDataProvider(
            dataset,
            num_readers=FLAGS.num_readers,
            shuffle=False,
            common_queue_capacity=2 * FLAGS.batch_size,
            common_queue_min=FLAGS.batch_size)
        [image, label] = provider.get(['image', 'label'])
        label -= FLAGS.labels_offset

    #####################################
    # Select the preprocessing function #
    #####################################
    preprocessing_name = FLAGS.preprocessing_name or FLAGS.model_name
    image_preprocessing_fn = preprocessing_factory.get_preprocessing(
        preprocessing_name,
        is_training=False)

    eval_image_size = FLAGS.eval_image_size or network_fn.default_image_size

    image = image_preprocessing_fn(image, eval_image_size, eval_image_size)

    images, labels = tf.train.batch(
        [image, label],
        batch_size=FLAGS.batch_size,
        num_threads=FLAGS.num_preprocessing_threads,
        capacity=5 * FLAGS.batch_size)
    batch_queue = slim.prefetch_queue.prefetch_queue(
        [images, labels], capacity=2 * deploy_config.num_clones)
    ####################
    # Define the model #
    ####################
    def clone_fn(batch_queue):
      """Allows data parallelism by creating multiple clones of network_fn."""
      with tf.device(deploy_config.inputs_device()):
        images, labels = batch_queue.dequeue()
      logits, end_points = network_fn(images)
      logits = tf.squeeze(logits)

      #############################
      # Specify the loss function #
      #############################
      if 'AuxLogits' in end_points:
        tf.losses.mean_squared_error(
            predictions=end_points['AuxLogits'], labels=labels, weights=0.4, scope='aux_loss')
      tf.losses.mean_squared_error(
          predictions=logits, labels=labels, weights=1.0)
      return end_points

    #clones = model_deploy.create_clones(deploy_config, clone_fn, [batch_queue])
    #first_clone_scope = deploy_config.clone_scope(0)
    ####################
    # Define the model #
    ####################
    logits, _ = network_fn(images)

    if FLAGS.moving_average_decay:
      variable_averages = tf.train.ExponentialMovingAverage(
          FLAGS.moving_average_decay, tf_global_step)
      variables_to_restore = variable_averages.variables_to_restore(
          slim.get_model_variables())
      variables_to_restore[tf_global_step.op.name] = tf_global_step
    else:
      variables_to_restore = slim.get_variables_to_restore()

    logits = tf.squeeze(logits)

    # Define the metrics:
    predictions = logits

    names_to_values, names_to_updates = slim.metrics.aggregate_metric_map({
        'Accuracy': tf.metrics.root_mean_squared_error(predictions, labels),
        'Recall_5': slim.metrics.streaming_recall(
            logits, labels),
    })

    # Print the summaries to screen.
    print_ops = []
    summary_ops = []
    for name, value in names_to_values.items():
      summary_name = 'eval/%s' % name
      op = tf.summary.scalar(summary_name, value, collections=[])
      op = tf.Print(op, [value], summary_name)
      summary_ops.append(op)
      print_ops.append(tf.Print(value, [value], summary_name))
      tf.add_to_collection(tf.GraphKeys.SUMMARIES, op)

    # TODO(sguada) use num_epochs=1
    if FLAGS.max_num_batches:
      num_batches = FLAGS.max_num_batches
    else:
      # This ensures that we make a single pass over all of the data.
      num_batches = math.ceil(dataset.num_samples / float(FLAGS.batch_size))

    if tf.gfile.IsDirectory(FLAGS.checkpoint_path):
        if tf.train.latest_checkpoint(FLAGS.checkpoint_path):
            checkpoint_path = tf.train.latest_checkpoint(FLAGS.checkpoint_path)
        else:
            checkpoint_path = FLAGS.checkpoint_path

    eval_interval_secs = 6

    tf.logging.info('Evaluating %s' % checkpoint_path)

    slim.evaluation.evaluation_loop(
        master=FLAGS.master,
        checkpoint_dir=checkpoint_path,
        logdir=FLAGS.eval_dir,
        num_evals=num_batches,
        eval_op=list(names_to_updates.values()) + print_ops,
        variables_to_restore=variables_to_restore,
        eval_interval_secs = eval_interval_secs )
if __name__ == '__main__':
  tf.app.run()

I tried to use some code like Tensorflow tutorial as well:

# Creates a graph.
with tf.device('/gpu:2'):
a = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[2, 3], name='a')
b = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[3, 2], name='b')
c = tf.matmul(a, b)
# Creates a session with allow_soft_placement and log_device_placement set to True.
sess = tf.Session(config=tf.ConfigProto(
         allow_soft_placement=True, log_device_placement=True))
# Runs the op.
print(sess.run(c))

I modified the code in this way:

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import math
import tensorflow as tf

from datasets import dataset_factory
from nets import nets_factory
from preprocessing import preprocessing_factory

slim = tf.contrib.slim

tf.app.flags.DEFINE_integer(
    'batch_size', 32, 'The number of samples in each batch.')

tf.app.flags.DEFINE_integer(
    'max_num_batches', None,
    'Max number of batches to evaluate by default use all.')

tf.app.flags.DEFINE_string(
    'master', '', 'The address of the TensorFlow master to use.')

tf.app.flags.DEFINE_string(
    'checkpoint_path', '...',
    'The directory where the model was written to or an absolute path to a '
    'checkpoint file.')

tf.app.flags.DEFINE_string(
    'eval_dir', '...',
    'Directory where the results are saved to.')

tf.app.flags.DEFINE_integer(
    'num_preprocessing_threads', 4,
    'The number of threads used to create the batches.')

tf.app.flags.DEFINE_string(
    'dataset_name', '...', 'The name of the dataset to load.')

tf.app.flags.DEFINE_string(
    'dataset_split_name', 'validation', 'The name of the train/test split.')

tf.app.flags.DEFINE_string(
    'dataset_dir', '...', 
    'The directory where the dataset files are stored.')

tf.app.flags.DEFINE_integer(
    'labels_offset', 0,
    'An offset for the labels in the dataset. This flag is primarily used to '
    'evaluate the VGG and ResNet architectures which do not use a background '
    'class for the ImageNet dataset.')

tf.app.flags.DEFINE_string(
    'model_name', 'densenet161', 'The name of the architecture to evaluate.')

tf.app.flags.DEFINE_string(
    'preprocessing_name', None, 'The name of the preprocessing to use. If left '
    'as `None`, then the model_name flag is used.')

tf.app.flags.DEFINE_float(
    'moving_average_decay', None,
    'The decay to use for the moving average.'
    'If left as None, then moving averages are not used.')

tf.app.flags.DEFINE_integer(
    'eval_image_size', None, 'Eval image size')

FLAGS = tf.app.flags.FLAGS

# Initialize all global and local variables
init = tf.group(tf.global_variables_initializer(),
                   tf.local_variables_initializer())

def main(_):
  if not FLAGS.dataset_dir:
    raise ValueError('You must supply the dataset directory with --dataset_dir')

  tf.logging.set_verbosity(tf.logging.INFO)

  sess = tf.Session(config=tf.ConfigProto(allow_soft_placement=True,
                                          log_device_placement=True))

  with tf.Graph().as_default(), tf.device('/gpu:0'):

    sess.run(init)
    tf_global_step = slim.get_or_create_global_step()

    ######################
    # Select the dataset #
    ######################
    dataset = dataset_factory.get_dataset(
        FLAGS.dataset_name, FLAGS.dataset_split_name, FLAGS.dataset_dir)

    ####################
    # Select the model #
    ####################
    network_fn = nets_factory.get_network_fn(
        FLAGS.model_name,
        num_classes=(dataset.num_classes - FLAGS.labels_offset),
        is_training=False)

    ##############################################################
    # Create a dataset provider that loads data from the dataset #
    ##############################################################
    provider = slim.dataset_data_provider.DatasetDataProvider(
        dataset,
        shuffle=False,
        common_queue_capacity=2 * FLAGS.batch_size,
        common_queue_min=FLAGS.batch_size)
    [image, label] = provider.get(['image', 'label'])
    label -= FLAGS.labels_offset

    #####################################
    # Select the preprocessing function #
    #####################################
    preprocessing_name = FLAGS.preprocessing_name or FLAGS.model_name
    image_preprocessing_fn = preprocessing_factory.get_preprocessing(
        preprocessing_name,
        is_training=False)

    eval_image_size = FLAGS.eval_image_size or network_fn.default_image_size

    image = image_preprocessing_fn(image, eval_image_size, eval_image_size)

    images, labels = tf.train.batch(
        [image, label],
        batch_size=FLAGS.batch_size,
        num_threads=FLAGS.num_preprocessing_threads,
        capacity=5 * FLAGS.batch_size)

    ####################
    # Define the model #
    ####################
    logits, _ = network_fn(images)

    if FLAGS.moving_average_decay:
      variable_averages = tf.train.ExponentialMovingAverage(
          FLAGS.moving_average_decay, tf_global_step)
      variables_to_restore = variable_averages.variables_to_restore(
          slim.get_model_variables())
      variables_to_restore[tf_global_step.op.name] = tf_global_step
    else:
      variables_to_restore = slim.get_variables_to_restore()

    logits = tf.squeeze(logits)

    # Define the metrics:
    predictions = logits

    names_to_values, names_to_updates = slim.metrics.aggregate_metric_map({
        'Accuracy': tf.metrics.root_mean_squared_error(predictions, labels),
        'Recall_5': slim.metrics.streaming_recall(
            logits, labels),
    })

    # Print the summaries to screen.
    print_ops = []
    summary_ops = []
    for name, value in names_to_values.items():
      summary_name = 'eval/%s' % name
      op = tf.summary.scalar(summary_name, value, collections=[])
      op = tf.Print(op, [value], summary_name)
      summary_ops.append(op)
      print_ops.append(tf.Print(value, [value], summary_name))
      tf.add_to_collection(tf.GraphKeys.SUMMARIES, op)

    # TODO(sguada) use num_epochs=1
    if FLAGS.max_num_batches:
      num_batches = FLAGS.max_num_batches
    else:
      # This ensures that we make a single pass over all of the data.
      num_batches = math.ceil(dataset.num_samples / float(FLAGS.batch_size))

    if tf.gfile.IsDirectory(FLAGS.checkpoint_path):
        if tf.train.latest_checkpoint(FLAGS.checkpoint_path):
            checkpoint_path = tf.train.latest_checkpoint(FLAGS.checkpoint_path)
        else:
            checkpoint_path = FLAGS.checkpoint_path

    #print(checkpoint_path)
    eval_interval_secs = 6

    tf.logging.info('Evaluating %s' % checkpoint_path)

    slim.evaluation.evaluation_loop(
        master=FLAGS.master,
        checkpoint_dir=checkpoint_path,
        logdir=FLAGS.eval_dir,
        num_evals=num_batches,
        eval_op=list(names_to_updates.values()) + print_ops,
        variables_to_restore=variables_to_restore,
        eval_interval_secs = eval_interval_secs )


if __name__ == '__main__':
  tf.app.run()

When I run this code, I face this error:

    Traceback (most recent call last):
  File "/home/zgholami/test1/GZ_Project/GZ_DenseNet_TF-slim/eval_image_classifier.py", line 210, in <module>
    tf.app.run()
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/platform/app.py", line 48, in run
    _sys.exit(main(_sys.argv[:1] + flags_passthrough))
  File "/home/zgholami/test1/GZ_Project/GZ_DenseNet_TF-slim/eval_image_classifier.py", line 206, in main
    eval_interval_secs = 60 )
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/contrib/slim/python/slim/evaluation.py", line 296, in evaluation_loo
p
    timeout=timeout)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/contrib/training/python/training/evaluation.py", line 447, in evalua
te_repeatedly
    session_creator=session_creator, hooks=hooks) as session:
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/training/monitored_session.py", line 668, in __init__
    stop_grace_period_secs=stop_grace_period_secs)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/training/monitored_session.py", line 490, in __init__
    self._sess = _RecoverableSession(self._coordinated_creator)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/training/monitored_session.py", line 842, in __init__
    _WrappedSession.__init__(self, self._create_session())
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/training/monitored_session.py", line 847, in _create_session
    return self._sess_creator.create_session()
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/training/monitored_session.py", line 551, in create_session
    self.tf_sess = self._session_creator.create_session()
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/training/monitored_session.py", line 425, in create_session
    init_fn=self._scaffold.init_fn)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/training/session_manager.py", line 273, in prepare_session
    config=config)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/training/session_manager.py", line 189, in _restore_checkpoin
t
    saver.restore(sess, checkpoint_filename_with_path)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/training/saver.py", line 1560, in restore
    {self.saver_def.filename_tensor_name: save_path})
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 895, in run
    run_metadata_ptr)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1124, in _run
    feed_dict_tensor, options, run_metadata)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1321, in _do_run
    options, run_metadata)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1340, in _do_call
    raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.InvalidArgumentError: Cannot assign a device for operation 'eval_step': Could not satisfy explicit device s
pecification '/device:GPU:0' because no supported kernel for GPU devices is available.
Colocation Debug Info:
Colocation group had the following types and devices: 
Const: GPU CPU 
AssignAdd: CPU 
VariableV2: CPU 
Identity: GPU CPU 
Assign: CPU 
IsVariableInitialized: CPU 
     [[Node: eval_step = VariableV2[_class=["loc:@eval_step"], container="", dtype=DT_INT64, shape=[], shared_name="", _device="/device:GPU:0"]
()]]

Caused by op u'eval_step', defined at:
  File "/home/zgholami/test1/GZ_Project/GZ_DenseNet_TF-slim/eval_image_classifier.py", line 210, in <module>
    tf.app.run()
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/platform/app.py", line 48, in run
    _sys.exit(main(_sys.argv[:1] + flags_passthrough))
  File "/home/zgholami/test1/GZ_Project/GZ_DenseNet_TF-slim/eval_image_classifier.py", line 206, in main
    eval_interval_secs = 60 )
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/contrib/slim/python/slim/evaluation.py", line 296, in evaluation_loo
p
    timeout=timeout)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/contrib/training/python/training/evaluation.py", line 410, in evalua
te_repeatedly
    eval_step = get_or_create_eval_step()
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/training/evaluation.py", line 57, in _get_or_create_eval_step
    collections=[ops.GraphKeys.LOCAL_VARIABLES, ops.GraphKeys.EVAL_STEP])
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 1065, in get_variable
    use_resource=use_resource, custom_getter=custom_getter)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 962, in get_variable
    use_resource=use_resource, custom_getter=custom_getter)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 367, in get_variable
    validate_shape=validate_shape, use_resource=use_resource)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 352, in _true_getter
    use_resource=use_resource)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/ops/variable_scope.py", line 725, in _get_single_variable
    validate_shape=validate_shape)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/ops/variables.py", line 199, in __init__
    expected_shape=expected_shape)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/ops/variables.py", line 283, in _init_from_args
    name=name)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/ops/state_ops.py", line 131, in variable_op_v2
    shared_name=shared_name)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/ops/gen_state_ops.py", line 682, in _variable_v2
    name=name)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 767, in apply_op
    op_def=op_def)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2630, in create_op
    original_op=self._default_original_op, op_def=op_def)
  File "/group/pawsey0245/zgholami/pyml/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1204, in __init__
    self._traceback = self._graph._extract_stack()  # pylint: disable=protected-access

InvalidArgumentError (see above for traceback): Cannot assign a device for operation 'eval_step': Could not satisfy explicit device specification '
/device:GPU:0' because no supported kernel for GPU devices is available.
Colocation Debug Info:
Colocation group had the following types and devices: 
Const: GPU CPU 
AssignAdd: CPU 
VariableV2: CPU 
Identity: GPU CPU 
Assign: CPU 
IsVariableInitialized: CPU 
     [[Node: eval_step = VariableV2[_class=["loc:@eval_step"], container="", dtype=DT_INT64, shape=[], shared_name="", _device="/device:GPU:0"]
()]]

ERROR:tensorflow:==================================
Object was never used (type <class 'tensorflow.python.framework.ops.Tensor'>):
<tf.Tensor 'report_uninitialized_variables_1/boolean_mask/Gather:0' shape=(?,) dtype=string>
If you want to mark it as used call its "mark_used()" method.

Answer 1

您尚未检查TensorFlow的文档： 使用GPU

Answer 2

If you've installed TF with GPU support, which I assume you've as you've trained it using train_image_classifier.py , then eval_image_classifier.py also used GPUs as default.

I've just ran default eval_image_classifier.py from models/research/slim and it is using GPUs to do evaluation. So, you don't have to modify anything to use GPUs to do eval.