Skip to content

GitLab

Commit 441c8f40 authored by qianyj's avatar qianyj
Browse files

update TF code

parent ec90ad8e
Hide whitespace changes
Inline Side-by-side
Showing with 2346 additions and 0 deletions
TensorFlow/ComputeVision/Accuracy_Validation/benchmarks-master/scripts/tf_cnn_benchmarks/test_data/tfrecord_image_generator.py 0 → 100644
View file @ 441c8f40
# 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.
# ==============================================================================
"""Generate black and white test TFRecords with Example protos.
Each record within the TFRecord file is a
serialized Example proto. The Example proto contains the following fields:
image/encoded: string containing JPEG encoded image in RGB colorspace
image/height: integer, image height in pixels
image/width: integer, image width in pixels
image/colorspace: string, specifying the colorspace, always 'RGB'
image/channels: integer, specifying the number of channels, always 3
image/format: string, specifying the format, always'JPEG'
image/filename: string containing the basename of the image file
e.g. 'n01440764_10026.JPEG' or 'ILSVRC2012_val_00000293.JPEG'
image/class/label: integer specifying the index in a classification layer.
The label ranges from [1, 1000] where 0 is not used.
image/class/synset: string specifying the unique ID of the label,
e.g. 'n01440764'
image/class/text: string specifying the human-readable version of the label
e.g. 'red fox, Vulpes vulpes'
image/object/bbox/xmin: list of integers specifying the 0+ human annotated
bounding boxes
image/object/bbox/xmax: list of integers specifying the 0+ human annotated
bounding boxes
image/object/bbox/ymin: list of integers specifying the 0+ human annotated
bounding boxes
image/object/bbox/ymax: list of integers specifying the 0+ human annotated
bounding boxes
image/object/bbox/label: integer specifying the index in a classification
layer. The label ranges from [1, 1000] where 0 is not used. Note this is
always identical to the image label.
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import os
import random
import numpy as np
import six
import tensorflow.compat.v1 as tf
def _int64_feature(value):
"""Wrapper for inserting int64 features into Example proto."""
if not isinstance(value, list):
value = [value]
return tf.train.Feature(int64_list=tf.train.Int64List(value=value))
def _float_feature(value):
"""Wrapper for inserting float features into Example proto."""
if not isinstance(value, list):
value = [value]
return tf.train.Feature(float_list=tf.train.FloatList(value=value))
def _bytes_feature(value):
"""Wrapper for inserting bytes features into Example proto."""
return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value]))
def _convert_to_example(filename, image_buffer, label, synset, human, bbox,
height, width):
"""Build an Example proto for an example.
Args:
filename: string, path to an image file, e.g., '/path/to/example.JPG'
image_buffer: bytes, JPEG encoding of RGB image
label: integer, identifier for the ground truth for the network
synset: string, unique WordNet ID specifying the label, e.g., 'n02323233'
human: string, human-readable label, e.g., 'red fox, Vulpes vulpes'
bbox: list of bounding boxes; each box is a list of integers
specifying [xmin, ymin, xmax, ymax]. All boxes are assumed to belong to
the same label as the image label.
height: integer, image height in pixels
width: integer, image width in pixels
Returns:
Example proto
"""
xmin = []
ymin = []
xmax = []
ymax = []
for b in bbox:
assert len(b) == 4
# pylint: disable=expression-not-assigned
[l.append(point) for l, point in zip([xmin, ymin, xmax, ymax], b)]
# pylint: enable=expression-not-assigned
colorspace = b'RGB'
channels = 3
image_format = b'JPEG'
example = tf.train.Example(features=tf.train.Features(feature={
'image/height': _int64_feature(height),
'image/width': _int64_feature(width),
'image/colorspace': _bytes_feature(colorspace),
'image/channels': _int64_feature(channels),
'image/class/label': _int64_feature(label),
'image/class/synset': _bytes_feature(six.ensure_binary(synset)),
'image/class/text': _bytes_feature(six.ensure_binary(human)),
'image/object/bbox/xmin': _float_feature(xmin),
'image/object/bbox/xmax': _float_feature(xmax),
'image/object/bbox/ymin': _float_feature(ymin),
'image/object/bbox/ymax': _float_feature(ymax),
'image/object/bbox/label': _int64_feature([label] * len(xmin)),
'image/format': _bytes_feature(image_format),
'image/filename': _bytes_feature(os.path.basename(six.ensure_binary(
filename))),
'image/encoded': _bytes_feature(image_buffer)}))
return example
class ImageCoder(object):
"""Helper class that provides TensorFlow image coding utilities."""
def __init__(self):
# Create a single Session to run all image coding calls.
self._sess = tf.Session()
# Initializes function that converts PNG to JPEG data.
self._image = tf.placeholder(dtype=tf.uint8)
self._encode_jpeg = tf.image.encode_jpeg(
self._image, format='rgb', quality=100)
def encode_jpeg(self, image):
jpeg_image = self._sess.run(self._encode_jpeg,
feed_dict={self._image: image})
return jpeg_image
def _process_image(coder, name):
"""Process a single image file.
If name is "train", a black image is returned. Otherwise, a white image is
returned.
Args:
coder: instance of ImageCoder to provide TensorFlow image coding utils.
name: string, unique identifier specifying the data set.
Returns:
image_buffer: bytes, JPEG encoding of RGB image.
height: integer, image height in pixels.
width: integer, image width in pixels.
"""
# Read the image file.
value = 0 if name == 'train' else 255
height = random.randint(30, 299)
width = random.randint(30, 299)
image = np.full((height, width, 3), value, np.uint8)
jpeg_data = coder.encode_jpeg(image)
return jpeg_data, height, width
def _process_dataset(output_directory, num_classes, coder, name, num_images,
num_shards):
"""Process a complete data set and save it as a TFRecord.
Args:
output_directory: Where to put outputs.
num_classes: number of classes.
coder: Instance of an ImageCoder.
name: string, unique identifier specifying the data set.
num_images: number of images to generate.
num_shards: integer number of shards to create.
"""
files_per_shard = num_images // num_shards
for shard in range(num_shards):
output_filename = '%s-%.5d-of-%.5d' % (name, shard, num_shards)
output_file = os.path.join(output_directory, output_filename)
with tf.python_io.TFRecordWriter(output_file) as writer:
for i in range(files_per_shard):
index = shard * files_per_shard + i
image_buffer, height, width = _process_image(coder, name)
filename = '{}_{}_{}'.format(name, shard, i)
label = index % num_classes
synset = str(index)
human = name
bbox = [[0.1, 0.1, 0.9, 0.9]]
example = _convert_to_example(filename, image_buffer, label,
synset, human, bbox,
height, width)
writer.write(example.SerializeToString())
def write_black_and_white_tfrecord_data(
output_directory, num_classes, num_train_images=512,
num_validation_images=128, train_shards=8, validation_shards=2):
"""Writes black and white images in tfrecord format.
Training images are black and validation images are white.
Args:
output_directory: Where to put outputs.
num_classes: number of classes.
num_train_images: number of training images to generate.
num_validation_images: number of validation images to generate.
train_shards: integer number of training shards to create.
validation_shards: integer number of validation shards to create.
"""
coder = ImageCoder()
_process_dataset(output_directory, num_classes, coder, 'validation',
num_validation_images, validation_shards)
_process_dataset(output_directory, num_classes, coder, 'train',
num_train_images, train_shards)
TensorFlow/ComputeVision/Accuracy_Validation/benchmarks-master/scripts/tf_cnn_benchmarks/test_util.py 0 → 100644
View file @ 441c8f40
# 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.
# ==============================================================================
"""Shared functionality across multiple test files."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from collections import namedtuple
from contextlib import contextmanager
import os
import numpy as np
import tensorflow.compat.v1 as tf
import benchmark_cnn
import cnn_util
import datasets
import preprocessing
from models import model
from platforms import util as platforms_util
from test_data import tfrecord_image_generator
from tensorflow.core.protobuf import rewriter_config_pb2 # pylint: disable=g-direct-tensorflow-import
from tensorflow.python.platform import test
@contextmanager
def monkey_patch(obj, **kwargs):
"""Context mgr to monkey patch attributes on an object (such as a module).
The attributes are patched back to their original value when the context
manager exits.
For example, to replace benchmark_cnn.get_data_type with an identity function,
do:
```
with monkey_patch(benchmark_cnn, get_data_type=lambda x: x)
loss1 = benchmark_cnn.loss_function(1) # loss1 will be 1
loss2 = benchmark_cnn.loss_function(params) # Call the original function
```
Args:
obj: The object (which can be a module) to monkey patch attributes on.
**kwargs: Dictionary mapping from attribute name to value that the attribute
will be patched with.
Yields:
Nothing.
"""
old_values = {key: getattr(obj, key) for key in kwargs}
try:
for key, value in kwargs.items():
setattr(obj, key, value)
yield
finally:
for key, value in old_values.items():
setattr(obj, key, value)
def monkey_patch_base_cluster_manager():
"""Monkey patches get_cluster_manager to return a BaseClusterManager.
This function replaces platforms_util.get_cluster_manager with a function that
always return a BaseClusterManager.
This is useful for testing creating a graph in distributed mode, with only a
single process. GrpcClusterManager's constructor blocks until a cluster is set
up, which requires multiple processes to be created.
"""
def get_test_cluster_manager(params, config_proto):
del config_proto
return cnn_util.BaseClusterManager(params)
platforms_util.get_cluster_manager = get_test_cluster_manager
def print_and_add_to_list(print_list):
"""Returns a function which prints the input, then adds it to print_list."""
def f(string):
print(string)
print_list.append(string)
return f
TrainingOutput = namedtuple('TrainingOutput',
['loss', 'top_1_accuracy', 'top_5_accuracy'])
EvalOutput = namedtuple('EvalOutput', ['top_1_accuracy', 'top_5_accuracy'])
def get_training_outputs_from_logs(logs, print_training_accuracy):
"""Returns a list of TrainingOutputs by parsing the logs of a training run.
Args:
logs: A list of strings, each which is a line from the standard output of
tf_cnn_benchmarks from training. Only lines in the form:
10 images/sec: 14.2 +/- 0.0 (jitter = 0.0) 7.020
are parsed (the line may also contain the training accuracies).
print_training_accuracy: The value of the param print_training_accuracy.
Returns:
A list of TrainingOutputs. The list has one element per element of logs
that is in the format above. top_1_accuracy and top_5_accuracy are set to -1
if the line does not contain accuracies.
"""
outputs = []
for log in logs:
if 'images/sec' in log and '+/-' in log:
parts = log.split()
if print_training_accuracy:
# Example log with training accuracy:
# 10 images/sec: 0.2 +/- 0.0 (jitter = 0.0) 6.908 0.500 1.000
assert len(parts) == 11
top_1_acc = float(parts[9])
top_5_acc = float(parts[10])
else:
# Example log without training accuracy:
# 10 images/sec: 0.2 +/- 0.0 (jitter = 0.0) 6.908
assert len(parts) == 9
top_1_acc = -1
top_5_acc = -1
loss = float(parts[8])
outputs.append(TrainingOutput(loss=loss, top_1_accuracy=top_1_acc,
top_5_accuracy=top_5_acc))
assert len(outputs) >= 1
return outputs
def get_evaluation_outputs_from_logs(logs):
"""Returns the top 1 and 5 accuracies by parsing the logs of an eval run.
Args:
logs: A list of strings, each which is a line from the standard output of
tf_cnn_benchmarks from evaluation. Only lines in the form:
Accuracy @ 1 = 0.5000 Accuracy @ 5 = 1.0000 [80 examples]
is parsed.
Returns:
A list of EvalOutputs. Normally this list only has one EvalOutput, but can
contain multiple if training is done and
--eval_during_training_every_n_steps is specified.
"""
eval_outputs = []
for log in logs:
if 'Accuracy @ ' in log:
# Example log:
# Accuracy @ 1 = 0.5000 Accuracy @ 5 = 1.0000 [80 examples]
parts = log.split()
assert len(parts) == 12
top_1_accuracy = float(parts[4])
top_5_accuracy = float(parts[9])
eval_outputs.append(EvalOutput(top_1_accuracy, top_5_accuracy))
assert eval_outputs
return eval_outputs
def check_training_outputs_are_reasonable(testcase, training_outputs,
print_training_accuracy,
max_final_loss=10.,
previous_final_loss=None):
"""Checks the outputs from training a model are reasonable.
An assert is failed if the outputs are not reasonable. The final top-1 and
top-5 accuracies are asserted to be 1, and so the dataset used to train should
be trivial to learn. For example, the dataset could consist of a black image
with label 0 and a white image with label 1.
Args:
testcase: A tf.test.TestCase used for assertions.
training_outputs: A list of TrainingOutputs, as returned from
get_training_outputs_from_logs().
print_training_accuracy: Whether training accuracies were printed and stored
in training_outputs.
max_final_loss: The loss of the final training output is asserted to be at
most this value.
previous_final_loss: If training was resumed from a checkpoint, the loss of
the final step from the previous training run that saved the checkpoint.
"""
if previous_final_loss is not None:
# Ensure the loss hasn't raised significantly from the final loss of the
# previous training run.
testcase.assertLessEqual(training_outputs[0].loss,
previous_final_loss * 1.01)
for output in training_outputs:
testcase.assertLessEqual(output.loss, 100.)
last_output = training_outputs[-1]
if print_training_accuracy:
testcase.assertEqual(last_output.top_1_accuracy, 1.0)
testcase.assertEqual(last_output.top_5_accuracy, 1.0)
if max_final_loss is not None:
testcase.assertLessEqual(last_output.loss, max_final_loss)
def train_and_eval(testcase,
run_fn,
params,
check_output_values,
max_final_loss=10.,
skip=None):
"""Trains a model then evaluates it.
This function should be used to verify training and evaluating
BenchmarkCNN works without crashing and that it outputs reasonable
values. BenchmarkCNN will be run three times. First, it will train a
model from scratch, saving a checkpoint. Second, it will load the checkpoint
to continue training. Finally, it evaluates based on the loaded checkpoint.
Args:
testcase: A tf.test.TestCase used for assertions.
run_fn: Must run `BenchmarkCNN` exactly once. BenchmarkCNN is
never used directly, but instead is only run through `run_fn`. `run_fn`
has the signature (run_type, inner_params) -> output_list, where:
* run_type is a string indicating how BenchmarkCNN will be run.
Either 'InitialTraining', 'TrainingFromCheckpoint' or 'Evaluation'.
* inner_params is the params BenchmarkCNN should be run with.
* output_list[i] is a list of lines from the ith worker's stdout.
params: The params BenchmarkCNN will be run with.
Will be passed to `run_fn` slightly modified in order to run with both
training and evaluation.
check_output_values: Whether the outputs of the workers, such as training
accuracy, should be checked to make sure their values are reasonable.
Fails an assert on `testcase` if a check fails.
max_final_loss: The loss of the final training output is asserted to be at
most this value for both training runs.
skip: If 'eval', evaluation is not done. if
'eval_and_train_from_checkpoint', evaluation and training from a
checkpoint are both not done.
"""
assert not skip or skip in {'eval', 'eval_and_train_from_checkpoint'}
# Part 1: Train from scratch.
tf.logging.info('Training model from scratch')
print_training_accuracy = (params.print_training_accuracy or
params.forward_only)
initial_train_logs = run_fn('InitialTraining', params)
testcase.assertGreaterEqual(len(initial_train_logs), 1)
for lines in initial_train_logs:
initial_train_outputs = get_training_outputs_from_logs(
lines, print_training_accuracy)
if params.cross_replica_sync and params.batch_group_size == 1:
testcase.assertEqual(len(initial_train_outputs), params.num_batches)
if check_output_values:
check_training_outputs_are_reasonable(testcase, initial_train_outputs,
print_training_accuracy,
max_final_loss=max_final_loss)
if params.train_dir is not None:
train_dir_entries = set(os.listdir(params.train_dir))
testcase.assertGreater(len(train_dir_entries), 0)
else:
train_dir_entries = None
if skip == 'eval_and_train_from_checkpoint':
return
# Part 2: Train from the loaded checkpoint.
testcase.assertIsNotNone(train_dir_entries)
tf.logging.info('Training model from loaded checkpoint')
# Run for same number of batches as before.
params = params._replace(num_batches=params.num_batches * 2)
train_logs_from_ckpt = run_fn('TrainingFromCheckpoint', params)
testcase.assertGreaterEqual(len(train_logs_from_ckpt), 1)
for lines in train_logs_from_ckpt:
train_outputs_from_ckpt = get_training_outputs_from_logs(
lines, print_training_accuracy)
if params.cross_replica_sync and params.batch_group_size == 1:
testcase.assertEqual(len(train_outputs_from_ckpt),
params.num_batches // 2 - params.num_warmup_batches)
if check_output_values:
check_training_outputs_are_reasonable(
testcase, train_outputs_from_ckpt, print_training_accuracy,
max_final_loss=max_final_loss,
previous_final_loss=initial_train_outputs[-1].loss)
# Ensure a new checkpoint was written out.
testcase.assertNotEqual(train_dir_entries, set(os.listdir(params.train_dir)))
if skip == 'eval':
return
# Part 3: Evaluate from the loaded checkpoint.
tf.logging.info('Evaluating model from checkpoint')
params = params._replace(num_batches=params.num_batches // 2, eval=True)
eval_logs = run_fn('Evaluation', params)
testcase.assertGreaterEqual(len(eval_logs), 1)
for lines in eval_logs:
eval_outputs = get_evaluation_outputs_from_logs(lines)
assert len(eval_outputs) == 1
top_1_accuracy, top_5_accuracy = eval_outputs[0]
if check_output_values:
testcase.assertEqual(top_1_accuracy, 1.0)
testcase.assertEqual(top_5_accuracy, 1.0)
def get_temp_dir(dir_name):
dir_path = os.path.join(test.get_temp_dir(), dir_name)
os.mkdir(dir_path)
return dir_path
def create_black_and_white_images():
dir_path = get_temp_dir('black_and_white_images')
tfrecord_image_generator.write_black_and_white_tfrecord_data(dir_path,
num_classes=1)
return dir_path
def get_params(train_dir_name):
"""Returns params that can be used to train."""
params = benchmark_cnn.make_params(
batch_size=2,
display_every=1,
init_learning_rate=0.005,
model='trivial',
num_batches=20,
num_gpus=2,
num_warmup_batches=5,
optimizer='sgd',
print_training_accuracy=True,
train_dir=get_temp_dir(train_dir_name),
variable_update='parameter_server',
weight_decay=0,
distortions=True,
distort_color_in_yiq=False)
return benchmark_cnn.set_default_param_values_and_env_vars(params)
def get_var_update_params():
"""Returns params that are used when testing variable updates."""
params = benchmark_cnn.make_params(
batch_size=2,
model='test_model',
num_gpus=2,
display_every=1,
num_warmup_batches=0,
num_batches=4,
weight_decay=2 ** -4,
init_learning_rate=2 ** -4,
optimizer='sgd')
return benchmark_cnn.set_default_param_values_and_env_vars(params)
def get_fake_var_update_inputs():
"""Returns fake input 1x1 images to use in variable update tests."""
# BenchmarkCNN divides by 127.5 then subtracts 1.0 from the images, so after
# that, the images will be -1., 0., 1., ..., 14.
return np.resize(127.5 * np.array(range(16)), (16, 1, 1, 1))
def _worker_batches_in_numpy_array(numpy_inputs, batch_size, shift_ratio):
"""Yields batches from a numpy array, for a single worker."""
numpy_inputs = cnn_util.roll_numpy_batches(numpy_inputs, batch_size,
shift_ratio)
i = 0
total_batches = numpy_inputs.shape[0]
assert total_batches % batch_size == 0
while True:
yield numpy_inputs[i:i + batch_size, ...]
i = (i + batch_size) % total_batches
def manually_compute_losses(numpy_inputs, inputs_placeholder, loss, num_workers,
params):
"""Manually compute the losses each worker should report in tf_cnn_benchmarks.
This function essentially simulates tf_cnn_benchmarks, computing what the loss
of each worker should be. The caller should create a model, that takes in
images from `inputs_placeholder`, a tf.placeholder, and computes `loss`.
This function, and all ops passed to this function, must be run under a
tf.device('cpu:0') context manager.
Non-SGD optimizers are not supported with multiple workers.
Args:
numpy_inputs: A Numpy array to use as the input images.
inputs_placeholder: A tf.placeholder tensor, where input images can be fed
into.
loss: A scalar tensor representing the loss of the model, which is obtained
from the input images in inputs_placeholder.
num_workers: How many workers should be simulated.
params: Params tuple. This doesn't have to have information about the
distributed cluster, such as --num_workers, as num_workers is passed in
separately.
Returns:
A list of list of losses. return_value[i][j] is the loss of the ith worker
after the jth step.
"""
batch_size = params.batch_size * params.num_gpus
assert numpy_inputs.shape[0] % (num_workers * batch_size) == 0
l2_loss = tf.add_n([tf.nn.l2_loss(x) for x in tf.trainable_variables()])
total_loss = loss + params.weight_decay * l2_loss
reported_loss = (loss if params.loss_type_to_report == 'base_loss'
else total_loss)
gradient_multiplier = 1
if params.variable_update in ('replicated', 'distributed_all_reduce'):
# In certain variable updates, tf_cnn_benchmarks add the gradients of the
# GPUs instead of taking their mean, making the gradients effectively
# params.num_gpu times higher.
# TODO(b/62722498): Make all variable updates consistent.
gradient_multiplier = params.num_gpus
opt = benchmark_cnn.get_optimizer(params, params.init_learning_rate)
grad_vars = opt.compute_gradients(
total_loss, grad_loss=tf.constant(gradient_multiplier, dtype=tf.float32))
grads = [g for g, _ in grad_vars]
# We apply gradients from a placeholder. That way, we can first compute the
# gradients from each worker, then afterwards apply them one by one by feeding
# them into the placeholder.
placeholder_grad_vars = [(tf.placeholder(g.dtype, g.shape), v)
for g, v in grad_vars]
placeholder_grads = [g for g, _ in placeholder_grad_vars]
apply_grads_op = opt.apply_gradients(placeholder_grad_vars)
batch_iterators = [_worker_batches_in_numpy_array(numpy_inputs, batch_size,
shift_ratio=i / num_workers)
for i in range(num_workers)]
# Set the GPU count to 0, to avoid taking all the GPU memory. Unfortunately,
# doing so still takes up about ~1GB for some reason.
config = tf.ConfigProto(device_count={'GPU': 0})
config.graph_options.rewrite_options.pin_to_host_optimization = (
rewriter_config_pb2.RewriterConfig.OFF)
with tf.Session(config=config) as sess:
sess.run(tf.global_variables_initializer())
losses = [[] for _ in range(num_workers)]
for i in range(params.num_batches):
computed_grads = []
for j in range(num_workers):
batch_feed = next(batch_iterators[j])
batch_feed = batch_feed / 127.5 - 1
worker_loss, worker_grads = sess.run((reported_loss, grads),
{inputs_placeholder: batch_feed})
losses[j].append(worker_loss)
computed_grads.append(worker_grads)
for worker_grads in computed_grads:
# TODO(reedwm): With multiple workers, applying the gradients
# sequentially per worker is not equivalent to what tf_cnn_benchmarks
# does when the optmizer is not SGD. Therefore, this currently does not
# work currently when num_workers > 1 and params.optimizer != 'sgd'.
feed_dict = dict(zip(placeholder_grads, worker_grads))
sess.run(apply_grads_op, feed_dict)
return losses
class TestCNNModel(model.CNNModel):
"""A simple model used for testing.
The input is a 1-channel 1x1 image, consisting of a single number. The model
has two scalar variables: A and B, initialized to 1 and 2 respectively. Given
an image x, the loss is defined as:
loss = x * A * B
"""
def __init__(self):
super(TestCNNModel, self).__init__(
'test_cnn_model', image_size=1, batch_size=1, learning_rate=1)
self.depth = 1
VAR_A_INITIAL_VALUE = 1.
VAR_B_INITIAL_VALUE = 2.
def add_inference(self, cnn):
# This model only supports 1x1 images with 1 channel
assert cnn.top_layer.shape[1:] == (1, 1, 1)
# Multiply by variable A.
with tf.name_scope('mult_by_var_A'):
cnn.conv(1, 1, 1, 1, 1, use_batch_norm=None, activation=None, bias=None,
kernel_initializer=tf.constant_initializer(
self.VAR_A_INITIAL_VALUE))
# Multiply by variable B.
with tf.name_scope('mult_by_var_B'):
cnn.conv(1, 1, 1, 1, 1, use_batch_norm=None, activation=None, bias=None,
kernel_initializer=tf.constant_initializer(
self.VAR_B_INITIAL_VALUE))
with tf.name_scope('reshape_to_scalar'):
cnn.reshape([-1, 1])
def skip_final_affine_layer(self):
return True
def loss_function(self, inputs, build_network_result):
del inputs
return tf.reduce_mean(build_network_result.logits)
def manually_compute_losses(self, inputs, num_workers, params):
with tf.Graph().as_default(), tf.device('/cpu:0'):
a = tf.Variable(self.VAR_A_INITIAL_VALUE, name='A')
b = tf.Variable(self.VAR_B_INITIAL_VALUE, name='B')
inputs_placeholder = tf.placeholder(tf.float32,
(None, 1, 1, 1),
name='inputs_placeholder')
inputs_reshaped = tf.reshape(inputs_placeholder, (-1, 1))
loss = self.loss_function(
None,
model.BuildNetworkResult(logits=inputs_reshaped * a * b,
extra_info=None))
return manually_compute_losses(inputs, inputs_placeholder, loss,
num_workers, params)
def accuracy_function(self, inputs, logits):
del inputs
# Let the accuracy be the same as the loss function.
return {'top_1_accuracy': logits, 'top_5_accuracy': logits}
class TestDataSet(datasets.ImageDataset):
"""A Dataset consisting of 1x1 images with a depth of 1."""
def __init__(self, height=1, width=1, depth=1):
super(TestDataSet, self).__init__('test_dataset', height=height,
width=width, depth=depth, data_dir=None,
queue_runner_required=True, num_classes=1)
def num_examples_per_epoch(self, subset='train'):
del subset
return 1
def get_input_preprocessor(self, input_preprocessor='default'):
return preprocessing.TestImagePreprocessor
def use_synthetic_gpu_inputs(self):
return False
TensorFlow/ComputeVision/Accuracy_Validation/ResNet50_Official/official/utils/logs/hooks_helper_test.py → TensorFlow/ComputeVision/Accuracy_Validation/benchmarks-master/scripts/tf_cnn_benchmarks/tf_cnn_benchmarks.py
View file @ 441c8f40
......@@ -13,55 +13,61 @@
# limitations under the License.
# ==============================================================================
"""Tests for hooks_helper."""
"""Benchmark script for TensorFlow.
See the README for more information.
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import unittest
import tensorflow as tf # pylint: disable=g-bad-import-order
from absl import app
from absl import flags as absl_flags
import tensorflow.compat.v1 as tf
from official.utils.logs import hooks_helper
import benchmark_cnn
import cnn_util
import flags
import mlperf
from cnn_util import log_fn
class BaseTest(unittest.TestCase):
flags.define_flags()
for name in flags.param_specs.keys():
absl_flags.declare_key_flag(name)
def test_raise_in_non_list_names(self):
with self.assertRaises(ValueError):
hooks_helper.get_train_hooks(
'LoggingTensorHook, ProfilerHook', model_dir="", batch_size=256)
absl_flags.DEFINE_boolean(
'ml_perf_compliance_logging', False,
'Print logs required to be compliant with MLPerf. If set, must clone the '
'MLPerf training repo https://github.com/mlperf/training and add '
'https://github.com/mlperf/training/tree/master/compliance to the '
'PYTHONPATH')
def test_raise_in_invalid_names(self):
invalid_names = ['StepCounterHook', 'StopAtStepHook']
with self.assertRaises(ValueError):
hooks_helper.get_train_hooks(invalid_names, model_dir="", batch_size=256)
def validate_train_hook_name(self,
test_hook_name,
expected_hook_name,
**kwargs):
returned_hook = hooks_helper.get_train_hooks(
[test_hook_name], model_dir="", **kwargs)
self.assertEqual(len(returned_hook), 1)
self.assertIsInstance(returned_hook[0], tf.train.SessionRunHook)
self.assertEqual(returned_hook[0].__class__.__name__.lower(),
expected_hook_name)
def main(positional_arguments):
# Command-line arguments like '--distortions False' are equivalent to
# '--distortions=True False', where False is a positional argument. To prevent
# this from silently running with distortions, we do not allow positional
# arguments.
assert len(positional_arguments) >= 1
if len(positional_arguments) > 1:
raise ValueError('Received unknown positional arguments: %s'
% positional_arguments[1:])
def test_get_train_hooks_logging_tensor_hook(self):
self.validate_train_hook_name('LoggingTensorHook', 'loggingtensorhook')
params = benchmark_cnn.make_params_from_flags()
with mlperf.mlperf_logger(absl_flags.FLAGS.ml_perf_compliance_logging,
params.model):
params = benchmark_cnn.setup(params)
bench = benchmark_cnn.BenchmarkCNN(params)
def test_get_train_hooks_profiler_hook(self):
self.validate_train_hook_name('ProfilerHook', 'profilerhook')
tfversion = cnn_util.tensorflow_version_tuple()
log_fn('TensorFlow: %i.%i' % (tfversion[0], tfversion[1]))
def test_get_train_hooks_examples_per_second_hook(self):
self.validate_train_hook_name('ExamplesPerSecondHook',
'examplespersecondhook')
bench.print_info()
bench.run()
def test_get_logging_metric_hook(self):
test_hook_name = 'LoggingMetricHook'
self.validate_train_hook_name(test_hook_name, 'loggingmetrichook')
if __name__ == '__main__':
tf.test.main()
tf.disable_v2_behavior()
app.run(main) # Raises error on invalid flags, unlike tf.app.run()
TensorFlow/ComputeVision/Accuracy_Validation/benchmarks-master/scripts/tf_cnn_benchmarks/variable_mgr.py 0 → 100644
View file @ 441c8f40
# 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.
# ==============================================================================
"""Defines VariableMgr and subclasses used to manage variables.
"""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import contextlib
import re
import tensorflow.compat.v1 as tf
import allreduce
import batch_allreduce
import variable_mgr_util
class VariableMgr(object):
"""Abstract superclass for class used by BenchmarkCNN to control variables.
Functions on this class are used to control how variables are created and
managed, and how gradients are computed and applied.
"""
def __init__(self, benchmark_cnn):
self.benchmark_cnn = benchmark_cnn
self.staging_delta_ops = []
self.use_resource_vars = benchmark_cnn.params.use_resource_vars
# A variable for automatic loss scaling.
self.grad_has_inf_nan = None
self._reuse_vars = False
def each_tower_has_variables(self):
"""Returns True if each GPU tower of the model has separate variables."""
assert False, 'Must be implemented in subclass'
def supports_staged_vars(self):
"""Whether staged variable management is supported."""
return False
def create_outer_variable_scope(self, device_num):
"""Create the tf.variable_scope around all model graph operations."""
del device_num # unused by this implementation
assert False, 'Must be implemented in subclass'
def preprocess_device_grads(self, device_grads):
"""Preprocess the device gradients prior to applying them.
Args:
device_grads: List of lists of (gradient, variable) tuples.
device_grads[t][g] = (gradient, variable), where t is the index of the
tower and g is the index of the gradient-variable pair.
Returns: a tuple of (apply_gradients_devices, gradient_state).
gradient_state is an opaque structure that should be passed to
get_gradients_to_apply() and append_apply_gradients_ops() (in that order).
apply_gradients_devices is a list of devices where the gradients will be
applied with get_gradients_to_apply() and append_apply_gradients_ops().
"""
del device_grads # unused by this implementation
assert False, 'Must be implemented in subclass'
def get_gradients_to_apply(self, device_num, gradient_state):
"""Returns the [(gradient, variable)] list to apply for device_num.
Args:
device_num: indexes into apply_gradients_devices, which was returned by an
earlier call to preprocess_device_grads.
gradient_state: from previous call to apply_gradients_devices.
"""
del device_num, gradient_state # unused by this implementation
assert False, 'Must be implemented in subclass'
def append_apply_gradients_ops(self, gradient_state, opt, grads, training_ops,
loss_scale_params):
"""Adds training ops for grads to 'training_ops'.
Args:
gradient_state: from previous call to apply_gradients_devices.
opt: the underlying optimizer
grads: [(grad, var)] to apply
training_ops: list to which to add ops
loss_scale_params: parameters for loss scaling.
"""
del gradient_state # unused by this implementation
def get_apply_gradients_ops_func():
"""Returns the apply_gradients op."""
return [opt.apply_gradients(grads)]
variable_mgr_util.append_gradients_with_loss_scale(
training_ops, get_apply_gradients_ops_func, loss_scale_params,
self.grad_has_inf_nan)
def get_post_init_ops(self):
"""Returns ops that should run post-initialization."""
return []
def get_devices(self):
"""Returns devices to use for computation; includes replica selection."""
assert False, 'Must be implemented in subclass'
def savable_variables(self):
"""Returns a list/dict of savable variables to pass to tf.train.Saver."""
return tf.global_variables()
def trainable_variables_on_device(self,
rel_device_num,
abs_device_num,
writable=False):
"""Return the set of trainable variables on device.
Args:
rel_device_num: local worker device index.
abs_device_num: global graph device index.
writable: whether to get a reference to the underlying variable.
Returns:
The set of trainable variables on the specified device.
"""
del rel_device_num, writable
if self.each_tower_has_variables():
params = [
v for v in tf.trainable_variables()
if v.name.startswith('v%s/' % abs_device_num)
]
else:
params = tf.trainable_variables()
return params
@contextlib.contextmanager
def reuse_variables(self):
"""Context manager that causes variables requested to be reused.
Variables requested under this context manager must already exist, and will
be reused instead of being created again. This should be used if the
evaluation model is being built after the training model has already been
built. This is because the evaluation model should reuse variables from the
training model.
Yields:
Nothing.
"""
old_reuse_vars = self._reuse_vars
try:
self._reuse_vars = True
yield
finally:
self._reuse_vars = old_reuse_vars
class VariableMgrIndependent(VariableMgr):
"""VariableMgr that implements the --independent mode for local jobs.
Each GPU has its own copy of the variables, and gradients are
not shared between towers. This can be used to check
performance when no data is moved between GPUs.
"""
def each_tower_has_variables(self):
return True
def create_outer_variable_scope(self, device_num):
return tf.variable_scope('v%s' % device_num, reuse=self._reuse_vars,
use_resource=self.use_resource_vars)
def preprocess_device_grads(self, device_grads):
return (self.benchmark_cnn.devices, device_grads)
def get_gradients_to_apply(self, device_num, gradient_state):
device_grads = gradient_state
tower_grad = device_grads[device_num]
if self.benchmark_cnn.enable_auto_loss_scale and device_num == 0:
# Since we don't aggregate variables in --independent mode, we cannot tell
# if there are NaNs on all GPUs. So we arbitrarily choose to only check
# NaNs on the first GPU.
has_inf_nan_list = []
for grad, _ in tower_grad:
has_inf_nan_list.append(tf.reduce_all(tf.is_finite(grad)))
self.grad_has_inf_nan = tf.logical_not(tf.reduce_all(has_inf_nan_list))
return tower_grad
def get_devices(self):
return self.benchmark_cnn.raw_devices
class VariableMgrLocalFetchFromPS(VariableMgr):
"""VariableMgr that implements the --parameter_server mode for local jobs.
Variables are stored on a parameter server. For each step, each tower gets
a copy of the variables from the parameter server, and sends its gradients
to the param server.
"""
def each_tower_has_variables(self):
return False
def create_outer_variable_scope(self, device_num):
return tf.variable_scope('v', reuse=bool(device_num) or self._reuse_vars,
use_resource=self.use_resource_vars)
def preprocess_device_grads(self, device_grads):
return ([self.benchmark_cnn.param_server_device], device_grads)
def get_gradients_to_apply(self, device_num, gradient_state):
assert device_num == 0
device_grads = gradient_state
agg_grads, self.grad_has_inf_nan = (
variable_mgr_util.
aggregate_gradients_using_copy_with_variable_colocation(
device_grads,
use_mean=True,
check_inf_nan=self.benchmark_cnn.enable_auto_loss_scale))
return agg_grads
def get_devices(self):
raw_devices = self.benchmark_cnn.raw_devices
if self.benchmark_cnn.local_parameter_device_flag == 'gpu':
return [
variable_mgr_util.ParamServerDeviceSetter(d, raw_devices)
for d in raw_devices
]
else:
return [
tf.train.replica_device_setter(
worker_device=d,
ps_device=self.benchmark_cnn.param_server_device,
ps_tasks=1) for d in raw_devices
]
class VariableMgrLocalFetchFromStagedPS(VariableMgrLocalFetchFromPS):
"""Implements fetching a local variable through staging buffers.
"""
def __init__(self, benchmark_cnn):
super(VariableMgrLocalFetchFromStagedPS, self).__init__(benchmark_cnn)
# A data structure to track where the variables are used on each device.
# Indexed by device_num and var_name, each entry stores the "put" and "get"
# ops used for that variable on that device:
# staging_vars_on_devices[device_num][var_name] == (put_op, get_op)
self.staging_vars_on_devices = [
dict() for _ in self.benchmark_cnn.raw_devices
]
def supports_staged_vars(self):
return True
def create_outer_variable_scope(self, device_num):
self._custom_getter = variable_mgr_util.StagedVariableGetter(
device_num, self.benchmark_cnn.raw_devices, None, self)
return tf.variable_scope(
'v', reuse=bool(device_num) or self._reuse_vars,
custom_getter=self._custom_getter, use_resource=self.use_resource_vars)
def trainable_variables_on_device(self,
rel_device_num,
abs_device_num,
writable=False):
return self._custom_getter.trainable_variables_on_device(
rel_device_num, abs_device_num, writable=writable)
class VariableMgrLocalReplicated(VariableMgr):
"""VariableMgr that implements the --replicated mode for local jobs.
Each GPU has its own copy of the variables. To apply gradients,
either a local all-reduce algorithm is applied or a regular
cross-device aggregation is used to replicate the combined
gradients to all towers.
"""
def __init__(self, benchmark_cnn, all_reduce_spec,
agg_small_grads_max_bytes, agg_small_grads_max_group,
allreduce_merge_scope):
super(VariableMgrLocalReplicated, self).__init__(benchmark_cnn)
if all_reduce_spec:
spec = allreduce.parse_all_reduce_spec(all_reduce_spec)
if len(spec) != 1:
raise ValueError(
'replicated mode does not support hybrid all-reduce strategies')
self._all_reduce_spec = spec[0]
else:
self._all_reduce_spec = None
self._agg_small_grads_max_bytes = agg_small_grads_max_bytes
self._agg_small_grads_max_group = agg_small_grads_max_group
self._warmup_ops = []
self._allreduce_merge_scope = allreduce_merge_scope
self._gradient_put_ops = None
def each_tower_has_variables(self):
return True
def create_outer_variable_scope(self, device_num):
return tf.variable_scope('v%s' % device_num, reuse=self._reuse_vars,
use_resource=self.use_resource_vars)
def preprocess_device_grads(self, device_grads):
compact_grads = (self.benchmark_cnn.params.use_fp16 and
self.benchmark_cnn.params.compact_gradient_transfer)
defer_grads = (self.benchmark_cnn.params.variable_consistency == 'relaxed')
grads_to_reduce = [[g for g, _ in grad_vars] for grad_vars in device_grads]
algorithm = batch_allreduce.algorithm_from_params(self.benchmark_cnn.params)
reduced_grads, self._warmup_ops = algorithm.batch_all_reduce(
grads_to_reduce, self.benchmark_cnn.params.gradient_repacking,
compact_grads, defer_grads, self.benchmark_cnn.params.xla_compile)
if self.benchmark_cnn.enable_auto_loss_scale:
# Check for infs or nans
is_finite_list = []
with tf.name_scope('check_for_inf_and_nan'):
for tower_grads in reduced_grads:
with tf.colocate_with(tower_grads[0]):
# TODO(tanmingxing): Create fused op that takes in a list of tensors
# as input and returns scalar boolean True if there are any
# infs/nans.
is_finite_list.append(tf.reduce_all(
[tf.reduce_all(tf.is_finite(g)) for g in tower_grads]))
self.grad_has_inf_nan = tf.logical_not(tf.reduce_all(is_finite_list))
reduced_device_grads = [[
(g, v) for g, (_, v) in zip(grads, grad_vars)
] for grads, grad_vars in zip(reduced_grads, device_grads)]
return self.benchmark_cnn.devices, reduced_device_grads
def get_gradients_to_apply(self, device_num, gradient_state):
device_grads = gradient_state
return device_grads[device_num]
def get_post_init_ops(self):
# Copy initialized values for variables on GPU 0 to other GPUs.
global_vars = tf.global_variables()
var_by_name = dict([(v.name, v) for v in global_vars])
post_init_ops = []
for v in global_vars:
split_name = v.name.split('/')
# TODO(b/62630508): use more specific prefix than v or v0.
if split_name[0] == 'v0' or not v.name.startswith('v'):
continue
split_name[0] = 'v0'
copy_from = var_by_name['/'.join(split_name)]
post_init_ops.append(v.assign(copy_from.read_value()))
post_init_ops += self._warmup_ops
return post_init_ops
def savable_variables(self):
"""Return the set of variables used for saving/loading the model."""
params = []
for v in tf.global_variables():
split_name = v.name.split('/')
if split_name[0] == 'v0' or not v.name.startswith('v'):
params.append(v)
return params
def get_devices(self):
return self.benchmark_cnn.raw_devices
class VariableMgrDistributedAllReduce(VariableMgr):
"""VariableMgr that implements the --distributed_all_reduce mode.
Each GPU has its own copy of the variables. To apply gradients,
the specified all-reduce algorithm is used to reduce the gradients
and replicate the final value to all GPUs.
"""
def __init__(self, benchmark_cnn, all_reduce_spec, job_name,
num_workers, agg_small_grads_max_bytes,
agg_small_grads_max_group, allreduce_merge_scope):
super(VariableMgrDistributedAllReduce, self).__init__(benchmark_cnn)
if not all_reduce_spec:
raise ValueError(
'distributed_all_reduce requires a non-empty all_reduce_spec')
self._all_reduce_spec = allreduce.parse_all_reduce_spec(all_reduce_spec)
self._all_reduce_device_prefixes = (
allreduce.build_all_reduce_device_prefixes(job_name, num_workers))
self._num_workers = num_workers
self._agg_small_grads_max_bytes = agg_small_grads_max_bytes
self._agg_small_grads_max_group = agg_small_grads_max_group
self._allreduce_merge_scope = allreduce_merge_scope
if not self._all_reduce_spec:
raise ValueError('all_reduce_spec must be specified')
self._single_session = True
def each_tower_has_variables(self):
return True
def create_outer_variable_scope(self, device_num):
"""Create a scope for the named device.
Args:
device_num: index of device for variable scope. (Note that
device_num spans all processes in cluster since a single global
graph is used.)
Returns:
the requested variable_scope
"""
return tf.variable_scope('v%s' % device_num, reuse=self._reuse_vars,
use_resource=self.use_resource_vars)
def preprocess_device_grads(self, device_grads):
remaining_grads = device_grads
aggregated_grads = []
for spec_tuple in self._all_reduce_spec:
if spec_tuple.limit < 0:
this_grads = remaining_grads
remaining_grads = []
else:
(this_grads, remaining_grads) = allreduce.split_grads_by_size(
spec_tuple.limit, remaining_grads)
if this_grads:
range_agg_grads = allreduce.sum_gradients_all_reduce(
self._single_session,
self._all_reduce_device_prefixes,
this_grads,
self._num_workers,
spec_tuple.alg,
spec_tuple.shards,
self.benchmark_cnn.gpu_indices,
agg_small_grads_max_bytes=self._agg_small_grads_max_bytes,
agg_small_grads_max_group=self._agg_small_grads_max_group,
allreduce_merge_scope=self._allreduce_merge_scope)
if not aggregated_grads:
aggregated_grads = range_agg_grads
else:
assert len(aggregated_grads) == len(range_agg_grads)
for i in range(len(aggregated_grads)):
aggregated_grads[i] += range_agg_grads[i]
assert not remaining_grads
full_device_set = []
for grads in device_grads:
g, v = grads[0]
del v
full_device_set.append(g.device)
return (full_device_set, aggregated_grads)
def get_gradients_to_apply(self, device_num, gradient_state):
device_grads = gradient_state
if device_num >= len(device_grads):
raise ValueError('device_num %d exceeds length of device_grads (%d)' %
(device_num, len(device_grads)))
return device_grads[device_num]
def get_post_init_ops(self):
"""Copy initialized values for variables to other devices."""
global_vars = tf.global_variables()
var_by_name = dict([(v.name, v) for v in global_vars])
post_init_ops = []
for v in global_vars:
split_name = v.name.split('/')
# TODO(b/62630508): use more specific prefix than v or v0.
if split_name[0] == 'v0' or not v.name.startswith('v'):
continue
split_name[0] = 'v0'
copy_from = var_by_name['/'.join(split_name)]
post_init_ops.append(v.assign(copy_from.read_value()))
return post_init_ops
def savable_variables(self):
"""Return the set of variables used for saving/loading the model."""
params = []
for v in tf.global_variables():
split_name = v.name.split('/')
if split_name[0] == 'v0' or not v.name.startswith('v'):
params.append(v)
return params
def get_devices(self):
return self.benchmark_cnn.raw_devices
# TODO(tucker): Merge this mode with DistributedAllReduce.
class VariableMgrCollectiveAllReduce(VariableMgr):
"""VariableMgr that implements the --collective_all_reduce mode.
Each GPU has its own copy of the variables. To apply gradients
the TF native collective all-reduce op is used to reduce the gradients
and replicate the final value to all GPUs.
"""
def __init__(self, benchmark_cnn, all_reduce_spec,
num_workers, num_gpus, task_id, allreduce_merge_scope):
super(VariableMgrCollectiveAllReduce, self).__init__(benchmark_cnn)
if not all_reduce_spec:
raise ValueError(
'collective_all_reduce requires a non-empty all_reduce_spec: %s'
% all_reduce_spec)
parsed_spec = allreduce.parse_all_reduce_spec(all_reduce_spec)
# So far we only support a length-1 all_reduce_spec
if len(parsed_spec) > 1 or parsed_spec[0].limit > 0:
raise ValueError(
'collective_all_reduce requires one single-range all_reduce_spec %s'
% parsed_spec)
self._all_reduce_spec = parsed_spec[0]
if self._all_reduce_spec.alg != 'collective':
raise ValueError(
'VariableMgrCollectiveAllReduce initialized with non-collective '
'all_reduce_spec %s' % self.all_reduce_spec)
self._num_workers = num_workers
self._num_gpus = num_gpus
self._task_id = task_id
self._allreduce_merge_scope = allreduce_merge_scope
self._instance_key_counter = 10000
self._instance_key_table = dict()
self._single_session = False
# List of prefixes for generating PS devices, unused here.
self._all_reduce_device_prefixes = None
def each_tower_has_variables(self):
return True
def create_outer_variable_scope(self, device_num):
"""Create a scope for the named device.
Args:
device_num: index of device for variable scope.
Returns:
the requested variable_scope
"""
return tf.variable_scope('v%s' % device_num, reuse=self._reuse_vars)
def preprocess_device_grads(self, device_grads):
reduced_grads = allreduce.sum_gradients_all_reduce(
self._single_session,
self._all_reduce_device_prefixes,
device_grads,
self._num_workers,
'collective',
self._all_reduce_spec.shards,
self.benchmark_cnn.gpu_indices,
allreduce_merge_scope=self._allreduce_merge_scope)
assert len(reduced_grads) == len(device_grads)
full_device_set = []
for grads in device_grads:
g, _ = grads[0]
full_device_set.append(g.device)
return (full_device_set, reduced_grads)
def get_gradients_to_apply(self, device_num, gradient_state):
device_grads = gradient_state
if device_num >= len(device_grads):
raise ValueError('device_num %d exceeds length of device_grads (%d)' %
(device_num, len(device_grads)))
return device_grads[device_num]
def _get_instance_key(self, name):
if name not in self._instance_key_table.keys():
self._instance_key_counter += 1
self._instance_key_table[name] = self._instance_key_counter
return self._instance_key_table[name]
def get_post_init_ops(self):
"""Broadcast initialized values of variables to other devices.
Returns:
At task 0 device 0, broadcast_send.
At all other devices and tasks, broadcast_recv.
"""
global_vars = tf.global_variables()
group_size = self._num_workers * self._num_gpus
post_init_ops = []
# Gather variables into same-var-different-device groups.
vars_by_suffix = dict()
for v in global_vars:
split_name = v.name.split('/')
mo = re.match(r'v(\d+)$', split_name[0])
if mo:
device_id = int(mo.group(1))
suffix = '/'.join(split_name[1:])
if suffix in vars_by_suffix.keys():
vars_by_suffix[suffix].append(v)
else:
vars_by_suffix[suffix] = [v]
# Generate broadcast ops for each such group.
for suffix in sorted(vars_by_suffix):
vlist = vars_by_suffix[suffix]
assert self._num_gpus == len(vlist)
devices = [v.device for v in vlist]
# NOTE: this key should generate the same value for all tasks
group_key = allreduce.collective_group_key(devices)
group_size = self._num_workers * len(devices)
instance_key = self._get_instance_key(suffix)
for v in vlist:
split_name = v.name.split('/')
mo = re.match(r'v(\d+)$', split_name[0])
if mo:
device_id = int(mo.group(1))
if (self._task_id == 0 and device_id == 0):
with tf.device(v.device):
bcast_send = allreduce.broadcast_send(
v, v.shape, v.dtype, group_size, group_key, instance_key)
post_init_ops.append(v.assign(bcast_send))
else:
with tf.device(v.device):
bcast_recv = allreduce.broadcast_recv(
v.shape, v.dtype, group_size, group_key, instance_key)
post_init_ops.append(v.assign(bcast_recv))
return post_init_ops
def savable_variables(self):
"""Return the set of variables used for saving/loading the model."""
params = []
if self._task_id == 0:
for v in tf.global_variables():
split_name = v.name.split('/')
if split_name[0] == 'v0' or not v.name.startswith('v'):
params.append(v)
return params
def get_devices(self):
return self.benchmark_cnn.raw_devices
class VariableMgrDistributedFetchFromPS(VariableMgr):
"""Implements --variable_update=parameter_server mode for distributed jobs.
Variables are stored on a parameter server. For each step, each tower gets
a copy of the variables from the parameter server, and sends its gradients
to the param server.
"""
def each_tower_has_variables(self):
return False
def create_outer_variable_scope(self, device_num):
if self.benchmark_cnn.local_parameter_device_flag == 'gpu':
caching_devices = self.benchmark_cnn.raw_devices
else:
caching_devices = [self.benchmark_cnn.cpu_device]
custom_getter = variable_mgr_util.OverrideCachingDevice(
caching_devices, self.benchmark_cnn.cpu_device, 1024 * 64)
return tf.variable_scope(
'v', reuse=bool(device_num) or self._reuse_vars,
custom_getter=custom_getter, use_resource=self.use_resource_vars)
def preprocess_device_grads(self, device_grads):
# Returns (gradient_devices, gradient_state)
return ([self.benchmark_cnn.param_server_device], device_grads)
def get_gradients_to_apply(self, device_num, gradient_state):
assert device_num == 0
agg_grads, self.grad_has_inf_nan = (
variable_mgr_util.aggregate_gradients_using_copy(
gradient_state,
use_mean=True,
check_inf_nan=self.benchmark_cnn.enable_auto_loss_scale))
return agg_grads
def get_devices(self):
ps_strategy = variable_mgr_util.GreedyLoadBalancingStrategy(
self.benchmark_cnn.num_ps, variable_mgr_util.byte_size_load_fn)
return [
tf.train.replica_device_setter(
worker_device=d,
cluster=self.benchmark_cnn.cluster_manager.get_cluster_spec(),
ps_strategy=ps_strategy) for d in self.benchmark_cnn.raw_devices
]
class VariableMgrDistributedFetchFromStagedPS(
VariableMgrDistributedFetchFromPS):
"""Extends VariableMgrDistributedFetchFromPS for --staged_vars."""
def __init__(self, benchmark_cnn):
super(VariableMgrDistributedFetchFromStagedPS, self).__init__(benchmark_cnn)
self.staging_vars_on_devices = [
dict() for _ in self.benchmark_cnn.raw_devices
]
self.staged_vars_on_cpu = {}
def create_outer_variable_scope(self, device_num):
self._custom_getter = variable_mgr_util.StagedVariableGetter(
device_num, self.benchmark_cnn.raw_devices,
self.benchmark_cnn.cpu_device, self)
return tf.variable_scope(
'v', reuse=bool(device_num) or self._reuse_vars,
custom_getter=self._custom_getter, use_resource=self.use_resource_vars)
def supports_staged_vars(self):
return True
def trainable_variables_on_device(self,
rel_device_num,
abs_device_num,
writable=False):
return self._custom_getter.trainable_variables_on_device(
rel_device_num, abs_device_num, writable=writable)
class VariableMgrDistributedReplicated(VariableMgr):
"""VariableMgr that implements the --distributed_replicated mode.
Each GPU has a copy of the variables, and updates its copy after the
parameter servers are all updated with the gradients from all servers. Only
works with cross_replica_sync=true. Unlike 'replicated', does not use nccl
all-reduce for replicating within a server.
"""
def each_tower_has_variables(self):
return True
def create_outer_variable_scope(self, device_num):
return tf.variable_scope(
'v%s' % device_num, reuse=self._reuse_vars,
custom_getter=variable_mgr_util.OverrideToLocalVariableIfNotPsVar(),
use_resource=self.use_resource_vars)
def preprocess_device_grads(self, device_grads):
return ([self.benchmark_cnn.param_server_device], device_grads)
def get_gradients_to_apply(self, device_num, gradient_state):
device_grads = gradient_state # From 2nd result of preprocess_device_grads.
avg_grads, self.grad_has_inf_nan = (
variable_mgr_util.aggregate_gradients_using_copy_with_device_selection(
self.benchmark_cnn,
device_grads,
use_mean=True,
check_inf_nan=self.benchmark_cnn.enable_auto_loss_scale))
# Make shadow variable on a parameter server for each original trainable
# variable.
for i, (g, v) in enumerate(avg_grads):
my_name = variable_mgr_util.PS_SHADOW_VAR_PREFIX + '/' + v.name
if my_name.endswith(':0'):
my_name = my_name[:-2]
new_v = tf.get_variable(
my_name,
dtype=v.dtype.base_dtype,
initializer=v.initial_value,
trainable=True)
avg_grads[i] = (g, new_v)
return avg_grads
def append_apply_gradients_ops(self, gradient_state, opt, grads, training_ops,
loss_scale_params):
device_grads = gradient_state # From 2nd result of preprocess_device_grads.
def get_apply_gradients_ops_func():
"""Returns a list of ops for updating gradients."""
apply_gradients_ops = []
# For each variable, apply the combined gradients for this server on
# the parameter server, and then wait for all other servers to do this.
for i, (g, v) in enumerate(grads):
apply_gradient_op = opt.apply_gradients([(g, v)])
barrier = self.benchmark_cnn.add_sync_queues_and_barrier(
'replicate_variable_%s' % i, [apply_gradient_op])
with tf.control_dependencies([barrier]):
with tf.device(self.benchmark_cnn.cpu_device):
updated_value = v.read_value()
for my_d in range(len(self.benchmark_cnn.devices)):
apply_gradients_ops.append(
device_grads[my_d][i][1].assign(updated_value))
return apply_gradients_ops
variable_mgr_util.append_gradients_with_loss_scale(
training_ops, get_apply_gradients_ops_func, loss_scale_params,
self.grad_has_inf_nan)
def _strip_port(self, s):
if s.endswith(':0'):
return s[:-2]
return s
def get_post_init_ops(self):
# Copy initialized variables for variables on the parameter server
# to the local copy of the variable.
local_vars = tf.local_variables()
local_var_by_name = dict(
[(self._strip_port(v.name), v) for v in local_vars])
post_init_ops = []
for v in tf.global_variables():
if v.name.startswith(variable_mgr_util.PS_SHADOW_VAR_PREFIX + '/v0/'):
prefix = self._strip_port(
v.name[len(variable_mgr_util.PS_SHADOW_VAR_PREFIX + '/v0'):])
for i in range(self.benchmark_cnn.num_gpus):
name = 'v%s%s' % (i, prefix)
if name in local_var_by_name:
copy_to = local_var_by_name[name]
post_init_ops.append(copy_to.assign(v.read_value()))
return post_init_ops
def _remove_shadow_var_prefix_if_present(self, var_name):
if var_name.startswith(variable_mgr_util.PS_SHADOW_VAR_PREFIX + '/'):
return var_name[len(variable_mgr_util.PS_SHADOW_VAR_PREFIX + '/'):]
else:
return var_name
def var_dict_name(self, v):
return self._strip_port(self._remove_shadow_var_prefix_if_present(v.name))
def savable_variables(self):
"""Returns a list/dict of savable variables to pass to tf.train.Saver."""
params = {}
for v in tf.global_variables():
assert (v.name.startswith(variable_mgr_util.PS_SHADOW_VAR_PREFIX + '/v0/')
or v.name in ('global_step:0', 'loss_scale:0',
'loss_scale_normal_steps:0')), (
'Invalid global variable: %s' % v)
# We store variables in the checkpoint with the shadow variable prefix
# removed so we can evaluate checkpoints in non-distributed replicated
# mode. The checkpoints can also be loaded for training in
# distributed_replicated mode.
name = self._strip_port(self._remove_shadow_var_prefix_if_present(v.name))
params[name] = v
for v in tf.local_variables():
# Non-trainable variables, such as batch norm moving averages, do not have
# corresponding global shadow variables, so we add them here. Trainable
# local variables have corresponding global shadow variables, which were
# added in the global variable loop above.
if v.name.startswith('v0/') and v not in tf.trainable_variables():
params[self._strip_port(v.name)] = v
return params
def get_devices(self):
return self.benchmark_cnn.raw_devices
TensorFlow/ComputeVision/Accuracy_Validation/benchmarks-master/scripts/tf_cnn_benchmarks/variable_mgr_util.py 0 → 100644
View file @ 441c8f40
# 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.
# ==============================================================================
"""Utilities for VariableMgr."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import collections as pycoll
import operator
import numpy as np
import tensorflow.compat.v1 as tf
# pylint: disable=g-direct-tensorflow-import
from tensorflow.python.framework import ops
from tensorflow.python.framework import tensor_shape
from tensorflow.python.ops import array_ops
from tensorflow.python.ops import data_flow_ops
from tensorflow.python.ops import math_ops
PS_SHADOW_VAR_PREFIX = 'ps_var'
AutoLossScaleParams = pycoll.namedtuple(
'AutoLossScaleParams',
[
# If true, enable automatic loss scaling.
'enable_auto_loss_scale',
# The value to scale the loss before computing gradients.
'loss_scale',
# Number of normal steps with the current `loss_scale`.
'loss_scale_normal_steps',
# Increase loss scale every n steps.
'inc_loss_scale_every_n',
# If true, the current worker is chief. The current implementation
# relies on the chief to update loss_scale value, but in future, we
# might change this to ask the parameter server to update loss_scales
# for better performance.
# TODO(tanmingxing): remove this if loss_scale is updated in ps.
'is_chief',
])
def get_loss_scale_update_op(loss_scale, loss_scale_normal_steps,
inc_loss_scale_every_n):
"""Returns the update op for loss scaling variables.
We maintain the counter `loss_scale_normal_steps` to count the number of steps
we have been using the current `loss_scale`. In most cases, this function
increments `loss_scale_normal_steps`. However, if `loss_scale_normal_steps` is
greater than the threshold `inc_loss_scale_every_n`, we double `loss_scale`
and reset `loss_scale_normal_steps` to zero.
This op is only called if the gradients don't have any infs or nans. Instead,
if infs or nans occur in the gradients, we immeditately halve `loss_scale` and
reset `loss_scale_normal_steps` to zero.
Args:
loss_scale: a tf.Variable represneting the loss_scale value.
loss_scale_normal_steps: a tf.Variable representing the number of training
steps that have run since the loss_scale last changed.
inc_loss_scale_every_n: a Python integer threshold. `loss_scale` is
increased every `inc_loss_scale_every_n` steps, unless the gradients have
infs or nans.
Returns:
An op for updating `loss_scale` and `loss_scale_normal_steps`.
"""
def increment_loss_scale_normal_steps_func():
return tf.group(loss_scale_normal_steps.assign_add(1))
def increase_loss_scale_func():
return tf.group(
tf.assign(loss_scale_normal_steps, 0),
tf.assign(loss_scale, loss_scale * 2))
# true_fn and false_fn must have the same type.
return tf.cond(loss_scale_normal_steps < inc_loss_scale_every_n,
increment_loss_scale_normal_steps_func,
increase_loss_scale_func)
def append_gradients_with_loss_scale(training_ops, get_apply_gradients_ops_func,
loss_scale_params, grad_has_inf_nan):
"""Selectively appends gradients update ops with loss scaling.
Args:
training_ops: a list of training ops to be executed.
get_apply_gradients_ops_func: a function that returns a list of ops for
applying gradients. Here, we must pass a function instead of the actual
list of ops; otherwise, those ops would be executed unconditionally due to
the semantics of tf.cond.
loss_scale_params: An AutoLossScaleParams tuple.
grad_has_inf_nan: Boolean tensor indicating whether the gradients have infs
or nans.
"""
is_chief = loss_scale_params.is_chief
loss_scale = loss_scale_params.loss_scale
loss_scale_normal_steps = loss_scale_params.loss_scale_normal_steps
inc_loss_scale_every_n = loss_scale_params.inc_loss_scale_every_n
enable_auto_loss_scale = loss_scale_params.enable_auto_loss_scale
if loss_scale is None or not enable_auto_loss_scale or not is_chief:
training_ops.extend(get_apply_gradients_ops_func())
else:
# If nans/infs occurred, skip applying gradients and instead update
# loss_scale (halve loss_scale and reset loss_scale_normal_steps to zero).
def update_op_if_nan_or_inf():
"""Update loss_scale and discard gradients if nans/infs occurred."""
return tf.group(
tf.assign(loss_scale, loss_scale / 2.),
tf.assign(loss_scale_normal_steps, 0))
# Otherwise, apply gradients, and update loss_scale and
# loss_scale_normal_steps.
def update_op_if_no_nan_or_inf():
"""Apply gradients, and update loss scaling."""
return tf.group(
get_loss_scale_update_op(loss_scale, loss_scale_normal_steps,
inc_loss_scale_every_n),
*get_apply_gradients_ops_func())
# TODO(tanmingxing): Add support for independent and distributed all_reduce.
assert grad_has_inf_nan is not None
update_op = tf.cond(
grad_has_inf_nan,
update_op_if_nan_or_inf,
update_op_if_no_nan_or_inf,
name='cond_if_grad_has_inf_nan'
)
training_ops.append(update_op)
# To be used with custom_getter on tf.get_variable.
class OverrideCachingDevice(object):
"""Variable getter which caches variables on the least loaded device.
Variables smaller than a certain threshold are cached on a single specific
device, as specified in the constructor. All other variables are load balanced
across a pool of devices, by caching each variable on the least loaded device.
Note that variable creation only happen when building the model graph on the
first device (see how it sets the 'reuse' parameter in
VariableMgr.*.create_outer_variable_scope()). That means, for all other
devices, the variable scope will reuse the variables created before, which
requires that we set the caching_device correctly as otherwise it may not be
able to find the previously created variable and will create a new one. This
requires when building the model graph on different devices, variables with
the same name should have same size.
TODO(laigd): consider adding tests or verification logic to enforce this, or
refactor it.
"""
def __init__(self, devices, device_for_small_variables,
small_variable_size_threshold):
self.devices = devices
self.sizes = [0] * len(self.devices)
self.device_for_small_variables = device_for_small_variables
self.small_variable_size_threshold = small_variable_size_threshold
def __call__(self, getter, *args, **kwargs):
size = tf.TensorShape(kwargs['shape']).num_elements()
if size < self.small_variable_size_threshold:
device_name = self.device_for_small_variables
else:
device_index, _ = min(enumerate(self.sizes), key=operator.itemgetter(1))
device_name = self.devices[device_index]
self.sizes[device_index] += size
kwargs['caching_device'] = device_name
var = getter(*args, **kwargs)
return var
# To be used with custom_getter on tf.get_variable. Ensures the created variable
# is in LOCAL_VARIABLES and not GLOBAL_VARIBLES collection.
class OverrideToLocalVariableIfNotPsVar(object):
# args and kwargs come from the custom_getter interface for Tensorflow
# variables, and matches tf.get_variable's signature, with the addition of
# 'getter' at the beginning.
def __call__(self, getter, name, *args, **kwargs):
if name.startswith(PS_SHADOW_VAR_PREFIX):
return getter(*args, **kwargs)
if 'collections' in kwargs:
collections = kwargs['collections']
if not collections:
collections = [tf.GraphKeys.GLOBAL_VARIABLES]
else:
collections = collections[:]
collections.remove(tf.GraphKeys.GLOBAL_VARIABLES)
collections.append(tf.GraphKeys.LOCAL_VARIABLES)
kwargs['collections'] = list(collections)
return getter(name, *args, **kwargs)
class ParamServerDeviceSetter(object):
"""Helper class to assign variables on the least loaded ps-device."""
def __init__(self, worker_device, ps_devices):
"""Initializer for ParamServerDevicSetter.
Args:
worker_device: the device to use for computer ops.
ps_devices: a list of device to use for Variable ops. Each variable is
assigned to the least loaded device.
"""
self.ps_devices = ps_devices
self.worker_device = worker_device
self.ps_sizes = [0] * len(self.ps_devices)
def __call__(self, op):
if op.device:
return op.device
if op.type not in ['Variable', 'VariableV2']:
return self.worker_device
device_index, _ = min(enumerate(self.ps_sizes), key=operator.itemgetter(1))
device_name = self.ps_devices[device_index]
var_size = op.outputs[0].get_shape().num_elements()
self.ps_sizes[device_index] += var_size
return device_name
class StagedModelVariable(object):
"""Staging variable wrapper that decouples reads and updates.
This class represents a variable through a staging buffer. Reads from this
variable directly gets from the staging buffer. Updates are stacked into
another staging buffer, and will be processed later.
"""
def __init__(self, real_var, var_stage_get, variable_mgr):
"""Initializer for the model variables through a staging buffer.
Args:
real_var: the underlying real variable.
var_stage_get: the read op from the staging buffer.
variable_mgr: the parent variable-manager.
"""
self.real_var = real_var
self.var_stage_get = var_stage_get
self.variable_mgr = variable_mgr
def _value(self):
"""The read access of this variable. The content from the staging buffer."""
return self.var_stage_get
def _ref(self):
"""Return the underlying variable ref, required by tf.colocate_with."""
return self.real_var._ref() # pylint: disable=protected-access
def read_value(self):
"""Mimics tf.Variable.read_value()."""
return tf.identity(self.var_stage_get, name='read')
@property
def dtype(self):
"""Return the non-reference dtype."""
return self.var_stage_get.dtype
def assign_sub(self, delta, name=None, read_value=True):
"""Mimic the updates to the variable.
Args:
delta: is pushed into a staging buffer and will be pumped later.
name: currently ignored; names of ops and the StagingArea are
computed without using this pass name.
read_value: if True, will return something which evaluates to the new
value of the variable; if False will return the assign op.
Returns:
The actual updates. The colocation constraint will be reapplied.
"""
# This parameter is ignored: the StagingArea only supports setting
# the shared name, not the names of individual ops it uses.
del name
# colocate_with(None, True) clears the colocation constraints.
# Push the delta into a staging buffer.
with ops.colocate_with(None, True), tf.device(self.var_stage_get.device):
delta_staging_area = data_flow_ops.StagingArea(
[self.var_stage_get.dtype], shapes=[self.var_stage_get.shape])
delta_put_op = delta_staging_area.put([delta])
self.variable_mgr.staging_delta_ops.append(delta_put_op)
delta_get_op = delta_staging_area.get()[0]
# Return the actual updates. The colocation constraint will be reapplied.
return self.real_var.assign_sub(delta_get_op, read_value=read_value)
@staticmethod
# pylint: disable=bad-staticmethod-argument,invalid-name
def _TensorConversionFunction(self, dtype=None, name=None, as_ref=False):
"""Utility function for converting a StagedModelVariable to a Tensor."""
del dtype, name # unused: this function returns the cached ref or value.
if as_ref:
return self._ref()
else:
return self._value()
ops.register_tensor_conversion_function(
StagedModelVariable, StagedModelVariable._TensorConversionFunction) # pylint: disable=protected-access
class StagedVariableGetter(object):
"""A variable getter through staging buffers on devices.
Instead of a caching device, this getter tracks where the variable is used.
And on each device, it goes through a staging buffer.
"""
def __init__(self, device_num, devices, cpu_device, variable_mgr):
"""Initializer for StagedVariableGetter.
Args:
device_num: the current device index.
devices: a list of all the devices to build towers.
cpu_device: a cpu_device for this replica. If None, no cpu-caching is
done.
variable_mgr: the parent variable manager.
"""
self.device_num = device_num
self.devices = devices
self.cpu_device = cpu_device
self.variable_mgr = variable_mgr
def __call__(self, getter, name, *args, **kwargs):
staging_ops = self.variable_mgr.staging_vars_on_devices[self.device_num]
if name in staging_ops:
put_op, get_op = staging_ops[name]
return get_op
real_var = getter(name, *args, **kwargs)
shape = kwargs['shape']
dtype = kwargs['dtype']
trainable = kwargs['trainable']
if self.cpu_device:
with tf.device(self.cpu_device):
# This helps copying the weights from the parameter to this server only
# once.
if name in self.variable_mgr.staged_vars_on_cpu:
cpu_var = self.variable_mgr.staged_vars_on_cpu[name]
else:
cpu_var = tf.identity(real_var)
self.variable_mgr.staged_vars_on_cpu[name] = cpu_var
var_to_stage = cpu_var
else:
var_to_stage = tf.identity(real_var) # de-reference the variable.
with tf.device(self.devices[self.device_num]):
staging_area = data_flow_ops.StagingArea([dtype], shapes=[shape])
put_op = staging_area.put([var_to_stage])
get_op = staging_area.get()[0]
staging_ops[name] = (put_op, get_op)
if trainable:
# For trainable variables, they are managed separatedly through
# apply_gradients.
return get_op
else:
# For other shadow variables, the access is decoupled through a wrapper
# class.
return StagedModelVariable(real_var, get_op, self.variable_mgr)
def trainable_variables_on_device(self, rel_device_num, abs_device_num,
writable):
"""Return the set of trainable variables on the specified device.
Args:
rel_device_num: local worker device index.
abs_device_num: global graph device index.
writable: whether the returned variables is writable or read-only.
Returns:
Return the set of trainable variables on the specified device.
"""
del abs_device_num
params_refs = tf.trainable_variables()
if writable:
return params_refs
params = []
for param in params_refs:
var_name = param.name.split(':')[0]
_, var_get_op = self.variable_mgr.staging_vars_on_devices[rel_device_num][
var_name]
params.append(var_get_op)
return params
def aggregate_gradients_using_copy_with_device_selection(
benchmark_cnn, tower_grads, use_mean, check_inf_nan):
"""Aggregate gradients, controlling device for the aggregation.
Args:
benchmark_cnn: benchmark_cnn class.
tower_grads: List of lists of (gradient, variable) tuples. The outer list
is over towers. The inner list is over individual gradients.
use_mean: if True, mean is taken, else sum of gradients is taken.
check_inf_nan: If true, check grads for nans and infs.
Returns:
The tuple ([(average_gradient, variable),], has_nan_or_inf) where the
gradient has been averaged across all towers. The variable is chosen from
the first tower. The has_nan_or_inf indicates the grads has nan or inf.
"""
if benchmark_cnn.local_parameter_device_flag == 'gpu':
avail_devices = benchmark_cnn.raw_devices
else:
avail_devices = [benchmark_cnn.param_server_device]
agg_grads = []
has_nan_or_inf_list = []
for i, single_grads in enumerate(zip(*tower_grads)):
with tf.device(avail_devices[i % len(avail_devices)]):
grad_and_var, has_nan_or_inf = aggregate_single_gradient_using_copy(
single_grads, use_mean, check_inf_nan)
agg_grads.append(grad_and_var)
has_nan_or_inf_list.append(has_nan_or_inf)
if check_inf_nan:
return agg_grads, tf.reduce_any(has_nan_or_inf_list)
else:
return agg_grads, None
def aggregate_gradients_using_copy_with_variable_colocation(
tower_grads, use_mean, check_inf_nan):
"""Aggregate gradients, colocating computation with the gradient's variable.
Args:
tower_grads: List of lists of (gradient, variable) tuples. The outer list
is over towers. The inner list is over individual gradients. All variables
of the same gradient across towers must be the same (that is,
tower_grads[x][a][1] == tower_grads[y][a][1] for all indices x, y, and a)
use_mean: if True, mean is taken, else sum of gradients is taken.
check_inf_nan: If true, check grads for nans and infs.
Returns:
The tuple ([(average_gradient, variable),], has_nan_or_inf) where the
gradient has been averaged across all towers. The variable is chosen from
the first tower. The has_nan_or_inf indicates the grads has nan or inf.
"""
agg_grads = []
has_nan_or_inf_list = []
for single_grads in zip(*tower_grads):
# Note that each single_grads looks like the following:
# ((grad0_gpu0, var0_gpu0), ... , (grad0_gpuN, var0_gpuN))
var = single_grads[0][1]
for _, v in single_grads:
assert v == var
with tf.device(var.device):
grad_and_var, has_nan_or_inf = aggregate_single_gradient_using_copy(
single_grads, use_mean, check_inf_nan)
agg_grads.append(grad_and_var)
has_nan_or_inf_list.append(has_nan_or_inf)
if check_inf_nan:
return agg_grads, tf.reduce_any(has_nan_or_inf_list)
else:
return agg_grads, None
def aggregate_gradients_using_copy(tower_grads, use_mean, check_inf_nan):
"""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 towers. The inner list is over individual gradients.
use_mean: if True, mean is taken, else sum of gradients is taken.
check_inf_nan: check grads for nans and infs.
Returns:
The tuple ([(average_gradient, variable),], has_nan_or_inf) where the
gradient has been averaged across all towers. The variable is chosen from
the first tower. The has_nan_or_inf indicates the grads has nan or inf.
"""
agg_grads = []
has_nan_or_inf_list = []
for single_grads in zip(*tower_grads):
grad_and_var, has_nan_or_inf = aggregate_single_gradient_using_copy(
single_grads, use_mean, check_inf_nan)
agg_grads.append(grad_and_var)
has_nan_or_inf_list.append(has_nan_or_inf)
if check_inf_nan:
return agg_grads, tf.reduce_any(has_nan_or_inf_list)
else:
return agg_grads, None
# The following two functions are copied from
# tensorflow/python/eager/backprop.py. We do not directly use them as they are
# not exported and subject to change at any time.
def flatten_nested_indexed_slices(grad):
assert isinstance(grad, ops.IndexedSlices)
if isinstance(grad.values, ops.Tensor):
return grad
else:
assert isinstance(grad.values, ops.IndexedSlices)
g = flatten_nested_indexed_slices(grad.values)
return ops.IndexedSlices(g.values, array_ops.gather(grad.indices,
g.indices),
g.dense_shape)
def aggregate_indexed_slices_gradients(grads):
"""Aggregates gradients containing `IndexedSlices`s."""
if len(grads) < 1:
return None
elif len(grads) == 1:
return grads[0]
else:
grads = [g for g in grads if g is not None]
# If any gradient is a `Tensor`, sum them up and return a dense tensor
# object.
if any(isinstance(g, ops.Tensor) for g in grads):
return math_ops.add_n(grads)
# The following `_as_indexed_slices_list` casts ids of IndexedSlices into
# int64. It is to make sure the inputs of `concat` all have same the data
# type.
grads = math_ops._as_indexed_slices_list(grads) # pylint: disable=protected-access
grads = [flatten_nested_indexed_slices(x) for x in grads]
# Form IndexedSlices out of the concatenated values and indices.
concat_grad = ops.IndexedSlices(
array_ops.concat([x.values for x in grads], axis=0),
array_ops.concat([x.indices for x in grads], axis=0),
grads[0].dense_shape)
return concat_grad
def aggregate_single_gradient_using_copy(grad_and_vars, use_mean,
check_inf_nan):
"""Calculate the average gradient for a shared variable across all towers.
Note that this function provides a synchronization point across all towers.
Args:
grad_and_vars: A list or tuple of (gradient, variable) tuples. Each
(gradient, variable) pair within the outer list represents the gradient
of the variable calculated for a single tower, and the number of pairs
equals the number of towers.
use_mean: if True, mean is taken, else sum of gradients is taken.
check_inf_nan: check grads for nans and infs.
Returns:
The tuple ([(average_gradient, variable),], has_nan_or_inf) where the
gradient has been averaged across all towers. The variable is chosen from
the first tower. The has_nan_or_inf indicates the grads has nan or inf.
"""
grads = [g for g, _ in grad_and_vars]
if any(isinstance(g, tf.IndexedSlices) for g in grads):
# TODO(reedwm): All-reduce IndexedSlices more effectively.
grad = aggregate_indexed_slices_gradients(grads)
else:
grad = tf.add_n(grads)
if use_mean and len(grads) > 1:
grad = tf.scalar_mul(1.0 / len(grads), grad)
v = grad_and_vars[0][1]
if check_inf_nan:
with tf.name_scope('check_for_inf_and_nan'):
has_nan_or_inf = tf.logical_not(tf.reduce_all(tf.is_finite(grads)))
return (grad, v), has_nan_or_inf
else:
return (grad, v), None
# This class is copied from
# https://github.com/tensorflow/tensorflow/blob/590d6eef7e91a6a7392c8ffffb7b58f2e0c8bc6b/tensorflow/contrib/training/python/training/device_setter.py#L56.
# We copy it since contrib has been removed from TensorFlow.
class GreedyLoadBalancingStrategy(object):
"""Returns the least-loaded ps task for op placement.
The load is calculated by a user-specified load function passed in at
construction. There are no units for load, and the load function is
responsible for providing an internally consistent measure.
Note that this strategy is very sensitive to the exact order in which
ps ops (typically variables) are created, as it greedily places ops
on the least-loaded ps at the point each op is processed.
One reasonable heuristic is the `byte_size_load_fn`, which
estimates load as the number of bytes that would be used to store and
transmit the entire variable. More advanced load functions
could consider the difference in access patterns across ops, or trade
off CPU-intensive ops with RAM-intensive ops with network bandwidth.
This class is intended to be used as a `ps_strategy` in
`tf.compat.v1.train.replica_device_setter`.
"""
def __init__(self, num_tasks, load_fn):
"""Create a new `LoadBalancingStrategy`.
Args:
num_tasks: Number of ps tasks to cycle among.
load_fn: A callable that takes an `Operation` and returns a
numeric load value for that op.
"""
self._num_tasks = num_tasks
self._load_fn = load_fn
self._ps_loads = np.zeros(num_tasks)
def __call__(self, op):
"""Choose a ps task index for the given `Operation`.
Args:
op: A `Operation` to be placed on ps.
Returns:
The next ps task index to use for the `Operation`. Greedily
places the op on the least-loaded ps task so far, as determined
by the load function.
"""
task = np.argmin(self._ps_loads)
self._ps_loads[task] += self._load_fn(op)
return task
# This function is copied from
# https://github.com/tensorflow/tensorflow/blob/590d6eef7e91a6a7392c8ffffb7b58f2e0c8bc6b/tensorflow/contrib/training/python/training/device_setter.py#L105.
# We copy it since contrib has been removed from TensorFlow.
def byte_size_load_fn(op):
"""Load function that computes the byte size of a single-output `Operation`.
This is intended to be used with `"Variable"` ops, which have a single
`Tensor` output with the contents of the variable. However, it can also be
used for calculating the size of any op that has a single output.
Intended to be used with `GreedyLoadBalancingStrategy`.
Args:
op: An `Operation` with a single output, typically a "Variable" op.
Returns:
The number of bytes in the output `Tensor`.
Raises:
ValueError: if `op` does not have a single output, or if the shape of the
single output is not fully-defined.
"""
if len(op.outputs) != 1:
raise ValueError('Op %s must have a single output' % op)
output = op.outputs[0]
elem_size = output.dtype.size
shape = output.get_shape()
if not shape.is_fully_defined():
# Due to legacy behavior, scalar "Variable" ops have output Tensors that
# have unknown shape when the op is created (and hence passed to this
# load function for placement), even though the scalar shape is set
# explicitly immediately afterward.
shape = tensor_shape.TensorShape(op.get_attr('shape'))
shape.assert_is_fully_defined()
return shape.num_elements() * elem_size
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