def testCreateLocalTheta(self):
methods = [py_utils.WeightInit.Gaussian, py_utils.WeightInit.Uniform]
dtypes = [tf.float32, tf.complex64]
shapes = [[2, 4], [3]]
test_vars = py_utils.NestedMap()
for i, (m, dt, sp) in enumerate(itertools.product(methods, dtypes, shapes)):
pc = py_utils.WeightParams(sp, m(), dt, 'col1')
test_vars['var_%d' % i] = py_utils.CreateVariable('var_%d' % i, pc)[0]
test_devices = [
'/job:worker/replica:0/device:GPU:0',
'/job:worker/replica:0/device:GPU:1'
]
sharded_local_vars = py_utils.CreateLocalTheta(test_vars, test_devices)
sharded_local_vars_list = sharded_local_vars.Flatten()
# assert the name is now Identity*
for v in sharded_local_vars_list:
self.assertTrue('Identity' in v.name)
# assert proper device placement
for i, v in enumerate(sharded_local_vars_list):
expected_device = test_devices[i % len(test_devices)]
self.assertEqual(v.device, expected_device)
def _FPropSplitInputBatch(self, theta, input_batch):
"""Splits the input batch on the input device."""
if py_utils.use_tpu():
return self._FPropTpu(theta, input_batch)
cluster = self.cluster
num_splits = cluster.num_splits_per_client
if not isinstance(input_batch, list):
input_batch = [input_batch]
assert len(input_batch) == num_splits, (len(input_batch), num_splits)
# dev_list_per_replica[i][j] is the i-th worker's j-th device.
dev_list_per_replica = cluster.available_devices.tolist()
# Asserts invariant of the total number of splits w.r.t.,
# splits per worker.
splits_per_replica = cluster.num_splits_per_replica
assert num_splits == splits_per_replica * len(dev_list_per_replica), (
num_splits, splits_per_replica, len(dev_list_per_replica))
all_metrics = []
all_per_example_tensors = []
with cluster:
for w_id, w_devs in enumerate(dev_list_per_replica):
# Make local copy of the vars, shard on devices for this worker.
theta_local = py_utils.CreateLocalTheta(
theta, w_devs, label='worker %d' % w_id)
for s_id in range(splits_per_replica):
# s_id-th split for the w_id-th worker.
split_id = splits_per_replica * w_id + s_id
with cluster_factory.SetModelSplit(split_id) as c:
with tf.device(c.WorkerDeviceInModelSplit(0)):
with tf.name_scope('tower_%d_%d' % (w_id, s_id)):
batch = input_batch[split_id]
metrics, per_example = self.FPropTower(theta_local, batch)
all_metrics.append(metrics)
all_per_example_tensors.append(per_example)
return py_utils.WeightedAvgOfMetrics(
all_metrics), py_utils.ConcatPerExampleTensors(all_per_example_tensors)
def _FPropSplitInputBatch(self, theta, input_batch):
"""Splits the input batch on the input device."""
cluster = self.cluster
num_splits = cluster.num_splits_per_client
if not isinstance(input_batch, list):
input_batch = [input_batch]
assert len(input_batch) == num_splits, (len(input_batch), num_splits)
# dev_list_per_replica[i][j] is the i-th worker's j-th device.
dev_list_per_replica = cluster.available_devices.tolist()
# Asserts invariant of the total number of splits w.r.t.,
# splits per worker.
splits_per_replica = cluster.num_splits_per_replica
assert num_splits == splits_per_replica * len(dev_list_per_replica), (
num_splits, splits_per_replica, len(dev_list_per_replica))
all_fprop_metrics = []
for w_id, w_devs in enumerate(dev_list_per_replica):
# Make local copy of the vars, shard on devices for this worker.
theta_local = py_utils.CreateLocalTheta(theta,
w_devs,
label='worker %d' % w_id)
for s_id in range(splits_per_replica):
# s_id-th split for the w_id-th worker.
split_id = splits_per_replica * w_id + s_id
with py_utils.ModelSplit(split_id):
with tf.device(cluster.WorkerDeviceInModelSplit(0)):
with tf.name_scope('tower_%d_%d' % (w_id, s_id)):
batch = self.input_generator.PreprocessInputBatch(
input_batch[split_id])
dec_metrics = self.FPropTower(theta_local, batch)
all_fprop_metrics.append(dec_metrics)
return py_utils.WeightedAvgOfMetrics(all_fprop_metrics)
def FProp(self, theta):
"""Forward propagation.
This default `FProp` implementation here supports batch splitting in
synchronous and asynchronous training when sub-classes implement
`FPropTower`.
Args:
theta: A `.NestedMap` object containing weights' values of this
layer and its children layers.
Returns:
A dict containing metrics pairs. One of the keys should be 'loss' and its
value should be a (loss, num_predictions) pair.
"""
p = self.params
cluster = cluster_factory.Current()
with tf.name_scope('fprop'), tf.name_scope(p.name):
all_fprop_metrics = []
if py_utils.use_tpu():
batch = self.input_generator.CreateTpuFeeds()
with tf.name_scope('tower_0_0'):
dec_metrics = self.FPropTower(theta, batch)
all_fprop_metrics.append(dec_metrics)
else:
# Splits the input batch on the input device.
num_splits = cluster.num_splits_per_client
with tf.device(cluster.input_device):
batches = self.input_generator.SplitInputBatch(num_splits)
assert num_splits == len(batches)
# dev_list_per_replica[i][j] is the i-th worker's j-th device.
dev_list_per_replica = cluster.available_devices.tolist()
# Asserts invariant of the total number of splits w.r.t.,
# splits per worker.
splits_per_replica = cluster.num_splits_per_replica
assert num_splits == splits_per_replica * len(
dev_list_per_replica)
for w_id, w_devs in enumerate(dev_list_per_replica):
# Make local copy of the vars, shard on devices for this worker.
theta_local = py_utils.CreateLocalTheta(theta,
w_devs,
label='worker %d' %
w_id)
for s_id in range(splits_per_replica):
# s_id-th split for the w_id-th worker.
split_id = splits_per_replica * w_id + s_id
with py_utils.ModelSplit(split_id):
with tf.device(
cluster.WorkerDeviceInModelSplit(0)):
with tf.name_scope('tower_%d_%d' %
(w_id, s_id)):
batch = self.input_generator.PreprocessInputBatch(
batches[split_id])
dec_metrics = self.FPropTower(
theta_local, batch)
all_fprop_metrics.append(dec_metrics)
metrics = py_utils.WeightedAvgOfMetrics(all_fprop_metrics)
# Adds stats about the input batch.
metrics['num_samples_in_batch'] = (tf.convert_to_tensor(
self.input_generator.InputBatchSize()), tf.constant(1.0))
# Generates summaries.
for name, (value, weight) in six.iteritems(metrics):
self.AddEvalMetric(name, value, weight)
# Loss.
self._loss, self._num_predicts = metrics['loss']
self._loss = py_utils.CheckNumerics(self._loss)
return metrics
