def unpack(self, parallel_tensor):
"""Unpack a parallel tensor into its components.
Args:
parallel_tensor: A tensor or composite tensor placed on the
ParallelDevice.
Returns:
A flat list of tensors, one per `self.components`.
"""
self._assert_eager()
if not isinstance(parallel_tensor,
(ops.Tensor, composite_tensor.CompositeTensor,
variables.Variable)):
raise ValueError(
"Expected a tensor, got {}.".format(parallel_tensor))
with ops.device(self._name):
flattened = nest.flatten(parallel_tensor, expand_composites=True)
unpacked = []
for packed in flattened:
unpacked.append(
tpu_ops.tpu_replicated_output(packed,
num_replicas=len(
self.components)))
return [
nest.pack_sequence_as(parallel_tensor,
component_value,
expand_composites=True)
for component_value in zip(*unpacked)
]
def _unpack_tensor(self, parallel_tensor):
"""Helper to unpack a single tensor."""
if not isinstance(parallel_tensor, (
ops.Tensor, composite_tensor.CompositeTensor, variables.Variable)):
raise ValueError(
"Expected a tensor, got {}.".format(parallel_tensor))
with ops.device(self._name):
return tpu_ops.tpu_replicated_output(
parallel_tensor, num_replicas=len(self.components))
def unpack(self, parallel_tensor):
"""Unpack a parallel tensor into its components.
Args:
parallel_tensor: A tensor placed on the ParallelDevice.
Returns:
A flat list of tensors, one per `self.components`.
"""
with ops.device(self._name):
return tpu_ops.tpu_replicated_output(
parallel_tensor, num_replicas=len(self.components))