contrib.nccl.broadcast
tf.contrib.nccl.broadcast
tf.contrib.nccl.broadcast
broadcast( src_tensor, dst_devices )
Defined in tensorflow/contrib/nccl/python/ops/nccl_ops.py
.
Returns a list of tensors on dst_devices
, each with value tensor
.
The computation is done with a broadcast nccl operation, so if only some of the returned tensors and src_tensor are evaluated then the computation will hang.
Args:
-
src_tensor
: The tensor to send; must be assigned to a GPU device. -
dst_devices
: The GPU devices to receive the sent tensor.
Returns:
List of tensors, each with the value of src_tensor
, which the device of tensor i is dst_devices[i]
.
© 2017 The TensorFlow Authors. All rights reserved.
Licensed under the Creative Commons Attribution License 3.0.
Code samples licensed under the Apache 2.0 License.
https://www.tensorflow.org/api_docs/python/tf/contrib/nccl/broadcast