projects / platform / upstream / tensorflow.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 36d77e8) | patch
[TF:XLA] Add direct implementation of AssignVariableOp for XLA devices.
authorPeter Hawkins <phawkins@google.com>
Sat, 26 May 2018 00:29:37 +0000 (17:29 -0700)
committerTensorFlower Gardener <gardener@tensorflow.org>
Sat, 26 May 2018 00:32:18 +0000 (17:32 -0700)
commit5ef609b6e542dc1e3f0eaf195a1f8d8d4e7ff8af
treeea3e3298bb5f1e7a2a045bf2dc2e735bef1ac14dtree | snapshot
parent36d77e84af08ded6c9b964d900231f7d04d507a8commit | diff
[TF:XLA] Add direct implementation of AssignVariableOp for XLA devices.

This allows us to avoid an XLA compilation and tensor copies when assigning to a variable placed on an XLA device.

PiperOrigin-RevId: 198127062
tensorflow/compiler/jit/xla_device_context.cc diff | blob | history
tensorflow/compiler/jit/xla_device_context.h diff | blob | history
tensorflow/compiler/jit/xla_device_ops.cc diff | blob | history
tensorflow/compiler/jit/xla_device_ops.h diff | blob | history
tensorflow/compiler/tests/variable_ops_test.py diff | blob | history
tensorflow/compiler/tf2xla/kernels/variable_ops.cc diff | blob | history
Domain: Machine Learning / ML Framework;