Peter Hawkins [Sat, 23 Dec 2017 16:59:26 +0000 (08:59 -0800)]
[TF:XLA] Refactor large list of hard-coded compile-time constant arguments to operators. Add a new .CompileTimeConstInput() annotation on kernel registrations instead.
PiperOrigin-RevId:
180008567
A. Unique TensorFlower [Sat, 23 Dec 2017 12:22:33 +0000 (04:22 -0800)]
Automated g4 rollback of changelist
179983419
PiperOrigin-RevId:
180000981
A. Unique TensorFlower [Sat, 23 Dec 2017 03:28:38 +0000 (19:28 -0800)]
Adds FFT for XLA: CPU via Eigen, GPU via cuFFT.
GPU support includes plan reuse with new scratch allocator per execution in fft_thunk.
PiperOrigin-RevId:
179983419
Yuefeng Zhou [Sat, 23 Dec 2017 02:37:27 +0000 (18:37 -0800)]
Fill the 4th and 5th output of FusedBatchNormGradOp with zeros so that check_op won't fail.
PiperOrigin-RevId:
179981448
Jianwei Xie [Sat, 23 Dec 2017 01:37:51 +0000 (17:37 -0800)]
Adds train_hooks into _TrainingExecutor.
PiperOrigin-RevId:
179978502
A. Unique TensorFlower [Sat, 23 Dec 2017 01:37:12 +0000 (17:37 -0800)]
Remove a few unused constructions and simplfy some code
PiperOrigin-RevId:
179978470
A. Unique TensorFlower [Sat, 23 Dec 2017 01:25:41 +0000 (17:25 -0800)]
Go: Update generated wrapper functions for TensorFlow ops.
PiperOrigin-RevId:
179977826
A. Unique TensorFlower [Sat, 23 Dec 2017 01:07:23 +0000 (17:07 -0800)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
179976854
Sergio Guadarrama [Sat, 23 Dec 2017 01:05:51 +0000 (17:05 -0800)]
Clarify Eager Metrics docs.
Add Metric.value()
PiperOrigin-RevId:
179976743
Jiri Simsa [Sat, 23 Dec 2017 00:42:18 +0000 (16:42 -0800)]
Changing tf.contrib.data.dense_to_sparse_batch to return tf.SparseTensor.
PiperOrigin-RevId:
179975122
Alexandre Passos [Sat, 23 Dec 2017 00:30:04 +0000 (16:30 -0800)]
Produces RunMetadata from eager execution when tracing is explicitly enabled.
This will allow downstream profiling tools to work with eager execution.
PiperOrigin-RevId:
179974238
Alexandre Passos [Sat, 23 Dec 2017 00:29:52 +0000 (16:29 -0800)]
Allow ImportGraphDef to ignore shape validation to match semantics of ConvertGraphDefToGraph optionally.
PiperOrigin-RevId:
179974220
Benoit Steiner [Sat, 23 Dec 2017 00:27:21 +0000 (16:27 -0800)]
Shard and turn on the layout optimizer test
Tag the dataset test
PiperOrigin-RevId:
179974079
A. Unique TensorFlower [Fri, 22 Dec 2017 23:50:19 +0000 (15:50 -0800)]
Replaces custom _lengths_to_masks function with the official, more efficient sequence_mask function that supersedes it.
PiperOrigin-RevId:
179971521
Michael Case [Fri, 22 Dec 2017 23:44:29 +0000 (15:44 -0800)]
Run gen_git_source.py inside of a repo_rule instead of configure.
PiperOrigin-RevId:
179971055
Derek Murray [Fri, 22 Dec 2017 23:39:18 +0000 (15:39 -0800)]
Avoid unnecessary copying and allocation in `ExecutorBarrier::WhenDone()`.
Previously the code would make a copy of the `done_cb_` and update the status
on a successful completion, neither of which is necessary, and both of which
showed up in trace amounts in profiling.
PiperOrigin-RevId:
179970701
A. Unique TensorFlower [Fri, 22 Dec 2017 23:34:13 +0000 (15:34 -0800)]
Adding page to tensorflow.org with directions for building the TFLite demo on Android.
PiperOrigin-RevId:
179970218
A. Unique TensorFlower [Fri, 22 Dec 2017 23:23:01 +0000 (15:23 -0800)]
[XLA] Hlo parser: update readme.
PiperOrigin-RevId:
179969357
A. Unique TensorFlower [Fri, 22 Dec 2017 23:22:36 +0000 (15:22 -0800)]
Output of a slice op can alias its operand.
PiperOrigin-RevId:
179969317
A. Unique TensorFlower [Fri, 22 Dec 2017 23:05:24 +0000 (15:05 -0800)]
Make the analytical estimator and the virtual cluster own the ready_node_manager to be used by the virtual_scheduler. This allows us to customize the ready_node_manager for each analyzer instance.
PiperOrigin-RevId:
179967778
Jianwei Xie [Fri, 22 Dec 2017 22:49:14 +0000 (14:49 -0800)]
Extracts the tpu_job_name from TF_CONFIG if set.
PiperOrigin-RevId:
179966184
Ian Langmore [Fri, 22 Dec 2017 22:33:44 +0000 (14:33 -0800)]
Use assert_near in mvn_full_covariance to assert symmetric.
PiperOrigin-RevId:
179964717
A. Unique TensorFlower [Fri, 22 Dec 2017 22:25:36 +0000 (14:25 -0800)]
Automated g4 rollback of changelist
179846085
PiperOrigin-RevId:
179963793
A. Unique TensorFlower [Fri, 22 Dec 2017 22:21:11 +0000 (14:21 -0800)]
Fix documentation error in tensorflow feature_utils.h
Need to take address of the example for modifying its feature
PiperOrigin-RevId:
179963317
Yao Zhang [Fri, 22 Dec 2017 22:10:56 +0000 (14:10 -0800)]
Support non-constant param input of AvgPoolGrad and Sum.
PiperOrigin-RevId:
179962212
Derek Murray [Fri, 22 Dec 2017 22:01:11 +0000 (14:01 -0800)]
Fix typo in direct_session_test.cc.
PiperOrigin-RevId:
179960797
Shanqing Cai [Fri, 22 Dec 2017 21:54:11 +0000 (13:54 -0800)]
[TFE] Add link to other exapmles from spinn/README.md
PiperOrigin-RevId:
179960166
A. Unique TensorFlower [Fri, 22 Dec 2017 21:31:15 +0000 (13:31 -0800)]
Adds more useful debug information to an error message when no kernels are found.
PiperOrigin-RevId:
179957977
A. Unique TensorFlower [Fri, 22 Dec 2017 21:05:26 +0000 (13:05 -0800)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
179955796
A. Unique TensorFlower [Fri, 22 Dec 2017 20:55:54 +0000 (12:55 -0800)]
Go: Update generated wrapper functions for TensorFlow ops.
PiperOrigin-RevId:
179954704
A. Unique TensorFlower [Fri, 22 Dec 2017 20:42:59 +0000 (12:42 -0800)]
Merge changes from github.
PiperOrigin-RevId:
179953488
Benoit Steiner [Fri, 22 Dec 2017 20:37:14 +0000 (12:37 -0800)]
Updated the virtual cluster to return the proper error code if the simulated
peak memory usage exceeds the available memory.
PiperOrigin-RevId:
179952918
Alexandre Passos [Fri, 22 Dec 2017 20:21:16 +0000 (12:21 -0800)]
Uses ImportGraphDef instead of ConvertGraphDefToGraph in tflite.
This has the right semantics for backwards compatibility.
PiperOrigin-RevId:
179951639
Jiri Simsa [Fri, 22 Dec 2017 19:40:43 +0000 (11:40 -0800)]
Adding performance guide for the `tf.data` API.
PiperOrigin-RevId:
179947300
Ian Langmore [Fri, 22 Dec 2017 19:28:19 +0000 (11:28 -0800)]
assert_near added to check_ops.py
PiperOrigin-RevId:
179946066
Peter Hawkins [Fri, 22 Dec 2017 19:20:43 +0000 (11:20 -0800)]
[TF:XLA] Implement AdjustContrastv2, AdjustHue, AdjustSaturation, HSVToRGB, RGBToHSV in XLA bridge.
PiperOrigin-RevId:
179945310
Jiri Simsa [Fri, 22 Dec 2017 19:04:18 +0000 (11:04 -0800)]
Fixing tf.contrib.data API declarations.
PiperOrigin-RevId:
179943582
Peter Hawkins [Fri, 22 Dec 2017 18:53:34 +0000 (10:53 -0800)]
[XLA] Remove Power architecture support from CPU backend.
The Power support is unused, untested, and buggy. Remove it until it is fixed.
PiperOrigin-RevId:
179942340
A. Unique TensorFlower [Fri, 22 Dec 2017 18:35:05 +0000 (10:35 -0800)]
In FirstReady node manager, use node name as tie-breaker when multiple nodes
have same time_ready.
PiperOrigin-RevId:
179940344
Matthew Schulkind [Fri, 22 Dec 2017 18:32:27 +0000 (10:32 -0800)]
Fix allow_smaller_final_batches for bucket_by_sequence_length.
Closes #14420.
PiperOrigin-RevId:
179940036
A. Unique TensorFlower [Fri, 22 Dec 2017 18:32:07 +0000 (10:32 -0800)]
Adds support in stream executor interface to update the scratch allocator used with a cuFFT plan. This enables plan reuse without requiring we keep the scratch allocation alive between executions.
PiperOrigin-RevId:
179939994
Benoit Steiner [Fri, 22 Dec 2017 18:29:16 +0000 (10:29 -0800)]
Extract placement restrictions for a graph
PiperOrigin-RevId:
179939679
Justin Lebar [Fri, 22 Dec 2017 18:18:18 +0000 (10:18 -0800)]
[XLA:GPU] Fix implementation of Thunk::ShouldBlockFutureThunks.
The old implementation did not do what it was intended to do. Instead
of making the next thunk on each other stream wait for this thunk to
complete, it made all following thunks on this stream wait for this
thunk to complete. (Which is redundant, since streams are serialized
anyway.)
PiperOrigin-RevId:
179938424
Yao Zhang [Fri, 22 Dec 2017 17:44:32 +0000 (09:44 -0800)]
No need to and don't add layout transform to control nodes.
PiperOrigin-RevId:
179934839
Derek Murray [Fri, 22 Dec 2017 17:24:31 +0000 (09:24 -0800)]
Mark the Arg and RetVal ops as inexpensive.
This change avoids gratuitous context switches at the beginning of a
step (for each fed value) and end of a step (for each fetched
value). These OpKernels are trivial, so the cost of executing it
inline is typically smaller than waking up another thread to perform
the work.
PiperOrigin-RevId:
179933412
A. Unique TensorFlower [Fri, 22 Dec 2017 15:40:11 +0000 (07:40 -0800)]
Add comments to link convolutional factors code with definitions in the KFC paper.
PiperOrigin-RevId:
179925679
A. Unique TensorFlower [Fri, 22 Dec 2017 11:07:51 +0000 (03:07 -0800)]
Added an option to download a fresh release of clang for doing cuda_clang builds.
Tested only on Linux, but should also work for Mac.
No support for Windows is available yet.
PiperOrigin-RevId:
179910980
A. Unique TensorFlower [Fri, 22 Dec 2017 08:07:40 +0000 (00:07 -0800)]
[XLA] Support conditional in all backends.
PiperOrigin-RevId:
179900775
Igor Saprykin [Fri, 22 Dec 2017 07:43:57 +0000 (23:43 -0800)]
Support multiple loss and multiple optimizers in replicate_model_fn.
Instead of supplying `optimizer_fn`, the user is now expected to wrap their optimizer in GatheringOptimizer. The latter will gather gradients, reduce and apply them.
There can be multiple instances of GatheringOptimizer inside the model.
PiperOrigin-RevId:
179899422
Benoit Steiner [Fri, 22 Dec 2017 05:49:22 +0000 (21:49 -0800)]
Made hard colocation constraints (i.e constraints that must be met for the
model to be executable) available from python
PiperOrigin-RevId:
179892785
Justin Lebar [Fri, 22 Dec 2017 05:42:12 +0000 (21:42 -0800)]
[XLA] Print out missing extra-info for many instructions in the HLO graph dumper.
Now we use the same functionality as HloInstruction::ToString() to print
instructions' extra info. This fills in a lot of previously-missing
info, like reduce-windows' windows, and dots' dot-dimension-numbers.
PiperOrigin-RevId:
179892469
A. Unique TensorFlower [Fri, 22 Dec 2017 04:24:37 +0000 (20:24 -0800)]
[XLA] Update the parsing and stringfication of the Dot dimension numbers.
PiperOrigin-RevId:
179888764
Justin Lebar [Fri, 22 Dec 2017 04:23:40 +0000 (20:23 -0800)]
[XLA] Add zero-sized batchnorm test.
Also fix out-of-bounds read in ReferenceUtil::Reduce4DTo1D.
PiperOrigin-RevId:
179888724
Yao Zhang [Fri, 22 Dec 2017 04:10:02 +0000 (20:10 -0800)]
Support fill op.
PiperOrigin-RevId:
179888184
Justin Lebar [Fri, 22 Dec 2017 03:11:50 +0000 (19:11 -0800)]
[XLA] Run ZeroSizedHloElimination after BatchNormExpander.
BatchNormExpander can create zero-sized HLOs even if it's run after
ZeroSizedHloElimination. For example, you might have a
BatchNormTraining op with shape (f32[0, 10], f32[10], f32[10]). This
won't be eliminated by ZeroSizedHloElimination, but when it's
decomposed, it will result in many zero-sized HLOs.
PiperOrigin-RevId:
179885417
Justin Lebar [Fri, 22 Dec 2017 02:27:42 +0000 (18:27 -0800)]
[XLA:GPU] Implement BatchNormThunk as a call into cudnn.
Using cudnn for these calls is disabled by default, because it's not a
performance win on our benchmarks.
PiperOrigin-RevId:
179882911
Anna R [Fri, 22 Dec 2017 02:19:25 +0000 (18:19 -0800)]
Automated g4 rollback of changelist
179872807
PiperOrigin-RevId:
179882393
Dandelion Man? [Fri, 22 Dec 2017 02:09:03 +0000 (18:09 -0800)]
No external change.
PiperOrigin-RevId:
179881638
Anna R [Fri, 22 Dec 2017 01:51:46 +0000 (17:51 -0800)]
Internal change.
PiperOrigin-RevId:
179880234
Jiri Simsa [Fri, 22 Dec 2017 01:34:47 +0000 (17:34 -0800)]
Adding a test that checks that Grappler is able to infer shapes for core `tf.data` API.
PiperOrigin-RevId:
179878973
Benoit Steiner [Fri, 22 Dec 2017 01:31:11 +0000 (17:31 -0800)]
Added a comment to explain why we can't reorder BiasAdd
PiperOrigin-RevId:
179878687
Derek Murray [Fri, 22 Dec 2017 00:58:27 +0000 (16:58 -0800)]
Add test coverage for example parsing with a feature that is an empty string.
PiperOrigin-RevId:
179875758
Suharsh Sivakumar [Fri, 22 Dec 2017 00:49:17 +0000 (16:49 -0800)]
Fold FuseBatchNorm correctly.
PiperOrigin-RevId:
179874901
Benoit Steiner [Fri, 22 Dec 2017 00:27:19 +0000 (16:27 -0800)]
Extract placement restrictions for a graph
PiperOrigin-RevId:
179872807
A. Unique TensorFlower [Fri, 22 Dec 2017 00:24:45 +0000 (16:24 -0800)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
179872583
Blake Hechtman [Fri, 22 Dec 2017 00:05:21 +0000 (16:05 -0800)]
Automated g4 rollback of changelist
179858783
PiperOrigin-RevId:
179870508
Justin Lebar [Thu, 21 Dec 2017 23:57:09 +0000 (15:57 -0800)]
[XLA] Transform division-by-fp-constant to multiplication by its inverse.
Backends can do this transformation, but in general they won't if the
divisor is not a constant. In addition, this formulation allows more
aggressive fusion.
PiperOrigin-RevId:
179869591
Anna R [Thu, 21 Dec 2017 23:29:21 +0000 (15:29 -0800)]
Internal change.
PiperOrigin-RevId:
179866603
A. Unique TensorFlower [Thu, 21 Dec 2017 23:20:42 +0000 (15:20 -0800)]
Only squeeze the last dimension of outputs and indices in infer_graph of
kmeans.
PiperOrigin-RevId:
179865588
Jiri Simsa [Thu, 21 Dec 2017 23:06:32 +0000 (15:06 -0800)]
Saving a couple of us in IteratorGetNextOp::ComputeAsync.
PiperOrigin-RevId:
179863871
A. Unique TensorFlower [Thu, 21 Dec 2017 22:50:09 +0000 (14:50 -0800)]
Automated g4 rollback of changelist
179782067
PiperOrigin-RevId:
179861781
A. Unique TensorFlower [Thu, 21 Dec 2017 22:27:14 +0000 (14:27 -0800)]
Merged commit includes the following changes:
179858783 by gunan:
[XLA] Add a pass to replace all non-side-effecting and removable zero sized
HLOs.
--
179853916 by A. Unique TensorFlower:
Fix missing backtick in parse_example docstring, causing wrong markdown formatting.
--
179852823 by apassos:
Another attempt at adding int64 indices to Fill.
--
179852553 by A. Unique TensorFlower:
Stop grappler optimization to reorder BiasAdd ops.
--
179850703 by yaozhang:
Support reverse op.
--
179850204 by blakehechtman:
[XLA] Add a pass to replace all non-side-effecting and removable zero sized
HLOs.
--
PiperOrigin-RevId:
179858783
Gunhan Gulsoy [Thu, 21 Dec 2017 21:02:21 +0000 (13:02 -0800)]
Automated g4 rollback of changelist
179836027
PiperOrigin-RevId:
179848421
Akshay Agrawal [Thu, 21 Dec 2017 20:39:52 +0000 (12:39 -0800)]
Make it possible to wrap tf.layers.Layer's 'call' method in tfe.defun.
This change wraps Layer's `build` method in an `init_scope`, which has the effect of lifting all ops created in `build` out of function-building graphs and control flow scopes; the `init_scope` also pauses the gradient tape.
PiperOrigin-RevId:
179846085
Benoit Steiner [Thu, 21 Dec 2017 20:26:30 +0000 (12:26 -0800)]
Tagged the grappler python tests
PiperOrigin-RevId:
179844728
Mark Daoust [Thu, 21 Dec 2017 20:22:19 +0000 (12:22 -0800)]
Add links to low level API intro
PiperOrigin-RevId:
179844300
Akshay Agrawal [Thu, 21 Dec 2017 19:23:20 +0000 (11:23 -0800)]
This change adds a mechanism to the internal C API for updating an output handle's shapes and types after its source operation has been created.
Context: framework/ops.py was recently updated to use the C API when setting shapes for an op's outputs. This update broke shape inference for graph functions that captured resource handles; this, in turn, made it impossible to create graph functions from Python methods that required fully defined shapes (e.g., like MNIST's `call` method). In particular, the C API computes shapes for ops when they are created and does not update them thereafter; this is problematic because when a resource handle is captured while building a function, we need to update the captured operation's output handle in order to propagate its outputs shapes and dtypes.
PiperOrigin-RevId:
179837104
Jingyue Wu [Thu, 21 Dec 2017 19:21:22 +0000 (11:21 -0800)]
Fix padding for int8 fused convolution.
cudnnConvolutionBiasActivationForward doesn't work when filter size >= 6 and
there is padding. This CL works around this cudnn bug.
PiperOrigin-RevId:
179836819
A. Unique TensorFlower [Thu, 21 Dec 2017 19:15:26 +0000 (11:15 -0800)]
Adding support for to resolve constant FloorDiv, FloorMod, StridedSlice, Stack, Rank and Range.
PiperOrigin-RevId:
179836027
A. Unique TensorFlower [Thu, 21 Dec 2017 19:11:22 +0000 (11:11 -0800)]
[XLA] Make dumped_computation_to_text print the short parsable hlo text.
PiperOrigin-RevId:
179835513
Jianwei Xie [Thu, 21 Dec 2017 19:08:43 +0000 (11:08 -0800)]
Adds eval session master into TF_CONFIG
PiperOrigin-RevId:
179835208
Eugene Brevdo [Thu, 21 Dec 2017 17:21:59 +0000 (09:21 -0800)]
[TFE] Return a byte-encoded value for tfe.defun's name
This is necessary when the function object is passed to an op constructor;
the op's func attribute is a protobuf field and must be bytes in python3.
PiperOrigin-RevId:
179822007
Ian Langmore [Thu, 21 Dec 2017 17:04:11 +0000 (09:04 -0800)]
sparse_ops: Static shape update.
* .sparse_reshape: Set static shape on output when shape includes inferred dims.
* .transpose: Set static shape if all shapes are known.
PiperOrigin-RevId:
179820175
Shanqing Cai [Thu, 21 Dec 2017 16:35:19 +0000 (08:35 -0800)]
tfdbg: allow writing tensor values to filesystem as numpy files
Command example:
tfdbg> pt Softmax:0 -w /tmp/softmax.npy
tfdbg> eval 'np.sum(`Softmax:0`, axis=1)' -w /tmp/softmax_sum.npy
The saved file can later be loaded in Python:
>> import numpy as np
>> np.load('/tmp/softmax.npy')
>> np.load('/tmp/softmax_sum.npy')
RELNOTES: For `pt` and `eval` commands, allow writing tensor values to filesystem as numpy files.
PiperOrigin-RevId:
179817453
A. Unique TensorFlower [Thu, 21 Dec 2017 14:49:17 +0000 (06:49 -0800)]
Implement Clone for kSend,kRecv,kSendDone and kRecvDone.
PiperOrigin-RevId:
179808928
Mark Daoust [Thu, 21 Dec 2017 14:23:46 +0000 (06:23 -0800)]
Modernize old "get_started/get_started.md", as "programmers_guide/low_level_intro.md".
PiperOrigin-RevId:
179807033
A. Unique TensorFlower [Thu, 21 Dec 2017 08:33:01 +0000 (00:33 -0800)]
Roll CL
177989542 forward with fix: Wrappers for CUDA 9 warp-synchronous intrinsics.
PiperOrigin-RevId:
179782067
Eugene Brevdo [Thu, 21 Dec 2017 07:02:11 +0000 (23:02 -0800)]
[tf nest] Add additional key yielder.
This is just copying a utility function created by Malcolm Reynolds.
PiperOrigin-RevId:
179775504
Derek Murray [Thu, 21 Dec 2017 07:01:10 +0000 (23:01 -0800)]
Revert change to ParseBytesList that failed on some platforms.
PiperOrigin-RevId:
179775354
A. Unique TensorFlower [Thu, 21 Dec 2017 05:15:44 +0000 (21:15 -0800)]
Go: Update generated wrapper functions for TensorFlow ops.
PiperOrigin-RevId:
179769852
Anna R [Thu, 21 Dec 2017 05:06:48 +0000 (21:06 -0800)]
Internal change.
PiperOrigin-RevId:
179769382
A. Unique TensorFlower [Thu, 21 Dec 2017 05:04:31 +0000 (21:04 -0800)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
179769207
Derek Murray [Thu, 21 Dec 2017 04:54:25 +0000 (20:54 -0800)]
[tf.data] Add `tf.contrib.data.parse_single_example()`.
The new op is a fused implementation of the existing
`tf.parse_single_example()`, which is more efficient when parsing a
single Example at a time.
PiperOrigin-RevId:
179768512
Anna R [Thu, 21 Dec 2017 04:17:15 +0000 (20:17 -0800)]
Automated g4 rollback of changelist
179757188
PiperOrigin-RevId:
179766237
A. Unique TensorFlower [Thu, 21 Dec 2017 03:01:09 +0000 (19:01 -0800)]
Go: Update generated wrapper functions for TensorFlow ops.
PiperOrigin-RevId:
179761889
Anna R [Thu, 21 Dec 2017 02:59:24 +0000 (18:59 -0800)]
Automated g4 rollback of changelist
179735577
PiperOrigin-RevId:
179761787
A. Unique TensorFlower [Thu, 21 Dec 2017 01:59:07 +0000 (17:59 -0800)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
179757188
Chris Leary [Thu, 21 Dec 2017 01:55:13 +0000 (17:55 -0800)]
[XLA] Add CrossReplicaSum to Python API bindings.
Add support to CPU for the trivial replica_count=1 case.
PiperOrigin-RevId:
179756956
Justin Lebar [Thu, 21 Dec 2017 01:54:39 +0000 (17:54 -0800)]
[XLA] Fix HLO graph dumper not to assume that instruction names start with "%".
HLO graph dumper needs to be aware that we've gotten rid of the "%"
prefix in HLO names so it doesn't print e.g.
reduce
reduce.42
Subcomputation: add
...
but instead simply prints
reduce.42
Subcomputation: add
...
PiperOrigin-RevId:
179756922
A. Unique TensorFlower [Thu, 21 Dec 2017 01:53:22 +0000 (17:53 -0800)]
Fix typo in struct name: RecvNodeDescritorHash -> RecvNodeDescriptorHash.
PiperOrigin-RevId:
179756840