aboutsummaryrefslogtreecommitdiffhomepage
Commit message (Collapse)AuthorAge
...
* | | | | | | | | | | Create new interface for specifying TensorFlow third party importsGravatar Austin Anderson2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208913262
* | | | | | | | | | | Fix error message formatting.Gravatar Sanjoy Das2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208911623
* | | | | | | | | | | Automated rollback of commit 2a8e827c87149c49a1c47da010ee40f50a904055Gravatar Alexandre Passos2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208911179
* | | | | | | | | | | The code for splitting constants to ensure single user had a side effect of ↵Gravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | deleting constants which did not have any users. It also made an extra clone when number of users reduced to one. While it did not affect correctness, but it's cleaner and more efficient to address these issues. PiperOrigin-RevId: 208910596
* | | | | | | | | | | Internal changeGravatar Jared Duke2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208910402
* | | | | | | | | | | Implmenet cuDNN algorithm cross checking.Gravatar Tim Shen2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208910038
* | | | | | | | | | | Enable f64 convolutions for GPU backend. Currently, all layouts are NCHWs.Gravatar Tim Shen2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208908539
* | | | | | | | | | | NEON optimizations for most common-Mul uint8 kernel.Gravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208907698
* | | | | | | | | | | [tf.contrib.lookup] More cleanups and add some simple benchmarks.Gravatar Eugene Brevdo2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208906068
* | | | | | | | | | | TOCO Bugfix: Do not transform Conv() to DepthwiseConv() if weights is sharedGravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | with another op. PiperOrigin-RevId: 208903223
* | | | | | | | | | | Make sure tensorrt and nccl kernels compile with newer versions of Eigen.Gravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208902416
* | | | | | | | | | | Automated rollback of commit 6b13a15e0b9906abbb66f87d83db291d0099cb43Gravatar Gunhan Gulsoy2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208896224
* | | | | | | | | | | Fix the MKL build by correcting misspelled references to MKL's build package.Gravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208894550
* | | | | | | | | | | Bump gtest version to latest head.Gravatar Justin Lebar2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208893658
* | | | | | | | | | | Introduce and use an "and recurrence" in deadness analysis to represent loops.Gravatar Sanjoy Das2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | And recurrences denote an infinite list of predicates and thus can be used to represent the control dependence of loop bodies. This lets us prove that nodes like "Add" (which have all inputs controlled by a single controlling loop "CtrlLoopCond") below can be clustered. digraph G { _SRC -> CtrlInit _SRC -> Init0 _SRC -> Init1 CtrlInit -> CtrlMerge CtrlMerge -> CtrlLoopCond CtrlLoopCond -> CtrlSwitch CtrlSwitch -> "CtrlSwitch+1" "CtrlSwitch+1" -> CtrlMerge CtrlLoopCond -> CtrlMerge Init0 -> Merge0 Merge0 -> Switch0 Switch0 -> "Switch0+1" "Switch0+1" -> NextIteration0 NextIteration0 -> Merge0 CtrlLoopCond -> Switch0 Init1 -> Merge1 Merge1 -> Switch1 Switch1 -> "Switch1+1" "Switch1+1" -> NextIteration1 NextIteration1 -> Merge1 CtrlLoopCond -> Switch1 } This graph can be generated from Python code like this (thanks for skyewm@ for reverse engineering this!): _, x, y = tf.while_loop( lambda ctrl, x, y: ctrl < 10, lambda ctrl, x, y: ctrl + 1, x + 1, y + 1, [0, 0, 0]) add = x + y PiperOrigin-RevId: 208891426
* | | | | | | | | | | Add description on how `sparse_reduce_max` behaves with respect to the ↵Gravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | missing values in a SparseTensor. Also added an extra example. PiperOrigin-RevId: 208879761
* | | | | | | | | | | Fix TensorFlow V2 test in api_compatibility_test and update goldens.Gravatar Anna R2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208878966
* | | | | | | | | | | Move params to kernels/internal/types.h.Gravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208877784
* | | | | | | | | | | Ops API to author TFLite ops as functions directly.Gravatar Yu-Cheng Ling2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208875580
* | | | | | | | | | | * Fixing checks for fully-known shapes.Gravatar Andrew Cotter2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | * Adding pre_train_ops to ConstrainedMinimizationProblem. These are tf.Operations that should be executed before each training step. * Updating ConstrainedOptimizers to have the train_ops they create depend on the pre_train_ops from the ConstrainedMinimizationProblem. PiperOrigin-RevId: 208871419
* | | | | | | | | | | Support multiple index-space constraints in synthetic input generator.Gravatar Mark Heffernan2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208868489
* | | | | | | | | | | Avoid copying Tensor when Select op takes scalar cond input (CPU).Gravatar Sung Jin Hwang2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208868429
* | | | | | | | | | | Automated rollback of commit e045fd284075001e7f7585c581c4444e4e850534Gravatar Andrew Selle2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208868027
* | | | | | | | | | | [XLA] The a cross-module all-reduce is a global predecessor of an hlo,Gravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | the peer all-reduces should also be global predecessors. Same for global successors. PiperOrigin-RevId: 208865653
* | | | | | | | | | | Always use a fresh graph in run_in_graph_and_eager_modesGravatar Alexandre Passos2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208865526
* | | | | | | | | | | Handle aggregate arguments using op_hints.Gravatar Andrew Selle2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | - Create a pack/unstack transformation "rule". - Handle aggregates with tags that are aggregated against. - Make convert_to_stubs use graph_defs rather than just sessions. PiperOrigin-RevId: 208865349
* | | | | | | | | | | Simplify the boosted trees control flow.Gravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208864844
* | | | | | | | | | | Strip control dependencies in despecializer.Gravatar Mark Heffernan2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Control dependencies are an artifact left over from lowering. These should be removed when raising to a despecialized form. PiperOrigin-RevId: 208862729
* | | | | | | | | | | Merge pull request #20003 from samikama:BazelNoStripGravatar TensorFlower Gardener2018-08-15
|\ \ \ \ \ \ \ \ \ \ \ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208857243
* | | | | | | | | | | | [TF:XLA] Bump open source llvm revision to r339662Gravatar Sanjoy Das2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208856945
* | | | | | | | | | | | Add a virtual interface for the executor side of stats collection.Gravatar Derek Murray2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This will enable switching out the default stats collection mechanism (based on strings and protocol buffers) for a lighter-weight implementation that can be used in other settings. PiperOrigin-RevId: 208851452
* | | | | | | | | | | | Convert more kernel signatures to use runtime shapes.Gravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208850456
* | | | | | | | | | | | Allows feature inputs to be arbitrary nested tuples in TPUEstimator which is ↵Gravatar Olivia Nordquist2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | allowed behavior in tf.Estimator NB: Calls data/util/nest.flatten() which is slightly different than core/util/nest (already a dependency in tpu_estimator.py) as it maintains structure by treating tuples and lists differently PiperOrigin-RevId: 208848267
* | | | | | | | | | | | Small fix to MaybeGetMinimumShape() in op_level_cost_estimator.Gravatar Doe Hyun Yoon2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | (1) previously, it set unknown shape flag for scalar input, but now it returns TensorShapeProto with rank equal to the expected and all dims set to 1, and unknown shape flag is not set. (2) Also, fixed a bug; when a rank is known, but dim_size() < rank (note that dim_size() may be non-zero), we previously called add_dim() with dim 1 rank times, which then makes dim_size() is incremented by rank, but we expect dim_size() equal to rank. (3) Added test for MaybeGetMinimumShape(). PiperOrigin-RevId: 208845501
* | | | | | | | | | | | Dead code elimination.Gravatar Akshay Agrawal2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208843066
* | | | | | | | | | | | Run buildifier on tensorflow.bzlGravatar Gunhan Gulsoy2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208838993
* | | | | | | | | | | | [XLA] Fix a bug in scatter test.Gravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208836927
| | | | | * | | | | | | fix Ubuntu Sanity check failureGravatar weidankong2018-08-15
| | | | | | | | | | | |
* | | | | | | | | | | | Add batch_gather that assumes a series of leading (dense) dimensions.Gravatar Adria Puigdomenech2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This operation computes: result[i1, ..., in] = params[i1, ..., in-1, indices[i1, ..., in]] That is, it assumes that `params` and `indices` have n-1 leading dense dimensions, followed by a sparse dimensions over which we want to gather. PiperOrigin-RevId: 208836509
* | | | | | | | | | | | Fix toco execution on WindowsGravatar Jared Duke2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Provide the necessary flags for reading/writing binary protobuf/flatbuffer files. With this fix, toco behavior should now be consistent with other platforms. This fixes #21085. PiperOrigin-RevId: 208834940
* | | | | | | | | | | | Split unit test case for further parallelize.Gravatar Scott Zhu2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208834783
* | | | | | | | | | | | Remove unused Estimator overrides.Gravatar Jianwei Xie2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208834294
* | | | | | | | | | | | Improve the performance of ShapeUtil::GetLeafCountGravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The previous implementation used a very generic helper function what added a large amount of overhead to a very simple operation. PiperOrigin-RevId: 208799040
* | | | | | | | | | | | compat: Update forward compatibility horizon to 2018-08-15Gravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208786904
* | | | | | | | | | | | Update ops-related pbtxt files.Gravatar A. Unique TensorFlower2018-08-15
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208783215
* | | | | | | | | | | | Update community page with more specific links to security advisories and ↵Gravatar Frank Chen2018-08-14
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | other relevant security links. PiperOrigin-RevId: 208775844
* | | | | | | | | | | | Automated rollback of commit a00fba38681fa28d6d20cd730fe362864b819a0dGravatar Gunhan Gulsoy2018-08-14
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | PiperOrigin-RevId: 208775205
* | | | | | | | | | | | Change the initialization of the mean squared gradient variable of the ↵Gravatar A. Unique TensorFlower2018-08-14
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | optimizer_v2 version of RMSProp. PiperOrigin-RevId: 208774314
* | | | | | | | | | | | Propagate minmax information across input and outputs of Reshape.Gravatar Suharsh Sivakumar2018-08-14
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Since reshape shouldn't change minmax info. PiperOrigin-RevId: 208772357
* | | | | | | | | | | | The rewriter should not add nodes if followed pass-through op and then a ↵Gravatar Suharsh Sivakumar2018-08-14
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | FakeQuant PiperOrigin-RevId: 208767575
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-09 20:28:07 +0000