aboutsummaryrefslogtreecommitdiffhomepage
Commit message (Collapse)AuthorAge
* [TF:XLA] Remove special base case from BatchDot that has been redundant ever ↵Gravatar Peter Hawkins2018-09-14
| | | | | | since xla::DotGeneral was added. PiperOrigin-RevId: 213052269
* Generalize TransformFilter method in preparation of NHWC Conv supportGravatar Smit Hinsu2018-09-14
| | | | PiperOrigin-RevId: 213049674
* Added TFE_OpSetAttrTensor() to eager C API.Gravatar Mingsheng Hong2018-09-14
| | | | | | | Also added some experimental C APIs for facilitate the use of eager C APIs in S4TF compiler. PiperOrigin-RevId: 213041780
* Disable flaky gpu_base_testGravatar Gunhan Gulsoy2018-09-14
| | | | PiperOrigin-RevId: 213040362
* Convert more kernel signatures to use runtime shapes.Gravatar A. Unique TensorFlower2018-09-14
| | | | PiperOrigin-RevId: 213037039
* Disable the flaky test case in timeline_testGravatar Gunhan Gulsoy2018-09-14
| | | | PiperOrigin-RevId: 213034078
* Populate custom name in registration.Gravatar Andrew Selle2018-09-14
| | | | PiperOrigin-RevId: 213028338
* Add --config=v2 option to the .bazelrc file.Gravatar Anna R2018-09-14
| | | | PiperOrigin-RevId: 213027176
* Make ReLU layer use nn.leaky_relu when appropriate.Gravatar Francois Chollet2018-09-14
| | | | PiperOrigin-RevId: 213026080
* Define PreferBlockAccess enum to prepare for Eigen upgrade.Gravatar Eugene Zhulenev2018-09-14
| | | | PiperOrigin-RevId: 213025676
* Automated rollback of commit 5f28bab20d303e9f815bbe8611c24b7f751e6f9eGravatar A. Unique TensorFlower2018-09-14
| | | | PiperOrigin-RevId: 213023382
* Updates to parameters, and to kernel helper functions.Gravatar A. Unique TensorFlower2018-09-14
| | | | PiperOrigin-RevId: 213023245
* [TF:XLA] Bump open source llvm revision to r342210Gravatar Sanjoy Das2018-09-14
| | | | PiperOrigin-RevId: 213022233
* Improve output parsing for unsupported opsGravatar Jared Duke2018-09-14
| | | | PiperOrigin-RevId: 213017532
* [tf.data] Introducing an optimization that parallelizes map transformations.Gravatar Piotr Padlewski2018-09-14
| | | | | | | | Stateless MapDatasets can be paralellized by switching to ParallelMapDataset. We set `num_parallel_calls` to 2 for now, but in the future a special value will be used that result in the optimal value to be selected dynamically at runtime. This patch also exposed a memory leak which was fixed. PiperOrigin-RevId: 213015223
* [XLA] Support strength reducing bfloat16 dot productsGravatar David Majnemer2018-09-14
| | | | | | | | There is no reason to limit ourselves to float32 dot product operations, we simply convert to and from float32 around the reduction to simulate the precision change. PiperOrigin-RevId: 213014410
* Convert more kernel signatures to use runtime shapes.Gravatar A. Unique TensorFlower2018-09-14
| | | | PiperOrigin-RevId: 213012717
* parallel_for: add a bunch of converters for cwise ops and gradients.Gravatar A. Unique TensorFlower2018-09-14
| | | | PiperOrigin-RevId: 213010458
* Removed unnecessary includes from stringpiece.hGravatar A. Unique TensorFlower2018-09-14
| | | | PiperOrigin-RevId: 213008707
* Point VectorDiffeomixture to tf.linalg rather than tf.contrib.linalgGravatar Ian Langmore2018-09-14
| | | | PiperOrigin-RevId: 213008118
* Convert more kernel signatures to use runtime shapes.Gravatar A. Unique TensorFlower2018-09-14
| | | | PiperOrigin-RevId: 213007905
* Fix archive pathGravatar Mark Daoust2018-09-14
| | | | PiperOrigin-RevId: 213007422
* Make tf.gradients() correctly handle captured EagerTensors.Gravatar Skye Wanderman-Milne2018-09-14
| | | | | | | | | | tf.gradients() can't be used to take the gradient of eager computations, but it should handle ops that take captured eager tensors as input as long as the gradient computation doesn't depend on that input. This change makes sure the gradient algorithm doesn't try to access the op, etc. of EagerTensors. PiperOrigin-RevId: 213007155
* Export tf.keras.sparse_categorical_accuracy. Copied from PR #21790.Gravatar Martin Wicke2018-09-14
| | | | | | closes #21790, fixes #21735 PiperOrigin-RevId: 213003724
* Update description of contrib.quantizeGravatar Raghuraman Krishnamoorthi2018-09-14
| | | | PiperOrigin-RevId: 212997520
* Internal change.Gravatar Nupur Garg2018-09-14
| | | | PiperOrigin-RevId: 212991181
* Fix bug preventing one from not specifying additional function attributes.Gravatar Lasse Espeholt2018-09-14
| | | | PiperOrigin-RevId: 212989480
* global_step/sec renamed to global_steps/secGravatar A. Unique TensorFlower2018-09-14
| | | | PiperOrigin-RevId: 212986442
* Run buildifier on build_defs.bzlGravatar Benjamin Kramer2018-09-14
| | | | PiperOrigin-RevId: 212972521
* [TF:XLA] Split XLA Concat Ops that fail on large sets of inputs.Gravatar A. Unique TensorFlower2018-09-14
| | | | | | Make the test large to prevent occasional timeouts on CPU. This should normally complete in well under a minute. PiperOrigin-RevId: 212953337
* [Grappler] s/std::string/string/Gravatar James Keeling2018-09-14
| | | | | | string and std::string are not necessarily the same thing in TF, but this code assumed that they are. PiperOrigin-RevId: 212952877
* compat: Update forward compatibility horizon to 2018-09-14Gravatar A. Unique TensorFlower2018-09-14
| | | | PiperOrigin-RevId: 212949973
* [XLA:TF] Make FloorDiv not crash on unsigned typesGravatar Benjamin Kramer2018-09-14
| | | | | | | | | | | FloorDiv (which corresponds to the // operator in python) supports uint8 and uint16 (but not uint32) in TF. Using xla::Abs on unsigned types throws an error, but the rounding logic is trivial for unsigned types so just do a plain Div. This isn't tested yet because we don't have any targets supporting uint8 or uint16 yet. PiperOrigin-RevId: 212946132
* Automated rollback of commit eb5cd6926ef8d2a5a748f1aa978e51148e22dd97Gravatar Gunhan Gulsoy2018-09-13
| | | | PiperOrigin-RevId: 212936412
* [GraphCompiler] Remove the use of XLA context as arugmnet.Gravatar Yunxing Dai2018-09-13
| | | | | | | - XLAContext is never used in the class, remove it from member list. - Be more clear in the comment that the result is written to the context from the compilation device. PiperOrigin-RevId: 212924213
* Add missing #include to mkl_layout_pass.cc.Gravatar A. Unique TensorFlower2018-09-13
| | | | PiperOrigin-RevId: 212921868
* Added Pyclif binding rule for config.proto.Gravatar Sung Jin Hwang2018-09-13
| | | | PiperOrigin-RevId: 212920113
* [XLA] Add hook for dump directory expansion.Gravatar Chris Leary2018-09-13
| | | | | | | | | Also puts a ".unoptimized" suffix on dumped HLO protobuf files to avoid the unoptimized dumped HLO protobuf colliding with the optimized dumped HLO protobufs when the same dump directory is specified for both. PiperOrigin-RevId: 212914100
* Make Keras relu use nn.leaky_relu when appropriate.Gravatar Francois Chollet2018-09-13
| | | | PiperOrigin-RevId: 212912615
* Fix performance issue when training with keras model in eager mode.Gravatar Pavithra Vijay2018-09-13
| | | | PiperOrigin-RevId: 212908218
* Simplify the initialization function in algortihm picker. No functional change.Gravatar Tim Shen2018-09-13
| | | | PiperOrigin-RevId: 212905536
* Reland "Add basic type propagation for unsupported ops in TFLite conversion"Gravatar Jared Duke2018-09-13
| | | | | | | The original CL was rolled back due to op registration conflicts in the pip. Resolve the issue by only including core:ops in the toco binary itself, not in intermediate libraries. PiperOrigin-RevId: 212902838
* Use `dataset.batch(.., drop_remainder=True)` instead of map_and_batch to ↵Gravatar Priya Gupta2018-09-13
| | | | | | achieve the same effect. PiperOrigin-RevId: 212901207
* [tf.data] Changes `make_batched_features_dataset` and ↵Gravatar Shivani Agrawal2018-09-13
| | | | | | `make_tf_record_dataset` default `prefetch` buffer size to auto-tune (from 1). PiperOrigin-RevId: 212900920
* Convert logdir paths to strings.Gravatar Alexandre Passos2018-09-13
| | | | | | This supports pathlib and other non-string path types. PiperOrigin-RevId: 212897666
* Ensure that the input image of decode.bmp.op has valid dimensions.Gravatar Mihai Maruseac2018-09-13
| | | | | | | This prevents an undefined behavior with signed integer overflow in decode.bmp.op. PiperOrigin-RevId: 212897289
* Automated rollback of commit ac60b46e2c5962fd8099a4406c1788d826ad3c0dGravatar A. Unique TensorFlower2018-09-13
| | | | PiperOrigin-RevId: 212896336
* [TF:XLA] Bump open source abseil revision to ↵Gravatar Sanjoy Das2018-09-13
| | | | | | 8ff1374008259719b54a8cb128ef951c02da164c PiperOrigin-RevId: 212893036
* Put a deprecation notice in cmake readme.Gravatar Gunhan Gulsoy2018-09-13
| | | | PiperOrigin-RevId: 212892844
* Fixing error output in api_compatibility_test.py. Looks like it should be ↵Gravatar Anna R2018-09-13
| | | | | | | | self.maxDiff instead of self.maxDiffs: "Diff is 2708 characters long. Set self.maxDiff to None to see it." PiperOrigin-RevId: 212892831
generated by cgit v1.2.3 (git 2.39.1) at 2024-09-07 13:24:13 +0000