aboutsummaryrefslogtreecommitdiffhomepage
Commit message (Collapse)AuthorAge
...
* [TF:XLA] Copy elision does not need to know about existing copies.Gravatar A. Unique TensorFlower2018-06-28
| | | | | | | | It already detects layout-changing copies and those are already left unchanged by copy elision. Special case copies are also skipped because they are tagged separately (SetCopyElisionAllowed) PiperOrigin-RevId: 202574858
* Import package xla_test instead of class XLATestCase.Gravatar A. Unique TensorFlower2018-06-28
| | | | PiperOrigin-RevId: 202572322
* Don't cache RNN weights in while loops since it's possible that train steps ↵Gravatar A. Unique TensorFlower2018-06-28
| | | | | | could be updating the weights. This is specifically true on TPUS (tpu.repeat). Also, fix the `testDynamicRnnTrainLoop` unit test. PiperOrigin-RevId: 202565323
* Change inputs from multi-dimensional arrays to ByteBuffer in TF Lite Object ↵Gravatar A. Unique TensorFlower2018-06-28
| | | | | | Detection app PiperOrigin-RevId: 202564164
* Automated g4 rollback of changelist 202292422Gravatar A. Unique TensorFlower2018-06-28
| | | | PiperOrigin-RevId: 202551122
* Fix Missing r-string.Gravatar Mark Daoust2018-06-28
| | | | | | | | | | The docs generator is not happy about the "\a" in "\approx" is becoming a "alert" escape sequence. This should also fix a lot of the mathjax rendering on this page: https://www.tensorflow.org/api_docs/python/tf/contrib/bayesflow/monte_carlo/expectation PiperOrigin-RevId: 202550662
* Adds support for Type III DCT, and tf.spectral.idct(type=2|3).Gravatar Brian Patton2018-06-28
| | | | PiperOrigin-RevId: 202546469
* Rewrite master_service in terms of more up-to-date gRPC APIsGravatar Noah Eisen2018-06-28
| | | | PiperOrigin-RevId: 202544091
* tfdbg: Fix compatibility with C++ MakeCallable and _make_callable_from_optionsGravatar Shanqing Cai2018-06-28
| | | | | | | Fixes #20160 REL_NOTES: tfdbg: Fix compatibility with `tf.keras.Model`s training on `tf.data.Dataset`s. PiperOrigin-RevId: 202543231
* TPUEstimator adds use_tpu key in params when calling model_fn. Model_fn can ↵Gravatar Yanan Cao2018-06-28
| | | | | | rely on this key to know whether TPU specific functionalities should be used. PiperOrigin-RevId: 202542458
* Automated g4 rollback of changelist 201419522Gravatar Tony Wang2018-06-28
| | | | PiperOrigin-RevId: 202539762
* Dedupe incompatible ops message.Gravatar Jacques Pienaar2018-06-28
| | | | | | No need to repeat an incompatible op time multiple times. Used set to ensure deterministic/same ordering in error message. PiperOrigin-RevId: 202534388
* Adds reset op as an optional return from update_stats.Gravatar A. Unique TensorFlower2018-06-28
| | | | PiperOrigin-RevId: 202528760
* [XLA] Change code in tensorflow/compiler/xla that uses XlaBuilder:: methods ↵Gravatar Peter Hawkins2018-06-28
| | | | | | to build ops to use the corresponding free functions in namespace xla:: instead. PiperOrigin-RevId: 202526945
* Add an output context that can be used to specify outputs to capture when ↵Gravatar Priya Gupta2018-06-28
| | | | | | running multiple steps at a time using the `run_steps_on_dataset` API. It allows the user's step function to specify which outputs to emit at what frequency. Currently it only supports capturing output from the last step, but will soon be augmented to support other use cases such as output each N steps. PiperOrigin-RevId: 202520245
* tf.keras sync 2.2.0Gravatar Anjali Sridhar2018-06-28
| | | | PiperOrigin-RevId: 202520102
* Bugfix: In eager mode, bessel_i0 and bessel_i1 raise "TypeError: unhashable ↵Gravatar Brian Patton2018-06-28
| | | | | | | | type: 'list'" out of ops.name_scope because the second arg of name_scope is default_name, not values. Standardizes special_math_ops on the full name_scope(name, default_name, values) constructor. PiperOrigin-RevId: 202519452
* Add code to parse interpolatable error messages.Gravatar James Keeling2018-06-28
| | | | | | The interpolate function currently just reformats tags, without adding any useful information. This change is part of a chain which will add this. PiperOrigin-RevId: 202519204
* Replace Keras clip by value and clip by norm in Keras Optimizers with native ↵Gravatar A. Unique TensorFlower2018-06-28
| | | | | | TF clip_ops, also added user input check for clipnorm and clipvalue >= 0 if set PiperOrigin-RevId: 202516320
* _UnaryOpsComposition kernel: compose multiple shape&type preserving unary ↵Gravatar Eugene Zhulenev2018-06-28
| | | | | | ops at runtime. PiperOrigin-RevId: 202514848
* Include eager/graph mode in cache key so that one type of tensor doesn't spillGravatar Akshay Modi2018-06-28
| | | | | | into the other. PiperOrigin-RevId: 202513508
* Add helper for creating error tagsGravatar James Keeling2018-06-28
| | | | | | | error_format_tag is a helper for building interpolatable strings as part of a project to improve Python error messages in TensorFlow. PiperOrigin-RevId: 202509392
* Avoid overflow in flops calculations in nn_ops.py by forcingGravatar A. Unique TensorFlower2018-06-28
| | | | | | np.prod() to use np.int64 in a few places. PiperOrigin-RevId: 202505308
* [XLA] Change code in TF/XLA bridge that uses XlaBuilder:: methods to build ↵Gravatar Peter Hawkins2018-06-28
| | | | | | ops to use the corresponding free functions in namespace xla:: instead. PiperOrigin-RevId: 202505306
* tf.keras sync 2.2.0Gravatar Anjali Sridhar2018-06-28
| | | | PiperOrigin-RevId: 202505228
* [XLA] VerifyShape() should verify that max_sparse_elements is non-negative.Gravatar Michael Kuperstein2018-06-28
| | | | PiperOrigin-RevId: 202504925
* Add `--input_examples` option to `run` syntax and update `--input_examples` ↵Gravatar A. Unique TensorFlower2018-06-28
| | | | | | and `--input_exprs` headings to match option names. PiperOrigin-RevId: 202504009
* Jump to version 1.9 to sync with TensorFlow versions.Gravatar A. Unique TensorFlower2018-06-28
| | | | | | Also enable cloud tpu profiler to detect the TF version for better version compatibility. PiperOrigin-RevId: 202503162
* make identify_lstm independent of rnn_statesGravatar A. Unique TensorFlower2018-06-28
| | | | PiperOrigin-RevId: 202501055
* More un-fused quantized LSTM support in TFLite interpreterGravatar A. Unique TensorFlower2018-06-28
| | | | PiperOrigin-RevId: 202496488
* Run setUp in the same mode each time.Gravatar Tom Hennigan2018-06-28
| | | | PiperOrigin-RevId: 202489637
* Use a named tag for the nsync version, instead of a git hash.Gravatar A. Unique TensorFlower2018-06-28
| | | | | | Only the name of the version is changing here; the version is unchanged. PiperOrigin-RevId: 202486284
* Support more quantized unfused LSTMs in TFLite interpreter.Gravatar A. Unique TensorFlower2018-06-28
| | | | PiperOrigin-RevId: 202472329
* [tf2xla] Return zero-element tensors as tokens from FusedBatchNormGradGravatar A. Unique TensorFlower2018-06-28
| | | | | | We returned one-element tensors with uninitialized content, which msan didn't like. PiperOrigin-RevId: 202463090
* Improve the performance of ParseShapeStringInternalGravatar A. Unique TensorFlower2018-06-28
| | | | | | | | | The previous implementation recompiled the shape regex at every call what is an expensive opertaion. The new implementation improves the hlo text parsing time for very large models for up to 9x by eliminating this overhead. PiperOrigin-RevId: 202454354
* Fixed ShardingMetadata dump of null sharding from None to {}, to make itGravatar A. Unique TensorFlower2018-06-28
| | | | | | compatible with hlo string syntax. PiperOrigin-RevId: 202445509
* [XLA] Handle domain instructions in dataflow analysis.Gravatar A. Unique TensorFlower2018-06-28
| | | | | | Without domain propagation in dataflow analysis we end up in inconsistent domain instructions with BF16 as output and F32 as input. In case of tuple shapes these are not fixed by bfloat16_normalization, and later on they cause asserts once the domain instructions are removed. PiperOrigin-RevId: 202442786
* Exclude test sources from stream executor builds.Gravatar Gunhan Gulsoy2018-06-27
| | | | PiperOrigin-RevId: 202423156
* Create a constant from the feature column group index.Gravatar A. Unique TensorFlower2018-06-27
| | | | PiperOrigin-RevId: 202419595
* Expose SRUCell via tf.contrib.rnn.Gravatar RJ Ryan2018-06-27
| | | | PiperOrigin-RevId: 202415942
* Update LinearOperator tests to use array_ops.placeholder_with_default.Gravatar A. Unique TensorFlower2018-06-27
| | | | PiperOrigin-RevId: 202412660
* Improve export_tensorflow readability by explicitly using type instead of ↵Gravatar A. Unique TensorFlower2018-06-27
| | | | | | using auto. PiperOrigin-RevId: 202409729
* Add complex64 support to tf.lite runtime.Gravatar RJ Ryan2018-06-27
| | | | PiperOrigin-RevId: 202403235
* [XLA] Add test case for TOKEN constants. Make the test case pass.Gravatar Peter Hawkins2018-06-27
| | | | PiperOrigin-RevId: 202401460
* Minor changes in SpaceToBatchND / BatchToSpaceNDGravatar A. Unique TensorFlower2018-06-27
| | | | PiperOrigin-RevId: 202401380
* Use persistent tensor for Convolution HWCN weightsGravatar Yu-Cheng Ling2018-06-27
| | | | PiperOrigin-RevId: 202400843
* [TF:XLA] Bump open source llvm revision to r335708Gravatar Sanjoy Das2018-06-27
| | | | PiperOrigin-RevId: 202399218
* Ignore stop indices when shrink_axis_mask is set in tf.lite StridedSlice ↵Gravatar RJ Ryan2018-06-27
| | | | | | | | | | implementation. Due to an issue with negative StridedSlice indices in TensorFlow, the end indices can specify degenerate slices when negative indices are used to shrink an axis (e.g. for tf.range(4)[-1], start is -1, end is 0, and stride is 1). This fix works around the issue by ignoring stop indices entirely when an axis is shrinking, since in order to be shrunk the length is by definition 1. Fixes Issue #19260. PiperOrigin-RevId: 202398678
* Fix bug in runtime code for FullyConnected.Gravatar A. Unique TensorFlower2018-06-27
| | | | PiperOrigin-RevId: 202397475
* [TF:XLA] Implement QuantizeAndDequantizeV3.Gravatar Peter Hawkins2018-06-27
| | | | | | | | Change XLA lowering of QuantizeAndDequantizeV2/V3 to match the TF kernel much more closely. The main exception is the min_quantized and max_quantized values are calculated as floats to avoid the need for 64-bit integer math, which is not present on all accelerators. Reformats unary_ops_test.py in passing, but on the whole I don't mind the reformatting. PiperOrigin-RevId: 202395114
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-03 13:21:44 +0000