aboutsummaryrefslogtreecommitdiffhomepage
Commit message (Collapse)AuthorAge
* allows tf.Print to print empty data list and changes a noop test in ↵Gravatar Olivia Nordquist2017-11-10
| | | | | | function_test.py to verify that it doesn't raise a ValueError as an empty list would have previously PiperOrigin-RevId: 175079527
* Move MakeFakeLiteral from client/lib/testing.h to tests/test_utils.h. Also ↵Gravatar Mark Heffernan2017-11-10
| | | | | | | | remove superfluous literal creation methods in that file, and replace them with the existing ones in the Literal class. Also, optionally print layout in Literal::ToString. PiperOrigin-RevId: 175076277
* Add a --all_tensor_names option, which is useful if I only want to know all ↵Gravatar A. Unique TensorFlower2017-11-10
| | | | | | tensor names. It is especially useful in cases whether some of the tensors has huge size. Also update the usage description. PiperOrigin-RevId: 175074541
* Hlo parser: support window and convolution.Gravatar A. Unique TensorFlower2017-11-10
| | | | | | | | Also, to make the text format easier to write and unambiguous: - Print "window={}" around the window attribute; rename the "window" sub attribute to "size"; - Print the dim_lables in logical order, instead of physical order. PiperOrigin-RevId: 175074526
* Improve usability of TPUEstimator.Gravatar Jianwei Xie2017-11-10
| | | | | | | | 1) Log how many batches to enqueue. The old message is very confusing. 2) If input_pipeline has queue runner, generate a logging (legacy mode) or error out (new mode) 3) If input pipeline has summaries, generate a logging (legacy mode) or error out (new mode) PiperOrigin-RevId: 175073856
* Update tf.keras RNNs to the Keras 2.0.9 API. Does not include cuDNN layers.Gravatar Francois Chollet2017-11-10
| | | | | | Additionally, fix a bug with handling of activity_regularizer in tf.layers base Layer (and add test). PiperOrigin-RevId: 175070161
* Fix tensorflow.org rendering of the example code for run_step_fn.Gravatar Igor Saprykin2017-11-10
| | | | | | Python code isn't indented correctly. PiperOrigin-RevId: 175067065
* Run Estimator.export_savedmodel with the user's TFSession config.Gravatar Igor Saprykin2017-11-10
| | | | | | | | | | | | Estimator assumes a particular config_pb2.ConfigProto that configures the underlying session. The config is either the default one or a user-supplied one. The default config has allow_soft_placement=True, the option that allows silent placement of operations on devices with kernels when the requested device doesn't have a kernel for that operation. Estimator's train(), eval() and predict() calls run with the underlying session configured in accordance to the ConfigProto. However, export_savedmodel runs without such a configuration. This appears to be a problem when the ModeKeys.PREDICT graph has an op that was placed on GPU but doesn't have a GPU kernel. The graph works for predict(), but when export_savedmodel() is trying to restore the corresponding variable, the code fails with "no kernel for the op" error. I attempted to show that in a test. To fix this issue, I am passing the ConfigProto to the session inside export_savedmodel. An alternative conservative and ugly fix is to pass a new instance ConfigProto with only allow_soft_placement=Estimator._session_config.allow_soft_placement. Passing the whole ConfigProto feels like the right thing to do. Here's what else is in ConfigProto: https://github.com/tensorflow/tensorflow/blob/r1.4/tensorflow/core/protobuf/config.proto#L280. I verified by running an internal pipeline. Here's allow_soft_placement logic: https://github.com/tensorflow/tensorflow/blob/r1.4/tensorflow/core/common_runtime/placer.cc#L322. PiperOrigin-RevId: 175063803
* More idiomatic tests for defuns using variables.Gravatar Alexandre Passos2017-11-10
| | | | PiperOrigin-RevId: 175063558
* Fix typo in tensorflow/python/client/session_clusterspec_prop_test.pyGravatar Yifei Feng2017-11-10
| | | | PiperOrigin-RevId: 175061854
* [XLA:GPU] Add more logging to convolution autotuning.Gravatar Justin Lebar2017-11-10
| | | | PiperOrigin-RevId: 175057863
* Moves imperative_grad to CGravatar Alexandre Passos2017-11-10
| | | | | | Neutral-to-positive on all benchmarks. Also reduces overhead of should_record. PiperOrigin-RevId: 175057104
* Optimize gradient subgraphs by taking advantage of symbolic shapes wheneverGravatar Benoit Steiner2017-11-10
| | | | | | possible. PiperOrigin-RevId: 175055770
* Give a better error message when placeholders are used with eagerGravatar Igor Ganichev2017-11-10
| | | | PiperOrigin-RevId: 175053592
* Allow Operation._get_attr() to work with all attr types with C API enabledGravatar Skye Wanderman-Milne2017-11-10
| | | | | | | | | | | | | | | | | | This is achieved by accessing the AttrValue directly and using the existing Python code instead of dispatching to the specific C API attr getter for every type. I started going down the dispatch path, but it turns out to be a lot of code (spread across Python, C, and SWIG), and this is likely good enough from a performance standpoint. We can optimize in the future if necessary. In addition, changes the colocation group logic to use _set_attr() and get_attr(), and makes _set_attr() work with the C API disabled. This allows the colocation tests to pass with both the C API enabled and disabled. Without these additional changes, the "_class" attribute would be set on the C NodeDef, and then it would try to retrieve it from the Python NodeDef. PiperOrigin-RevId: 175050473
* Support replacing tpu_config.Gravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 175049981
* Add padded_batch_and_drop_remainder and factor out sharedGravatar A. Unique TensorFlower2017-11-10
| | | | | | filter_irregular_batches. PiperOrigin-RevId: 175045241
* Supports logits as a Tensor in MultiHead.Gravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 175042091
* Having with_gcp_support and windows causes build error.Gravatar Michael Case2017-11-10
| | | | | | | | Multiple statements in a select statement should not be able to be true at the same time (unless one rule is more 'specific' than another). PiperOrigin-RevId: 175040618
* Go: Update generated wrapper functions for TensorFlow ops.Gravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 175037663
* [XLA:CPU] Implement single threaded Matrix-Vector products in LLVM IRGravatar Sanjoy Das2017-11-10
| | | | | | | | | | | | | | | | | | | | | | | Right now we're always doing a 8x8 tiling on the matrix. This can probably be tuned further. There are some other follow-up items that I did not want to put in this already large CL: - Eigen has some smarts to avoid issuing unaligned vector loads and stores which the current CL does not. We need to investigate if being smart about alignment is worth it. - Prevent LLVM from vectorizing the epilogue. In fact we should disable loop vectorization for all the loops we've explicitly vectorized. - Cache the kernels by their shape to reduce code size impact. - Add aliasing information to the loads and stores emitted by the PacketSupportLibrary. This is probably not super critical since we've already vectorized the code, but we should do this for completeness. PiperOrigin-RevId: 175036991
* Minor docstring fixesGravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 175036743
* [XLA] More diagnostic information in the reshape shape inference error.Gravatar Chris Leary2017-11-10
| | | | PiperOrigin-RevId: 175036413
* Add comment describing how to get optimized builds in Dockerfile.Gravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 175036186
* Add database writer ops to contrib/summaryGravatar Justine Tunney2017-11-10
| | | | PiperOrigin-RevId: 175030602
* [XLA] Print constant literals of size <= 8 elements.Gravatar Justin Lebar2017-11-10
| | | | | | | Previously we'd only print scalars. But if you have a constant with just a few values, what the heck, show the whole thing. PiperOrigin-RevId: 175030210
* HParams.set_hparam doesn't fully check types, contrary to its docstring.Gravatar Thomas Schumm2017-11-10
| | | | PiperOrigin-RevId: 175028981
* Do not return a mutable HloComputation* from a entry_computation() on aGravatar A. Unique TensorFlower2017-11-10
| | | | | | const HloModule*. PiperOrigin-RevId: 175024608
* Fix incomplete spec of EagerTensor.numpy()Gravatar Igor Ganichev2017-11-10
| | | | PiperOrigin-RevId: 175023039
* tfdbg: Fix a test bug hidden in a child threadGravatar Shanqing Cai2017-11-10
| | | | PiperOrigin-RevId: 175004323
* tfdbg: Add test for loading DebugDumpDir with a relative pathGravatar Shanqing Cai2017-11-10
| | | | PiperOrigin-RevId: 174999937
* Check GPU availability after creating test session.Gravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 174983466
* Also register string types if __ANDROID_TYPES_FULL__ is definedGravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 174979678
* Update ops-related pbtxt files.Gravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 174964560
* Go: Update generated wrapper functions for TensorFlow ops.Gravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 174962378
* Automated g4 rollback of changelist 174912490Gravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 174961746
* Update nsync version---allow compilation with bazel on x86_32Gravatar A. Unique TensorFlower2017-11-10
| | | | | | | The new version of nsync has a BUILD file that detects x86_32 (which bazel currently calls piii). PiperOrigin-RevId: 174959924
* Fix tf.contrib.seq2seq._monotonic_probability_fn to use a hard sigmoid when ↵Gravatar Colin Raffel2017-11-10
| | | | | | | | mode='hard'. Also adds tests to make sure the attention probabilities are 0 or 1 when mode='hard'. PiperOrigin-RevId: 174956465
* Fix link (the link tool expects these to be on one line)Gravatar Mark Daoust2017-11-10
| | | | PiperOrigin-RevId: 174948909
* Silenced an unnecessary warningGravatar Benoit Steiner2017-11-10
| | | | PiperOrigin-RevId: 174947453
* MovingAvgQuantize and LastValueQuantize should use the updated value from ↵Gravatar Suharsh Sivakumar2017-11-10
| | | | | | the Assign op, otherwise min max variables never get updated. PiperOrigin-RevId: 174947421
* Fix Bazel builds for the TF Lite demo appGravatar Austin Anderson2017-11-10
| | | | | | | Adds a new remote repository for the mobilenet tflite models necessary for running the TF Lite demo app. PiperOrigin-RevId: 174946867
* Add functionality to perform training of additional fixed point layer on top ↵Gravatar Suharsh Sivakumar2017-11-10
| | | | | | | | of quantized base model. Also modify retrain_test to test creation of model info for fixed point mobilenet. PiperOrigin-RevId: 174946745
* Register int64 for GPU StridedSlice kernelGravatar Shanqing Cai2017-11-10
| | | | PiperOrigin-RevId: 174944857
* [tf.data] Saveable iterator for dataset.zip(..).Gravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 174941651
* [XLA] Fix comments for arg_literals parameter in HloEvaluator::Evaluate.Gravatar Kay Zhu2017-11-10
| | | | PiperOrigin-RevId: 174939009
* Removed an unused temporary variable from DeviceNameUtils::ParseFullName.Gravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 174938299
* Used tf.where to simplify conditional expression in div-sharding.Gravatar A. Unique TensorFlower2017-11-10
| | | | PiperOrigin-RevId: 174937860
* Make quant_delay work even if user didn't create global step.Gravatar Suharsh Sivakumar2017-11-10
| | | | PiperOrigin-RevId: 174937793
* [XLA] Add binary operation name to shape inference error message.Gravatar Chris Leary2017-11-10
| | | | PiperOrigin-RevId: 174937290
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-06 11:04:24 +0000