Commit message (Collapse) | Author | Age | |
---|---|---|---|
* | Add fused_conv2d_bias_activation operator for the forward phase. | Yangzihao Wang | 2017-07-20 |
| | | | | PiperOrigin-RevId: 162624917 | ||
* | Adds RELU6 to the list of cheap operations for potential recompute in ↵ | A. Unique TensorFlower | 2017-07-20 |
| | | | | | | memory_optimizer. PiperOrigin-RevId: 162624766 | ||
* | Modify batch_shape and event_shape for transformed distributions to handle ↵ | A. Unique TensorFlower | 2017-07-20 |
| | | | | | | dynamic shapes. PiperOrigin-RevId: 162619535 | ||
* | Generalize categorical CDF broadcasting over argument. | A. Unique TensorFlower | 2017-07-20 |
| | | | | PiperOrigin-RevId: 162601793 | ||
* | Use an ErrorSpec in R4ReduceWindowTest::DoIt and re-enable a disabled test. | A. Unique TensorFlower | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162570059 | ||
* | Fix missing spaces for several errors. | A. Unique TensorFlower | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162567926 | ||
* | Remove invalid usage of DISABLED_ON_(C|G)PU -- these don't work on ↵ | A. Unique TensorFlower | 2017-07-19 |
| | | | | | | reduce_window_test.cc PiperOrigin-RevId: 162566699 | ||
* | Add support for display_name and summary_description to the ↵ | Dandelion Man? | 2017-07-19 |
| | | | | | | | | | | | tf.summary.tensor_summary op. The display_name will be used to display the series in TensorBoard, in lieu of the tag, assuming that it is specified. (When it is not specified, behavior will stay the same.) The summary_description will allow the user to write a longform readable description of the summary series for display in TensorBoard. Markdown will be supported. This will make it possible for TensorBoard-2 summary ops to give the user direct control over the display name, solving https://github.com/tensorflow/tensorboard/issues/59. PiperOrigin-RevId: 162566261 | ||
* | Internal change. | Anna R | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162562631 | ||
* | Further BUILD cleanup in tensorflow/contrib/... | A. Unique TensorFlower | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162561536 | ||
* | Minor cleanup | A. Unique TensorFlower | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162561496 | ||
* | Disable a test temporarily while we fix it to account for fastmath-caused ↵ | A. Unique TensorFlower | 2017-07-19 |
| | | | | | | floating point imprecision. PiperOrigin-RevId: 162560345 | ||
* | Fully qualify namespaces in XLA macros. | Justin Lebar | 2017-07-19 |
| | | | | | | No functional change. PiperOrigin-RevId: 162558977 | ||
* | Add multi-head attention capabilities to AttentionWrapper via the ↵ | Adam Roberts | 2017-07-19 |
| | | | | | | specification of multiple AttentionMechanisms. PiperOrigin-RevId: 162557562 | ||
* | Allow kernel op tests to add resource inputs | A. Unique TensorFlower | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162556398 | ||
* | Fix comment: OperatorPDCholesky -> LinearOperatorTriL | A. Unique TensorFlower | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162544918 | ||
* | Split HLO profile display logic out of hlo_execution_profile.cc, moving | Justin Lebar | 2017-07-19 |
| | | | | | | it into execution_profile_builder.cc. PiperOrigin-RevId: 162541782 | ||
* | Merge changes from github. | Jonathan Hseu | 2017-07-19 |
| | | | | | | | | | | | | | | | | | END_PUBLIC --- Commit daa67ad17 authored by Jonathan Hseu<vomjom@vomjom.net> Committed by Frank Chen<frankchn@gmail.com>: Remove unittest import (#11596) --- Commit 491beb74c authored by A. Unique TensorFlower<gardener@tensorflow.org> Committed by TensorFlower Gardener<gardener@tensorflow.org>: BEGIN_PUBLIC Automated g4 rollback of changelist 162423171 PiperOrigin-RevId: 162541442 | ||
* | PiperOrigin-RevId: 162539241 | A. Unique TensorFlower | 2017-07-19 |
| | |||
* | Add functions for specifying a custom gradient. | Joshua V. Dillon | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162527721 | ||
* | Automated g4 rollback of changelist 160314706 | Yangzihao Wang | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162525519 | ||
* | Disable padding by default in tf.contrib.signal.stft to match ↵ | RJ Ryan | 2017-07-19 |
| | | | | | | tf.contrib.signal.frame. PiperOrigin-RevId: 162523648 | ||
* | Internal change. | A. Unique TensorFlower | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162522519 | ||
* | Use math_ops.{sinh,asin,cosh} instead of trig.{sinh,arc_sinh) in Sinh bijector. | A. Unique TensorFlower | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162519993 | ||
* | Add Short-time Fourier Transform (STFT) and inverse STFT support to ↵ | RJ Ryan | 2017-07-19 |
| | | | | | | | | tf.contrib.signal. These are implemented using GPU-capable ops and have gradient support. PiperOrigin-RevId: 162511440 | ||
* | Remove class BoolVector. Use uint8 in its place. | A. Unique TensorFlower | 2017-07-19 |
| | | | | | | | | Both preds() and uint8() use the same storage. In a subsequent CL, I will unify all the different storage types and make Literal a template class. It's now replicating storage that is never used. PiperOrigin-RevId: 162509649 | ||
* | [XLA] Move the flag from stream_assignment_flags into DebugOptions | Eli Bendersky | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162506964 | ||
* | [TF:XLA] Use a separate function library for functions generated by control ↵ | Peter Hawkins | 2017-07-19 |
| | | | | | | flow functionalization pass. Avoids copying the main function library for each compilation. PiperOrigin-RevId: 162506388 | ||
* | Minor improvements in TensorForestEstimator: reduce default stopping loss ↵ | A. Unique TensorFlower | 2017-07-19 |
| | | | | | | threshold, which makes it train a little longer and results in better models. Also add average tree size to summary for handy viewing in TensorBoard. PiperOrigin-RevId: 162505559 | ||
* | [TF:XLA] Relax numeric tolerances of some tests. | Peter Hawkins | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162505321 | ||
* | Internal change. | A. Unique TensorFlower | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162502710 | ||
* | Seals tf.contrib.factorization | A. Unique TensorFlower | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162494862 | ||
* | Adds a warning if the evaluation steps are 0 | Jianwei Xie | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162491844 | ||
* | tfdbg: open-source C++ and Python libraries of gRPC debugger mode | Shanqing Cai | 2017-07-19 |
| | | | | | | | | with the exception of Windows the session_debug_grpc_test is temporarily disabled on mac pending pip install of futures and grpcio on all slaves. PiperOrigin-RevId: 162482541 | ||
* | Further BUILD cleanup in tensorflow/contrib/... | A. Unique TensorFlower | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162466482 | ||
* | Add output_partitions support in distributed runtime. | Suharsh Sivakumar | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162456565 | ||
* | Added support for testing convolution of 3D arrays (i.e. 1d-convolution). | A. Unique TensorFlower | 2017-07-19 |
| | | | | PiperOrigin-RevId: 162453537 | ||
* | Fix is_floating for tf.qint*, tf.quint*. | A. Unique TensorFlower | 2017-07-18 |
| | | | | | | is_floating fails for tf.qint* because issubclass is called on tf.qint*, which is not a class. Use numpy.issubdtype to perform the check instead. PiperOrigin-RevId: 162441672 | ||
* | No public change | A. Unique TensorFlower | 2017-07-18 |
| | | | | PiperOrigin-RevId: 162441153 | ||
* | Automated g4 rollback of changelist 162423171 | A. Unique TensorFlower | 2017-07-18 |
| | | | | PiperOrigin-RevId: 162437318 | ||
* | [XLA] Get rid of ServiceFlags by absorbing it into DebugOptions. | A. Unique TensorFlower | 2017-07-18 |
| | | | | | | | After this change HloModuleConfig::hlo_profiling_enabled_ is redundant. I'll remove it in a future change. PiperOrigin-RevId: 162436163 | ||
* | adds missing "pass" to empty method in tf.contrib.learn.DNNClassifier doc | A. Unique TensorFlower | 2017-07-18 |
| | | | | PiperOrigin-RevId: 162431577 | ||
* | Introduces fuzzers for some complex operations and kernels in TensorFlow in ↵ | Frank Chen | 2017-07-18 |
| | | | | | | preparation for eventaul fuzzing using Google OSS-Fuzz. PiperOrigin-RevId: 162425120 | ||
* | Add autotuning code for matmul operator. | Yangzihao Wang | 2017-07-18 |
| | | | | | | Currently it is turned off by default. PiperOrigin-RevId: 162423171 | ||
* | Remove all references to OperatorPD and replace with LinearOperator. | A. Unique TensorFlower | 2017-07-18 |
| | | | | | | - Only dependent is Wishart, which is updated to use LinearOperator. PiperOrigin-RevId: 162421033 | ||
* | Add a utility function to build and write a SavedModel, setting reasonable | A. Unique TensorFlower | 2017-07-18 |
| | | | | | | defaults for the required arguments. PiperOrigin-RevId: 162420810 | ||
* | fix leftnav files for programmer's guide | Mark Daoust | 2017-07-18 |
| | | | | PiperOrigin-RevId: 162418535 | ||
* | FileBlockCache will now run a thread to prune files with expired blocks, if ↵ | A. Unique TensorFlower | 2017-07-18 |
| | | | | | | configured with a nonzero max staleness. PiperOrigin-RevId: 162416777 | ||
* | Fix crash when OOM error occurs with verbose logging enabled | A. Unique TensorFlower | 2017-07-18 |
| | | | | PiperOrigin-RevId: 162414962 | ||
* | Go: Update generated wrapper functions for TensorFlow ops. | A. Unique TensorFlower | 2017-07-18 |
| | | | | PiperOrigin-RevId: 162413543 |