Commit message (Collapse) | Author | Age | |
---|---|---|---|
* | Internal-only change. | Peter Hawkins | 2016-11-30 |
| | | | | Change: 140667141 | ||
* | Move properties in Layer to read-only; add output_size property. | Eugene Brevdo | 2016-11-30 |
| | | | | Change: 140661470 | ||
* | Bugfix in tf.contrib.distributions base class, which enables more flexible ↵ | Eugene Brevdo | 2016-11-30 |
| | | | | | | subclassing of Distribution. Change: 140659334 | ||
* | Eliminate the tf_kernel_libraries BUILD rule. Its "deps" parameter was used in | A. Unique TensorFlower | 2016-11-30 |
| | | | | | | | | | | an ambiguous way to designate (1) the deps attribute of each generated tf_kernel_library target, and (2) associated tf_kernel_library targets for ops which should be included as dependencies of the omnibus cc_library target (for example, :dilation_ops.) As a result, each individual op library brought in the deps of all the others along with dependencies on unrelated ops. Change: 140657643 | ||
* | Add a SavedModel ModelServer test. | Kiril Gorovoy | 2016-11-30 |
| | | | | | - Update saved_model_half_plus_two to export a prediction default serving signature. Change: 140657101 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-11-30 |
| | | | | Change: 140651442 | ||
* | Add tests for metrics and variables. | A. Unique TensorFlower | 2016-11-30 |
| | | | | Change: 140651312 | ||
* | Cleanup division operators after floor change | Andrew Selle | 2016-11-30 |
| | | | | | | | | | | | - Change the name scope names to be consistent with behavior. - Prepare to depreate tf.div() by putting its functionality in _div_without_cast. - Add note about cwise dense sparse ops and how division should be extended to integers - Add notes in tf.truediv and tf.div to prefer tf.divide - Add division operators to hidden ops Change: 140649903 | ||
* | Support all ops used in inception v3 in GraphTransferer | A. Unique TensorFlower | 2016-11-30 |
| | | | | Change: 140648750 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-11-30 |
| | | | | Change: 140644987 | ||
* | Internal-change | Andrew Harp | 2016-11-30 |
| | | | | Change: 140643948 | ||
* | Correct the names of base-class variables when no default name is given. | Lukasz Kaiser | 2016-11-30 |
| | | | | Change: 140643194 | ||
* | Exclude contrib.layers.stack from documentation as it conflicts with tf.stack. | A. Unique TensorFlower | 2016-11-30 |
| | | | | Change: 140641199 | ||
* | Migrate tf summary ops to use tf.contrib.deprecated endpoints. | Dan Mané | 2016-11-30 |
| | | | | Change: 140639557 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-11-30 |
| | | | | Change: 140639515 | ||
* | Preserves order when calling Session.run on an OrderedDict. | A. Unique TensorFlower | 2016-11-30 |
| | | | | Change: 140635682 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-11-30 |
| | | | | Change: 140634839 | ||
* | Deprecate tf.select since it is getting replaced by tf.where. | A. Unique TensorFlower | 2016-11-30 |
| | | | | Change: 140632089 | ||
* | Clear any existing bookmarks when the selected tensor changes. | Nikhil Thorat | 2016-11-30 |
| | | | | Change: 140623910 | ||
* | Add missing target to LLVM Build file. | Peter Hawkins | 2016-11-30 |
| | | | | Change: 140616374 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-11-30 |
| | | | | Change: 140615810 | ||
* | Add ability to disable zero-debiasing in ExponentialMovingAverage, for the ↵ | A. Unique TensorFlower | 2016-11-30 |
| | | | | | | purpose of backwards compatibility to support old checkpoints. For now, set this default value to avoid debiasing. Change: 140613316 | ||
* | Improve numerical stability of eigen dequantization routines and | A. Unique TensorFlower | 2016-11-30 |
| | | | | | corresponding test cases. Change: 140612620 | ||
* | Have ConvertSignaturesToSignatureDef() handle the case in which there are ↵ | Christopher Olston | 2016-11-30 |
| | | | | | | | both named and default signatures. Also, add some ASSERTs to avoid segfaulting when a test fails. Change: 140612556 | ||
* | 3d label visualizer doesn't have a point accessor anymore, just takes an array | Charles Nicholson | 2016-11-30 |
| | | | | | | of label text. Scatter plot adapter now uses point accessor to build label array and hands it to the 3d label visualizer. Change: 140610524 | ||
* | Update saved-model half plus two with a Predict signature def. | Sukriti Ramesh | 2016-11-30 |
| | | | | Change: 140609389 | ||
* | Add a trace annotator for python files. | Dan Smilkov | 2016-11-30 |
| | | | | Change: 140606458 | ||
* | projectorScatterPlotAdapter populates labels with strings, instead of passing | Charles Nicholson | 2016-11-30 |
| | | | | | | the labelAccessor to the canvas label visualizer. Continue decoupling dataSet accessors from the scatter plot. Remove DataSet from RenderContext! Change: 140604276 | ||
* | Add support for benchmarking with uint8 input types. | Yoni Ben-Meshulam | 2016-11-29 |
| | | | | Change: 140568567 | ||
* | StreamExecutor: Optimize kernel argument packing | Peter Hawkins | 2016-11-29 |
| | | | | | Create a single class to hold all kernel arguments and optimize how they are added into this class. Change: 140556725 | ||
* | Do not calculate train_op in eval mode for _BinaryLogisticHead. | A. Unique TensorFlower | 2016-11-29 |
| | | | | Change: 140555724 | ||
* | Added AVX-512 support to Eigen and TensorFlow. | A. Unique TensorFlower | 2016-11-29 |
| | | | | Change: 140553382 | ||
* | Seal the tf.layers interface. | Patrick Nguyen | 2016-11-29 |
| | | | | Change: 140552119 | ||
* | Ensuring that tuples and dicts are converted to lists correct for use by ↵ | Nathan Silberman | 2016-11-29 |
| | | | | | | tf.with_dependencies. Change: 140550034 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-11-29 |
| | | | | Change: 140548948 | ||
* | Remove net_lib since the tests that were using it are using portpicker library | A. Unique TensorFlower | 2016-11-29 |
| | | | | | instead now. Change: 140548344 | ||
* | Another instance of wrong context. | Yuan Yu | 2016-11-29 |
| | | | | Change: 140548009 | ||
* | Add Dropout layer, its functional interface, and `training` argument support ↵ | A. Unique TensorFlower | 2016-11-29 |
| | | | | | | in base layer `__call__`. Change: 140547974 | ||
* | Replace tf.train.SummaryWriter with tf.summary.FileWriter | Dan Mané | 2016-11-29 |
| | | | | Change: 140545243 | ||
* | Only run the tests for the NCHW layout on CUDA GPUs, since this is the only ↵ | Benoit Steiner | 2016-11-29 |
| | | | | | | type of GPUs for which they're implemented. Change: 140537838 | ||
* | Small fixes to logging and stacktrace platform libraries. | Peter Hawkins | 2016-11-29 |
| | | | | Change: 140533246 | ||
* | Fix C++ shape function for Concat with concat_dim=-1. | A. Unique TensorFlower | 2016-11-29 |
| | | | | | | | Use concat_dim=-1 in crelu. This lets shape inference infer the output shape correctly; until shape inference is driven completely by C++, the value of tf.Rank(...) - 1 is not inferrred as a constant. Change: 140529371 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-11-29 |
| | | | | Change: 140527730 | ||
* | Fixed a typo in the variables class description. | A. Unique TensorFlower | 2016-11-29 |
| | | | | Change: 140525356 | ||
* | Update comment | Wei Ho | 2016-11-29 |
| | | | | Change: 140524503 | ||
* | Dedup multiple control edges from one partition to a single destination in ↵ | Yuan Yu | 2016-11-29 |
| | | | | | | another partition. Change: 140521393 | ||
* | Add optional custom_getter_ param to make_template(). | Skye Wanderman-Milne | 2016-11-29 |
| | | | | Change: 140520520 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-11-29 |
| | | | | Change: 140520388 | ||
* | LinearOperator updates | Ian Langmore | 2016-11-29 |
| | | | | | | | | | | | SquareLinearOperatorDerivedClassTest. New base test class that reduces boilerplate in derived class tests for square operators. _tests_to_skip property. Provides a way to skip tests. add_to_tensor() method added. Some operators, like LinearOperatorDiag, can easily be added to a Tensor. Generic _to_dense() method now implemented in the base class. This only relies on .apply. This is somewhat efficient when .apply() is efficient. This is useful for some cases, such as a Circulant operator where .apply() is fast, but forming the dense matrix would be cumbersome. Change: 140519839 | ||
* | Add experimental option for compilation. | A. Unique TensorFlower | 2016-11-29 |
| | | | | Change: 140517341 |