Commit message (Collapse) | Author | Age | |
---|---|---|---|
* | [XLA] Use `ShapeUtil::PermuteDimensions` in ↵ | 2017-01-11 | |
| | | | | | | `ShapeInference::InferTransposeShape` Change: 144282152 | ||
* | Delete the FunctionDef.Node code paths, now that we have switched | 2017-01-11 | |
| | | | | | to the NodeDef representation. Change: 144281952 | ||
* | Python3 Compatibility for tf_upgrade_test.py | 2017-01-11 | |
| | | | | Change: 144279869 | ||
* | Convert more flags use to argparse in dist_test | 2017-01-11 | |
| | | | | Change: 144278086 | ||
* | Internal only change | 2017-01-11 | |
| | | | | Change: 144275556 | ||
* | Autogenerated Change: Change TensorBoard TAG to 40 | 2017-01-11 | |
| | | | | Change: 144273322 | ||
* | Fully deprecate mutability of RunConfig. | 2017-01-11 | |
| | | | | Change: 144273220 | ||
* | Fix documentation of rows vs columns of confusion matrix. | 2017-01-11 | |
| | | | | Change: 144271336 | ||
* | Added a slider beneath image summaries within Tensorboard's images panel. ↵ | 2017-01-11 | |
| | | | | | | The slider lets the user chose which step to view. Change: 144270020 | ||
* | Estimator fit method returns self even if max_steps has been reached. | 2017-01-11 | |
| | | | | Change: 144269545 | ||
* | Speed up tensorflow/contrib/linalg:linear_operator_composition_test. | 2017-01-11 | |
| | | | | | | | The test was appending to the existing Graph in a loop, which lead to an apparent quadratic behavior. Instead, create a new Graph for each test instance. Test improved from 553s to 28s wall time on my desktop machine. Change: 144269242 | ||
* | Add a placeholder for hexagon_controller | 2017-01-11 | |
| | | | | Change: 144268749 | ||
* | Switch open-source to use jemalloc for CPU Tensor memory allocation, gRPC, ↵ | 2017-01-11 | |
| | | | | | | | | and other places where we call malloc/free. - Only enabled on Linux for now. - Added as a ./configure option defaulting to enabled. Change: 144266237 | ||
* | Clarify ResetDevice operates on all devices associated with backend. | 2017-01-11 | |
| | | | | Change: 144258290 | ||
* | Update generated Python Op docs. | 2017-01-11 | |
| | | | | Change: 144255262 | ||
* | More conversions of flags library to argparse. | 2017-01-11 | |
| | | | | | | Add argv to benchmark/main function so they can handle passing command line arguments. Change: 144254260 | ||
* | Now that moments are numerically stable by default, not need to use moving_mean. | 2017-01-11 | |
| | | | | Change: 144251474 | ||
* | Android: add image stylization example demo based on "A Learned ↵ | 2017-01-11 | |
| | | | | | | Representation For Artistic Style" Change: 144247143 | ||
* | Add an option to run inference on hvx into build script | 2017-01-11 | |
| | | | | Change: 144246159 | ||
* | Internal-only change. | 2017-01-11 | |
| | | | | Change: 144239231 | ||
* | Update generated Python Op docs. | 2017-01-11 | |
| | | | | Change: 144238818 | ||
* | Removing TensorArray.pack and TensorArray.unpack | 2017-01-11 | |
| | | | | Change: 144235900 | ||
* | Fix to correctly identify dtype and shape of inputs in StreamingDataFeeder. | 2017-01-11 | |
| | | | | Change: 144230407 | ||
* | Add optimized reverse kernel for reversing rows of a 3D tensor. This is used in | 2017-01-11 | |
| | | | | | | | image preprocessing pipelines. Add benchmark for reverse. Change: 144224184 | ||
* | Update generated Python Op docs. | 2017-01-11 | |
| | | | | Change: 144217413 | ||
* | Add link back to contrib/android/README.md from contrib/android/cmake/README.md | 2017-01-11 | |
| | | | | Change: 144216903 | ||
* | Deprecate tf.neg, tf.mul, tf.sub (and remove math_ops.{neg,mul,sub} usages | 2017-01-11 | |
| | | | | | | tf.negative, tf.multiply, tf.subtract are the new names - Also enabled deprecation warning (to be completely removed by friday) Change: 144215355 | ||
* | Add support for list(func) AttrValues. | 2017-01-11 | |
| | | | | Change: 144211572 | ||
* | Build and expose benchmark_model binary as part of Android nightly build, ↵ | 2017-01-10 | |
| | | | | | | and also run build_all_android.sh for better build coverage. Change: 144158273 | ||
* | Remove --icf=all flag from benchmark_model to fix build for Android arm64-v8 | 2017-01-10 | |
| | | | | Change: 144158100 | ||
* | Update generated Python Op docs. | 2017-01-10 | |
| | | | | Change: 144155792 | ||
* | Android demo: Add support for YOLO-based object detection to object tracking ↵ | 2017-01-10 | |
| | | | | | | demo. Set USE_YOLO boolean flag in DetectorActivity.java to true and add model according to directions there to enable. Change: 144155665 | ||
* | Switch tf.concat_v2 references in third_party/tensorflow to tf.concat. | 2017-01-10 | |
| | | | | Change: 144153795 | ||
* | Update generated Python Op docs. | 2017-01-10 | |
| | | | | Change: 144152960 | ||
* | Change CheckpointSaverHook to only call SummaryWriterCache if the hook | 2017-01-10 | |
| | | | | | | | | is active. Before the tfevents file was created even if the hook is never installed, which confuses Tensorboard and can cause it to watch the wrong file and not display updates. Change: 144152846 | ||
* | Test that the dense layer respects the initializer from variable_scope above. | 2017-01-10 | |
| | | | | Change: 144152550 | ||
* | Refactor (DNN|Linear|DNNLinearCombined)(Classifier|Regressor) to inherit ↵ | 2017-01-10 | |
| | | | | | | from Estimator. Change: 144152544 | ||
* | Conditionally enable or disable XLA in TF tests. | 2017-01-10 | |
| | | | | Change: 144150966 | ||
* | Automated rollback of change 141622306 | 2017-01-10 | |
| | | | | Change: 144145884 | ||
* | Optimize eight-bit graphs by removing RequantizationRanges | 2017-01-10 | |
| | | | | Change: 144145086 | ||
* | Update generated Python Op docs. | 2017-01-10 | |
| | | | | Change: 144144091 | ||
* | Implement KL-divergence between two Gamma distributions. | 2017-01-10 | |
| | | | | Change: 144144041 | ||
* | Deprecate graph_actions.py | 2017-01-10 | |
| | | | | Change: 144142420 | ||
* | Allow tf.divide's name argument to be honored. | 2017-01-10 | |
| | | | | | | | | | - Prevent tf.divide from using operator overloads which cannot pass names - Use dummy python class to properly delegate to the right division mode (i.e. python2 vs python3 vs from __future__ import divide) Fixes #6741 Change: 144140254 | ||
* | Add tf.concat with the new argument order: tf.concat(values, axis, name). Also, | 2017-01-10 | |
| | | | | | set tf.concat_v2 to be an alias to tf.concat and remove concat_v2 doc. Change: 144139939 | ||
* | Remove std:: from std::round() calls in benchmark_model.cc to avoid issue in ↵ | 2017-01-10 | |
| | | | | | | Android ndk where std::round is not found: https://code.google.com/p/android/issues/detail?id=54418 Change: 144137666 | ||
* | Update generated Python Op docs. | 2017-01-10 | |
| | | | | Change: 144133445 | ||
* | Remove all remaining tf.pack,tf.unpack references and remove tf.pack/tf.unpack | 2017-01-10 | |
| | | | | | op. Change: 144130931 | ||
* | Removed extra / in text | 2017-01-10 | |
| | | | | Change: 144130687 | ||
* | Make models.BUILD filegroup include everything but metadata files and archives. | 2017-01-10 | |
| | | | | Change: 144127608 |