Commit message (Collapse) | Author | Age | |
---|---|---|---|
* | Remove unnecessary zero check from resize_bilinear_op.cc. | Brennan Saeta | 2016-12-15 |
| | | | | | Additionally, update the resize_bilinear_op tests to ensure they fail with a given error. Change: 142204006 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142199921 | ||
* | Update ops-related pbtxt files. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142198903 | ||
* | Add Expm1 Op. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142198004 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142194349 | ||
* | Test for CPU features on startup, to replace SIGILLs with useful errors and ↵ | Pete Warden | 2016-12-15 |
| | | | | | | suggest optimizations Change: 142193596 | ||
* | Remove obsolete comment in tf.abs docstring. tf.abs now delegates to | A. Unique TensorFlower | 2016-12-15 |
| | | | | | | | | | complex_abs for complex tensors. Hide generated complex_abs kernel in preparation for deprecation. Fix a few unrelated Python linter errors that have been getting on my nerves lately. Change: 142188440 | ||
* | Replace _BaseEstimatorForTest with a model_fn. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142187067 | ||
* | Upgraded to the latest version of Eigen | Benoit Steiner | 2016-12-15 |
| | | | | | | | Made it possible to upgrade Eigen without having to change the version in multiple places Made the cmake file check the signature of the package again Change: 142186076 | ||
* | Fix use-after-free bug when closing a MasterSession. | Derek Murray | 2016-12-15 |
| | | | | | | | | | | | | | | | | | | | | A concurrent run call could continue to use a MasterSession that had deleted itself (in `MasterSession::Close()`). Change `MasterSession` to be an explicitly shared, refcounted object; and modify the `Master` so that each call that uses a session will acquire a reference to that session, then release it before calling the callback. Add a `MasterSession::closed_` bit to ensure that no call uses a session after `MasterSession::Close()` has succeeded. Also fix a null-pointer access that could occur if a ReffedClientGraph was partially constructed. If `ReffedClientGraph::RegisterPartitions()` failed, some of the `Part::worker` fields would be uninitialized, and `ReffedClientGraph::DeregisterPartitions()` (called by the destructor) would attempt to invoke a method on that object. Fix this by only attempting to deregister partitions that have been successfully registered. Change: 142185174 | ||
* | Update SVM classifier to be consitent with other classifiers with respect to ↵ | A. Unique TensorFlower | 2016-12-15 |
| | | | | | | model_dir handling. Change: 142184770 | ||
* | Avoid bazel-mirror URLs for make and cmake | Justine Tunney | 2016-12-15 |
| | | | | | | People might make commits updating the version without actually mirroring the URL. Change: 142177676 | ||
* | Adding Scaffold to the tf.learn.ModelFn, so that users can play with ↵ | Mustafa Ispir | 2016-12-15 |
| | | | | | | initialization, saving, and so on. Change: 142175064 | ||
* | Add a debug -mavx option for Cloud ML builds. | Jonathan Hseu | 2016-12-15 |
| | | | | Change: 142174591 | ||
* | Sync the github and local versions of the ThreadPool header | Benoit Steiner | 2016-12-15 |
| | | | | Change: 142169284 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142166062 | ||
* | Make the context correct for the gradient of mixed nesting of conds and loops. | Yuan Yu | 2016-12-15 |
| | | | | Change: 142162324 | ||
* | Mutes warnings for some deprecated arguments with default values. Fixes a bug. | A. Unique TensorFlower | 2016-12-15 |
| | | | | | | | | | | Users were getting log warnings even when using the API correctly, because of argument passthrough within tf.learn. This change mutes warnings for calls to fit, estimate, and predict if the deprecated arguments receive default values. Also fixes a bug in the deprecation tool where "==" was used instead of "is" to compare objects, which was at times yeilding a non-boolean. Change: 142162193 | ||
* | Add tensor contraction op 'tensordot' to opensource TensorFlow. Original ↵ | A. Unique TensorFlower | 2016-12-15 |
| | | | | | | implementation by Moritz Hardt (mrtz@google.com). Fixes Github issue #5231. Change: 142161582 | ||
* | Automated rollback of change 141675118 | Yuan Yu | 2016-12-15 |
| | | | | Change: 142160746 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142159965 | ||
* | Add an op for matrix norm, implementing a subset of numpy.linalg.norm. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142156560 | ||
* | Switch shape inference handles resource variable shapes and dtypes. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142152862 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142149281 | ||
* | Fix linter errors left over from PRs. | Martin Wicke | 2016-12-15 |
| | | | | Change: 142149043 | ||
* | Update ops-related pbtxt files. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142148654 | ||
* | Move the implementation code of static_rnn, static_bidirectional_rnn and ↵ | Jianwei Xie | 2016-12-15 |
| | | | | | | static_state_saving_rnn from core to contrib. Change: 142148394 | ||
* | Add BroadcastShape op. | A. Unique TensorFlower | 2016-12-15 |
| | | | | | Specifically create a broadcast_static_shape and broadcast_dynamic_shape python wrappers. broadcast_static_shape returns an inferred TensorShape from the input TensorShapes, while broadcast_dynamic_shape returns an integer Tensor representing the broadcasted shape. Change: 142147719 | ||
* | Bump eigen version to get implementation of scalar_expm1_op. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142147027 | ||
* | Fix the typo in dnn_benchmark_test. | Jianwei Xie | 2016-12-15 |
| | | | | Change: 142140440 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-12-15 |
| | | | | Change: 142110419 | ||
* | Switch array_ops.pack/unpack to array_ops.stack/unstack. Also switch a few ↵ | A. Unique TensorFlower | 2016-12-15 |
| | | | | | | remaining references to tf.pack/unpack to tf.stack/unstack. Change: 142108785 | ||
* | Removed TensorFlow Mechanics 101 from how_to left navigation | A. Unique TensorFlower | 2016-12-14 |
| | | | | Change: 142087617 | ||
* | Automated rollback of change 142080248 | Mustafa Ispir | 2016-12-14 |
| | | | | Change: 142084980 | ||
* | Add tf-benchmark test for DNNClassifier. | A. Unique TensorFlower | 2016-12-14 |
| | | | | | | Add some more metric assertions to dnn_test. Make feature columns tuples instead of lists. Change: 142084798 | ||
* | Fix failures because of SparseTensor.__init__ signature change. | A. Unique TensorFlower | 2016-12-14 |
| | | | | Change: 142084456 | ||
* | Disable flaky test case in grpc_session_test. | Derek Murray | 2016-12-14 |
| | | | | Change: 142084021 | ||
* | Use log warnings instead of errors for up-converting generic signatures. | Sukriti Ramesh | 2016-12-14 |
| | | | | Change: 142082843 | ||
* | Added tf.train.Scaffold support for tf.learn.Estimator. This lets users to ↵ | Mustafa Ispir | 2016-12-14 |
| | | | | | | play with initialization, saver, and so on. Change: 142080248 | ||
* | Remove hourglass imports from kernel_tests | Justine Tunney | 2016-12-14 |
| | | | | Change: 142080137 | ||
* | Update generated Python Op docs. | A. Unique TensorFlower | 2016-12-14 |
| | | | | Change: 142075580 | ||
* | Implement sampled_scattered_embedding_lookup_sparse. This op is used in ↵ | A. Unique TensorFlower | 2016-12-14 |
| | | | | | | variable pool. Change: 142075501 | ||
* | Update ops-related pbtxt files. | A. Unique TensorFlower | 2016-12-14 |
| | | | | Change: 142074727 | ||
* | Merge changes from github. | Martin Wicke | 2016-12-14 |
| | | | | Change: 142074581 | ||
* | Add Tactivation attribute for QuantizedMatmul and QuantizedMatmulMinMax to ↵ | A. Unique TensorFlower | 2016-12-14 |
| | | | | | | explicitly notify which quantizedtype output would be produced by following activation function. Change: 142073731 | ||
* | Final breaking change, removing .shape from SparseTensor. | Martin Wicke | 2016-12-14 |
| | | | | Change: 142071642 | ||
* | Make comment more concise in C++ API. | A. Unique TensorFlower | 2016-12-14 |
| | | | | Change: 142065959 | ||
* | Fix a typo in the docstring of init_from_checkpoint. | A. Unique TensorFlower | 2016-12-14 |
| | | | | Change: 142065387 | ||
* | Added a microbenchmark for `tf.Session.run()` calls with different targets. | Derek Murray | 2016-12-14 |
| | | | | | We will use this benchmark to track the work towards addressing issue #6256. Change: 142065136 | ||
* | Fix erroneous references to tf.summary.histogram in deprecation string for ↵ | A. Unique TensorFlower | 2016-12-14 |
| | | | | | | non-histogram summary ops. Change: 142063275 |