Commit message (Collapse) | Author | Age | |
---|---|---|---|
* | Parse tf.Example using arena allocation. Dense string benchmark shows 10-20% ↵ | 2016-07-06 | |
| | | | | | | speedup. Change: 126726018 | ||
* | Add output_prefix parameter benchmark_model so that output directory can be ↵ | 2016-07-06 | |
| | | | | | | explicitly specified on the command-line. Change: 126725665 | ||
* | Split the demo datasets JSON file for the tf graph demo out into an external ↵ | 2016-07-06 | |
| | | | | | | file. Change: 126722080 | ||
* | Change the error level of Cuda host allocator memory allocation memory to | 2016-07-06 | |
| | | | | | | ERROR. This allows BFC allocator to backtrack when it fails to allocate a bigger chunk of memory. Change: 126718922 | ||
* | Disable flaky contrib/learn:export_test with the manual tag | 2016-07-06 | |
| | | | | Change: 126713223 | ||
* | Enhance test_installation.sh bazel tag filter | 2016-07-06 | |
| | | | | | | | | | | | | | With these changes, the environment-variable switch "TF_BUILD_FILTER_INSTALL_TESTS_BY_TAG" of test_installation.sh is no longer limited to a single tag or limited to only the inclusive filtering mode. Multiple tags can be used, separated by commas. The minus sign ("-") can be used to mark the exclusive filtering mode. The inclusive filtering mode is the default. For example: TF_BUILD_FILTER_INSTALL_TESTS_BY_TAG="local,-manual" will let the script run the Python unit tests that have the tag "local" and do not have the tag "manual". Change: 126707186 | ||
* | Fix sampling_ops test timeout by removing unneeded network inferences in ↵ | 2016-07-06 | |
| | | | | | | unit test. Change: 126691628 | ||
* | Update generated Python Op docs. | 2016-07-03 | |
| | | | | Change: 126531150 | ||
* | Add streaming_mean_iou to metric_ops. | 2016-07-03 | |
| | | | | | | | | | | streaming_mean_iou computes the mean intersection-over-union, a typical evaluation metric in the task of semantic image segmentation. Factor out the utility function, remove_squeezable_dimensions, to metric_ops_util.py, move its test and make changes for the other dependencies as well. Change: 126530424 | ||
* | Implement C++ shape inference functions for Select, Range, and LinSpace. | 2016-07-02 | |
| | | | | Change: 126488375 | ||
* | Make the seq2seq module work well with RNNCells with tuple-like states. (#3056) | 2016-07-02 | |
| | | | | Change: 126478627 | ||
* | Set a max of 0.2 for the learning rate of the Wide model. | 2016-07-01 | |
| | | | | Change: 126470847 | ||
* | Add K-Means clustering and WALS matrix factorization to tensorflow. | 2016-07-01 | |
| | | | | Change: 126465430 | ||
* | Unit test for iterators passed to .fit() | 2016-07-01 | |
| | | | | Change: 126464223 | ||
* | Update generated Python Op docs. | 2016-07-01 | |
| | | | | Change: 126463785 | ||
* | Update ops-related pbtxt files. | 2016-07-01 | |
| | | | | Change: 126463584 | ||
* | Support Sparse-Sparse cwise ops; use for tf.sparse_{minimum,maximum}(). | 2016-07-01 | |
| | | | | | | This change adds the CPU kernel and Python ifaces. For now, assumes both operands have the same shapes. Change: 126460669 | ||
* | Fix export utilities of Estimators and ExportMonitor. | 2016-07-01 | |
| | | | | | init_op should be properly specified when exporting the model. Change: 126458099 | ||
* | Remove dead code from Supervisor and SessionManager. | 2016-07-01 | |
| | | | | Change: 126456805 | ||
* | First step for adding the new TensorSummary op: Add fields to summary.proto | 2016-07-01 | |
| | | | | Change: 126453617 | ||
* | Update generated Python Op docs. | 2016-07-01 | |
| | | | | Change: 126452503 | ||
* | Handle input iterator exhaustion. | 2016-07-01 | |
| | | | | Change: 126450913 | ||
* | Add retry logic to all GCS operations. | 2016-07-01 | |
| | | | | | | | | Also adds a few minor fixes for Google auth: - support for the no-auth case (public objects will still work) - less tolerance for almost expired tokens - better error logging for failed token retrievals Change: 126449395 | ||
* | Clipping the negative padding for planes in conv3d. | 2016-07-01 | |
| | | | | Change: 126449225 | ||
* | Validate the argument to `tf.name_scope()` and fix bug in op name validation. | 2016-07-01 | |
| | | | | Change: 126447618 | ||
* | Add conv2d to testlib. | 2016-07-01 | |
| | | | | Change: 126441771 | ||
* | Update generated Python Op docs. | 2016-07-01 | |
| | | | | Change: 126439312 | ||
* | Make tensorflow_serving/session_bundle compatible with internal build rules. | 2016-07-01 | |
| | | | | Change: 126438387 | ||
* | Handle cudaMemcpyAsync_v3020 API callback in GpuTracer. | 2016-07-01 | |
| | | | | | Reduce verbosity of logging. Change: 126437651 | ||
* | Make SampleAndReshapeValue the default value type for StochasticTensor | 2016-07-01 | |
| | | | | Change: 126436874 | ||
* | Doc-only change clarifying the behavior of default Tensor constructors. | 2016-07-01 | |
| | | | | Change: 126432862 | ||
* | If you have a zoomed in TensorBoard chart, and you load new data, and you ↵ | 2016-07-01 | |
| | | | | | | unzoom, the chart should snap to the new extent of the data. Change: 126431405 | ||
* | Update generated Python Op docs. | 2016-07-01 | |
| | | | | Change: 126431173 | ||
* | Second of two CLs to simplify Polymer data-binding with TF.URIStorage. Adds ↵ | 2016-07-01 | |
| | | | | | | the URIStorage initializer and updater. Adds runs regex into URI. Change: 126429139 | ||
* | Add support for multi-arity inputs & outputs to map_fn and scan. | 2016-07-01 | |
| | | | | Change: 126427339 | ||
* | Add C++ shape inference functions for ZerosLike, Diag, DiagPart, BatchDiagPart, | 2016-07-01 | |
| | | | | | BatchMatrixDiagPart, and BatchMatrixBandPart. Change: 126427288 | ||
* | Improve error messages in feature_column_ops. | 2016-07-01 | |
| | | | | Change: 126416482 | ||
* | Removed caching of optimizer. Since optimizer may depend on a graph element ↵ | 2016-07-01 | |
| | | | | | | (global_step). Change: 126415442 | ||
* | Adds documentation for ubuntu-specific build issue seen by open-source users. | 2016-07-01 | |
| | | | | Change: 126415298 | ||
* | Update generated Python Op docs. | 2016-06-30 | |
| | | | | Change: 126374388 | ||
* | Automated rollback of change 126349886 | 2016-06-30 | |
| | | | | Change: 126374056 | ||
* | Internal change. | 2016-06-30 | |
| | | | | Change: 126373347 | ||
* | Autogenerated Change: Release TensorBoard at TAG: 21 | 2016-06-30 | |
| | | | | Change: 126371321 | ||
* | Use learn.DNNClassifier for Iris examples. Use default graph and existing ↵ | 2016-06-30 | |
| | | | | | | global step (if there is one) in Estimator, because they may have been created earlier when constructing an optimizer, as in the example iris_custom_decay_dnn. Change: 126369635 | ||
* | Add a unit test to test DNN weights/biases lookup with correct names from ↵ | 2016-06-30 | |
| | | | | | | saved checkpoints. Change: 126367935 | ||
* | Automated rollback of change 126348349 | 2016-06-30 | |
| | | | | Change: 126366720 | ||
* | Allow to specify the list of "clean termination" exceptions when creating a | 2016-06-30 | |
| | | | | | Coordinator Change: 126366456 | ||
* | Move POSIX-dependent tracing function from platform/default -> platform/posix. | 2016-06-30 | |
| | | | | Change: 126364522 | ||
* | Update generated Python Op docs. | 2016-06-30 | |
| | | | | Change: 126364170 | ||
* | Transformed Distribution | 2016-06-30 | |
| | | | | Change: 126362065 |