Commit message (Collapse) | Author | Age | |
---|---|---|---|
* | Add cuda_is_configured() build macro for testing whether CUDA was enabled at ↵ | 2016-12-08 | |
| | | | | | | configure time. Change: 141474932 | ||
* | Added config to the wide-n-deep model-fns. | 2016-12-08 | |
| | | | | Change: 141474204 | ||
* | Update generated Python Op docs. | 2016-12-08 | |
| | | | | Change: 141473210 | ||
* | Fix bug in handling of explicitly specified num parameter of split_v | 2016-12-08 | |
| | | | | | and also add associated test case. Change: 141469479 | ||
* | Update generated Python Op docs. | 2016-12-08 | |
| | | | | Change: 141468149 | ||
* | Bump the tolerance in linear_equations_test.py to fix the AVX build on macs. | 2016-12-08 | |
| | | | | Change: 141467544 | ||
* | Fix FileIO.read() to match file.read() behavior and not raise an exception ↵ | 2016-12-08 | |
| | | | | | | on EOF Change: 141466762 | ||
* | [Android]: Add -lm and -llog to libtensorflow_inference.so to solve ↵ | 2016-12-08 | |
| | | | | | | standalone lib link errors. Change: 141465701 | ||
* | Change the protobuf dependency from == 3.1.0 to >= 3.1.0 so it doesn't | 2016-12-08 | |
| | | | | | reinstall when installing tensorflow. Change: 141465425 | ||
* | Introduced tf.train.HookedSession which is designed to help single machine ↵ | 2016-12-08 | |
| | | | | | | training. Change: 141462689 | ||
* | Update generated Python Op docs. | 2016-12-08 | |
| | | | | Change: 141462624 | ||
* | Update ops-related pbtxt files. | 2016-12-08 | |
| | | | | Change: 141461377 | ||
* | Disable all tests depending on *half_plus_two* genrules. | 2016-12-08 | |
| | | | | Change: 141461172 | ||
* | SparseTensor.shape -> SparseTensor.dense_shape part 2... | 2016-12-08 | |
| | | | | Change: 141460319 | ||
* | Pass estimator config to the model-fn. | 2016-12-08 | |
| | | | | Change: 141460215 | ||
* | Tidy up some text in basic usage documentation. | 2016-12-08 | |
| | | | | Change: 141458853 | ||
* | Add an option for choosing the J_DCT_METHOD for jpeg decoding to | 2016-12-08 | |
| | | | | | DecodeJpegOp. Change: 141457003 | ||
* | Remove undefined function GetRunfilesDirectory. | 2016-12-08 | |
| | | | | Change: 141447099 | ||
* | Import tf.contrib.rnn for cells instead of tf.nn.rnn_cell. | 2016-12-08 | |
| | | | | Change: 141440475 | ||
* | Sort op nodes so as to make HVX execution more robust | 2016-12-08 | |
| | | | | Change: 141428065 | ||
* | Fix inference execution on HVX | 2016-12-08 | |
| | | | | Change: 141420253 | ||
* | Refactor Python imports: layers | 2016-12-08 | |
| | | | | Change: 141412543 | ||
* | Prepare all callers of sampled_softmax_loss to be ready for inputs and ↵ | 2016-12-08 | |
| | | | | | | labels swap. Change: 141410359 | ||
* | Fix static initialization based registry patterns with shared libraries when ↵ | 2016-12-08 | |
| | | | | | | built with --config=cuda. Change: 141406402 | ||
* | Make base core layer match the behavior of tf.get_variable with regard to ↵ | 2016-12-08 | |
| | | | | | | regularization. Change: 141405902 | ||
* | Use GraphTransferer to initialize graph on SOC | 2016-12-07 | |
| | | | | Change: 141394565 | ||
* | tfdbg core: remove unnecessary construction of DebuggerState when there is ↵ | 2016-12-07 | |
| | | | | | | | | no watch Hitchhiking change: * A typo fix Change: 141386349 | ||
* | Bug fix: Allow int64 inputs to tf.sequence_mask. | 2016-12-07 | |
| | | | | Change: 141385792 | ||
* | Improve experiment_test to not rely on actual sleep timing. Use patch instead. | 2016-12-07 | |
| | | | | Change: 141384001 | ||
* | Update generated Python Op docs. | 2016-12-07 | |
| | | | | Change: 141383834 | ||
* | Shorten summary tag names for queues now that the scopes are automatically | 2016-12-07 | |
| | | | | | added. Change: 141382469 | ||
* | Make RunConfig immutable. | 2016-12-07 | |
| | | | | Change: 141379749 | ||
* | Update generated Python Op docs. | 2016-12-07 | |
| | | | | Change: 141379710 | ||
* | Fix bug in tf.einsum() when input is placeholder. | 2016-12-07 | |
| | | | | Change: 141378762 | ||
* | Fix iteritems break in python3 | 2016-12-07 | |
| | | | | Change: 141377833 | ||
* | Add named arguments to all uses of tf.split and reverse the order of arguments | 2016-12-07 | |
| | | | | | | so as to conform to the order that will be standard once the split_v -> split rename happens. Change: 141377137 | ||
* | Update documentation to explain how to enable SSE/AVX when compiling from source | 2016-12-07 | |
| | | | | Change: 141376599 | ||
* | Export the framework/versions.h header file. | 2016-12-07 | |
| | | | | Change: 141376479 | ||
* | Import tf.contrib.rnn for cells instead of tf.nn.rnn_cell. | 2016-12-07 | |
| | | | | Change: 141376440 | ||
* | Separate the stream executor CUDA plugin into its own build target. This ↵ | 2016-12-07 | |
| | | | | | | | allows non --config=cuda builds to link the CUDA plugins if desired. Add build macro if_cuda_is_configured() that tests whether CUDA was enabled at configure time, rather than whether the current BUILD is using a CUDA compiler. Change: 141375030 | ||
* | Update generated Python Op docs. | 2016-12-07 | |
| | | | | Change: 141374489 | ||
* | Remove tf.select op. tf.where now supports tf.select functionality and | 2016-12-07 | |
| | | | | | tf.select is deprecated. Change: 141374194 | ||
* | Remove tf.batch_matmul Python interface. | 2016-12-07 | |
| | | | | Change: 141373915 | ||
* | Add a decorator to enforce function be called with keyworded args (only). | 2016-12-07 | |
| | | | | Change: 141373569 | ||
* | Invoke APIs of SOC wrapper library to execute inference on HVX | 2016-12-07 | |
| | | | | Change: 141373481 | ||
* | Set default config in canned estimators to avoid config discrepancy between ↵ | 2016-12-07 | |
| | | | | | | estimator and model_fn. Change: 141371292 | ||
* | Test summary tags in head. | 2016-12-07 | |
| | | | | Change: 141371278 | ||
* | Extend half-plus-two saved model to have an additional execution path, for ↵ | 2016-12-07 | |
| | | | | | | testing models with multiple signatures. Also fix its existing signature. Change: 141371192 | ||
* | Add named arguments to all uses of tf.split and reverse the order of arguments | 2016-12-07 | |
| | | | | | | so as to conform to the order that will be standard once the split_v -> split rename happens. Change: 141371070 | ||
* | Add a blacklist of variables to graph_freeze.py. | 2016-12-07 | |
| | | | | Change: 141369603 |