| Commit message (Collapse) | Author | Age |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Fix docstring.
PiperOrigin-RevId: 191747417
* Use constants in tf.zeros if the constant won't be too big.
Using fill saves on GraphDef size, but can slow down models since the
total number of ops is greater (fill + shape + constant op). This
change makes us only use fill for large shapes.
PiperOrigin-RevId: 191747456
* Fix typos in "Profile Model Float Operations" documentation.
PiperOrigin-RevId: 191751175
* Added a call in CheckpointSaverHook.after_create_session to always save
checkpoint before the first training step.
PiperOrigin-RevId: 191753026
* Document expected regular structure of the statistical testing library.
PiperOrigin-RevId: 191753693
* Refine BatchReshape error messages.
PiperOrigin-RevId: 191754120
* Include the operators module in the test framework as well.
PiperOrigin-RevId: 191756100
* Expand activity analysis to the test nodes of if and while statements.
PiperOrigin-RevId: 191756234
* Inline more functions
PiperOrigin-RevId: 191761109
* Sync only the convolutional_recurrent file to Keras 2.1.5.
PiperOrigin-RevId: 191763101
* Internal change
PiperOrigin-RevId: 191769724
* Expose odeint_fixed in tf.contrib.integrate
PiperOrigin-RevId: 191769890
* Automated g4 rollback of changelist 191761109
PiperOrigin-RevId: 191771969
* Fix final eval bottleneck creation to work in cases where it isn't cached already.
Fixes #17423
PiperOrigin-RevId: 191773001
* Fix regression caused by cl/191020868: Re-use materialized shapes for other broadcast gradient shape nodes.
PiperOrigin-RevId: 191779263
* Save the original from_proto method before calling it to avoid infinite loop.
PiperOrigin-RevId: 191784430
* Automated g4 rollback of changelist 191753026
PiperOrigin-RevId: 191784709
* [XLA] Remove a dead function and a stale todo.
PiperOrigin-RevId: 191786563
* Enable branch prediction in TensorFlow
PiperOrigin-RevId: 191788253
* Changes loss_reduction default to SUM_OVER_BATCH_SIZE for multi_class_head and binary_classification_head.
PiperOrigin-RevId: 191793392
* quantized LSTM support improvements
PiperOrigin-RevId: 191794956
* Fix TF_ImportGraphDefResults and TF_Function leaks in Python API.
PiperOrigin-RevId: 191797853
* [XLA] Better support for mul reductions in MakeFakeArguments()
Mul reductions want a 1 as their init value, not a 0 or a random value.
PiperOrigin-RevId: 191802819
* Disable tests that are currently failing with cuda 9
PiperOrigin-RevId: 191805453
* Make tf.contrib.estimator.add_metrics work with warm-starting.
PiperOrigin-RevId: 191805682
* Add Raspberry Pi section and link to github build instructions.
PiperOrigin-RevId: 191807862
* Add for and while loops to the list of operators. Do not use them yet.
PiperOrigin-RevId: 191807973
* [TF:XLA] No need to set return value in the while loop's condition.
PiperOrigin-RevId: 191809110
* Add functions to extract the basic symbols on which a composite name relies. This in turn allows to statically obtain a block's syntactic closure.
PiperOrigin-RevId: 191809965
* Add link for index file in performance tab.
PiperOrigin-RevId: 191811610
* Added an option to inline all functions in aggressive mode.
PiperOrigin-RevId: 191819577
* Make concat handler support mixed range input
PiperOrigin-RevId: 191822664
* Automated g4 rollback of changelist 191605505
PiperOrigin-RevId: 191824447
* Add a command line parameter to toco to change the way toco rescales input and output tensors.
PiperOrigin-RevId: 191825756
* refactor and add proto field required by POD support.
PiperOrigin-RevId: 191826636
* Lazily evaluate shapes with the C API enabled.
This change makes it so shapes are computed only when requested with
_USE_C_API = True. Note that the C API will still raise a shape error
if necessary when the op is created.
In addition, it cleans up the logic for _USE_C_SHAPES = True. In this
case, we lazily fetch and cache shapes directly from the C API. We no
longer need set_shapes_for_outputs at all in this case.
PiperOrigin-RevId: 191830565
* [XLA] Don't call Literal::Get in HloEvaluator's convolution loop.
This speeds up the implementation of conv because Literal::Get calls
Literal::Piece::data, which is relatively slow.
Instead, we call Literal::Data() once and cache the result.
Before: ConvolutionTest/0.StridedFilter (59094 ms)
After: ConvolutionTest/0.StridedFilter (41812 ms)
Speedup: 59/42 = 1.4x
PiperOrigin-RevId: 191830741
* Added `drop_final_batch` argument to make_batched_features_dataset. This allows the batch_and_drop_remainder function to be used instead of the default batch function.
PiperOrigin-RevId: 191831842
* Add RunMetadata logging to tf.train.ProfilerHook for Tensorboard Memeory/CPU usage visualization
PiperOrigin-RevId: 191832832
* [XLA] Don't call MultidimensionalIndexToLinearIndex in HloEvaluator's convolution routine.
Before: ConvolutionTest/0.StridedFilter (41812 ms)
After: ConvolutionTest/0.StridedFilter (28054 ms)
Speedup: 42 / 28 = 1.5x
PiperOrigin-RevId: 191835735
* Expose the adaptive sampling option for SDCA and shuffle the data when adaptive sampling is off.
PiperOrigin-RevId: 191836004
* Swap in the new implementation of while and for loops.
PiperOrigin-RevId: 191838806
* Upgrade libpng
PiperOrigin-RevId: 191840652
* Fix StringPiece use-after-free in MasterSession::ReffedClientGraph.
Use the owned ClientGraph as the source for the node_to_name_ map, rather than the borrowed GraphExecutionState (which can be deleted while the ReffedClientGraph is in use).
PiperOrigin-RevId: 191847023
* Add a test to check graceful handling of out-of-memory conditions.
PiperOrigin-RevId: 191860462
* internal change
PiperOrigin-RevId: 191869400
* Fix typos in XlaCompilationCache
PiperOrigin-RevId: 191881135
* Define PRNG seeding style for new code in Distributions and TF Probability, with rationales.
Implement lightweight PRNG for seed generation in that style.
Enables incremental refactoring of existing code into this style.
PiperOrigin-RevId: 191884573
* Avoid marking clusters containing only Identity ops for compilation.
This would produce clusters where XLA cannot optimize anything.
PiperOrigin-RevId: 191887414
* Add description to the LPIRC 2018 competition benchmarker.
PiperOrigin-RevId: 191889484
* The training model need not be built when the kfac optimizer is initialized so the
self._variables will be empty list. So pass a function which returns list of trainable variables to estimator.
PiperOrigin-RevId: 191893084
* Fix up the support for the case where a given array name occurs multiple
times in the inputs/outputs list of an op. The (non-essential) computation
of the optimal workspace size had not been updated for that case, causing it
to fail on a simple test case. Moreover, the initial implementation had some
redundant usage of std::find that this CL moves to a shared helper function.
PiperOrigin-RevId: 191894081
* Support override of device filters for gRPC, by overriding the requests with default session config.
PiperOrigin-RevId: 191895856
* Tweaked docstrings in LayerCollection.
PiperOrigin-RevId: 191897098
* [TPUClusterResolver] Start a TFServer when running in GKE
This change allows advanced input pipelines (e.g. StreamingFilesDataset, or split-pipelines that use py_func's) to run in GKE- and GKE-like enviornments.
PiperOrigin-RevId: 191897639
* [tf.data] Enable using `tf.contrib.data.prefetch_to_device()` in eager mode.
The added functionality is a substitute for the implicit prefetching in
`tfe.Iterator`, and the two paths will converge in a future change.
Fixes #18260.
PiperOrigin-RevId: 191897666
* Materialize tensor array sizes whenever possible
PiperOrigin-RevId: 191900015
* Object-based checkpointing support for unidirectional cuDNN LSTM cells
Once checked in, this will be the only way I know of to save canonical weights
when executing eagerly. Eager's name-based saving support will only do the
opaque parameter buffer.
I'm not going to try converting everything in one go, but it's a start at
least. And everything else should raise a NotImplementedError rather than
silently not saving correctly.
Single-layer cuDNN cells can be swapped for un-wrapped cuDNN compatible cells or
single cells wrapped in MultiRNNCells. Multi-layer cells need MultiRNNCell
wrapping.
PiperOrigin-RevId: 191905703
* Allow TFE_NewContext to fail more reasonably when SWIG is checking status.
Before:
TFE_Context would check nullptr, and the function would fail straight away.
Now:
TFE_Context is nullptr, so it skips down to checking the status, and an error
is raised.
I'm not able to find in SWIG documentation how to order typemaps in the
generated code - ideally, I'd order it to check the status typemap first. This
code makes it not dependent on this ordering either way.
PiperOrigin-RevId: 191905893
* Change GetInstructionCallContext to take an opcode instead of an
HloInstruction.
This enables use of the function without an actual instruction (eg, if you just
have an HloProto).
PiperOrigin-RevId: 191905914
* TPU Cost Estimator has been modified to also account for the memory cost in the execution time. Until more sophisticated methods are added, we resort to the roofline model to calculate such cost.
PiperOrigin-RevId: 191913626
* Properly handle callable objects.
PiperOrigin-RevId: 191913834
* Minor doc clarification for reduce_sum return type
PiperOrigin-RevId: 191914398
* Added headers only version of tensorflow/core/kernels:cwise_lib, cwise_lib_hdrs. This is for clients that want to use the cwise_ops machinery when making their own custom ops, including cwise_lib directly causes multiple definition linker errors.
PiperOrigin-RevId: 191914445
* [TF:XLA] Create Despecializing Pass Pipeline
When comparing backends, it is useful to take an HLO optimized for one backend and perform transformations in order to match numerics. This can be thought of as finding a lowest common denominator.
Move this grouping of passes into its own HloPassPipeline that can be reused in a few different places.
PiperOrigin-RevId: 191914799
* Update tf.keras to keras 2.1.5 version.
PiperOrigin-RevId: 191914904
* Remove `TF_InitializeTPU` and `TF_ShutdownTPU` from experimental C API as they are no longer needed. Also remove a duplicate function declaration.
PiperOrigin-RevId: 191918408
* Fix small performance regression in microbenchmarks.
PiperOrigin-RevId: 191919464
* Support RNN profiling in StreamExecutor for CUDA GPUs.
This change hasn't applied autotune on TF Cudnn kernels, only provides lower level support.
PiperOrigin-RevId: 191919566
* Validate errorReporter and improve the documentation on it.
PiperOrigin-RevId: 191920009
* Fix a few bugs in ArithmeticOptimizer and make it robust to failures of shape inference.
PiperOrigin-RevId: 191922788
* Update the rewriter options with the optimizer options
PiperOrigin-RevId: 191923287
* Pull changes from prefetching_ops to support dicts in prefetching_ops_v2 in distribute, and update estimator test to use prefetching.
Also update readme to reflect the support of dictionaries.
PiperOrigin-RevId: 191924990
* Replaced calls to deprecated tensorflow::StringPiece methods with their
tensorflow::str_util equivalents.
This will allow the deprecated methods to be removed.
PiperOrigin-RevId: 191925087
|
| |
|
|
| |
PiperOrigin-RevId: 190878279
|
| |
|
|
|
|
|
|
|
|
|
| |
This change makes CondContext._external_values more consistently store
Tensors external this context. These values are then not added to the
context when it's imported. This also removes the workaround I added
earlier to manually remove the predicate and pivot Tensors from the
context, instead adding them to _external_values were they're
automatically excluded.
PiperOrigin-RevId: 188083780
|
| |
|
|
|
|
|
| |
This is end-to-end in that it starts as TensorFlow python,
producing tflite models.
PiperOrigin-RevId: 162801924
|
| |
|
|
| |
PiperOrigin-RevId: 159591663
|
| |
|
|
| |
Change: 155209832
|
| |
|
|
| |
Change: 143230429
|
| |
|
|
| |
Change: 142318245
|
| |
|
|
| |
Change: 142080137
|
| |
|
|
|
|
| |
tf.train.SummaryWriterCache.
Change: 140406300
|
| |
|
|
| |
Change: 138675832
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Supervisor and graph_actions now do this by default if saving a graph is requested.
TensorBoard can now fall back to using the MetaGraphDef if the GraphDef is not
available in the Event.
Other changes include:
* factor out metagraph code from training.saver to framework.meta_graph
* update the summary writer, session hooks, and supervisor to support these new summaries
Change: 135424442
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- Split SummaryWriter implementation into two pieces:
*** SummaryToEventTransformer, which implements the functional API that transforms summaries and other data into Event protocol buffers
*** EventFileWriter, which takes Event protocol buffers and writes them to disk
As the summary-transformer logic in SummaryWriter grows more complicated, this decomposition makes it much easier to test. Also, it will faciltiate adding other event-writing backends in the future.
- Create LegacySummaryWriter, which is the composition of those two pieces and has an API that is identical to the current tf.train.SummaryWriter
- Move the SummaryWriter code into python/summary and out of python/training
- Reimport the code to training/summary_io, so that no one downstream notices any changes.
Change: 135417373
|
| |
|
|
| |
Change: 127109341
|
| |
|
|
|
| |
non-deterministic.
Change: 125689946
|
| |
|
|
| |
Change: 125134620
|
| |
|
|
| |
Change: 123903104
|
| |
|
|
| |
Change: 123901292
|
| |
|
|
| |
Change: 123791949
|
| |
|
|
| |
Change: 118590392
|
| |
|
|
| |
Change: 115626850
|
|
|
Change: 115578243
|
Rasmus Munk Larsen
Martin Wicke
Skye Wanderman-Milne
Andrew Selle
A. Unique TensorFlower
Dan Ringwalt
Justine Tunney
Dan Mané
Benoit Steiner
Eugene Brevdo
A. Unique TensorFlower
Olivia Nordquist
Vijay Vasudevan