aboutsummaryrefslogtreecommitdiffhomepage
Commit message (Collapse)AuthorAge
...
* Fix profiler to track some missed persistent bytes.Gravatar A. Unique TensorFlower2017-11-30
| | | | PiperOrigin-RevId: 177516249
* Extract out a MathUtil::GCD helperGravatar Sanjoy Das2017-11-30
| | | | | | This fixes a TODO. PiperOrigin-RevId: 177508258
* TF server should not crash when -v=1 is enabled.Gravatar A. Unique TensorFlower2017-11-30
| | | | | | These WriteTextProto() calls are purely for diagnostics (and are usually called within IF_VLOG_IS_ON(1) guards), but if they fail to write to a file, they'll take down the entire calling process. Which makes debugging difficult, and seems rather astonishing. PiperOrigin-RevId: 177506379
* Disable state_saving_rnn_estimator_test in asan mode.Gravatar Gunhan Gulsoy2017-11-30
| | | | PiperOrigin-RevId: 177506166
* Automated g4 rollback of changelist 177499365Gravatar Gunhan Gulsoy2017-11-30
| | | | PiperOrigin-RevId: 177505909
* Enable tests that pass now with the new copy insertion.Gravatar A. Unique TensorFlower2017-11-30
| | | | PiperOrigin-RevId: 177503567
* [TF:XLA] Allow bfloat16 types in more places.Gravatar Peter Hawkins2017-11-30
| | | | PiperOrigin-RevId: 177502497
* Add R1 slice tests.Gravatar A. Unique TensorFlower2017-11-30
| | | | PiperOrigin-RevId: 177502259
* Support binary operations with a scalar and a 4d tensor as input; refactorGravatar Yao Zhang2017-11-30
| | | | | | connectivity check code. PiperOrigin-RevId: 177499365
* Add an option to override maximum number of elements in the quantileGravatar A. Unique TensorFlower2017-11-30
| | | | | | accumulator. PiperOrigin-RevId: 177497240
* Add rules to replace nodes corresponding to operations with the neutral and ↵Gravatar A. Unique TensorFlower2017-11-30
| | | | | | | | absorbing elements for addition and multiplication with Identity. Get rid of a gratuitous copy of the entire graph in the main optimizer loop. PiperOrigin-RevId: 177491247
* [XLA:CPU] Add Hlo profiling support to XlaJitCompiledCpuFunctionGravatar Sanjoy Das2017-11-30
| | | | | | | | | | | | | | | Some of the functionality has bled into the generic XlaCompiledCpuFunction, but there still remains a fair amount of work to do before the AOT side of things start working. This CL also fixes a bug I introduced in a previous CL -- when I changed IrEmitter::hlo_to_profile_idx_ to a value, I changed the signature of the generated function to always have the "profile_counters" argument when the AOT client code expects the signature to not have that argument. In practice this wasn't an issue for the standard x86 calling convention, but it could easily have been problematic on other architectures and calling conventions. After this change the mismatch is no longer present. PiperOrigin-RevId: 177481998
* [XLA] Adds Dot with DotDimensionNumbers proto for specifying arbitrary ↵Gravatar A. Unique TensorFlower2017-11-30
| | | | | | contracting and batch dimensions. PiperOrigin-RevId: 177481231
* Change depthwise convolution filter expansion and contraction withGravatar Blake Hechtman2017-11-30
| | | | | | algebraic manipulation instead of slices and pads that are more difficult to fuse. PiperOrigin-RevId: 177480353
* Disable dnn_linear_combined_testGravatar Gunhan Gulsoy2017-11-30
| | | | PiperOrigin-RevId: 177478106
* Internal testing change.Gravatar Russell Power2017-11-30
| | | | PiperOrigin-RevId: 177478003
* Turned a verbose log into a vlogGravatar Benoit Steiner2017-11-30
| | | | PiperOrigin-RevId: 177474943
* Change "Datasets" to "`tf.data`" in the "Reading Data" API guide.Gravatar Derek Murray2017-11-30
| | | | PiperOrigin-RevId: 177473833
* Uses C API for eager functions.Gravatar Alexandre Passos2017-11-30
| | | | | | Rolls back the rollback with some swiggery to get python3 to work. PiperOrigin-RevId: 177470328
* [TF:XLA] Add support for the V2 variants of the FusedBatchNorm operators, ↵Gravatar Peter Hawkins2017-11-30
| | | | | | | | which support mixed precision training. Until the necessary support for mixed precision fused batch norm is added to XLA, implement by casting to a common type. PiperOrigin-RevId: 177468202
* Rename tests.Gravatar A. Unique TensorFlower2017-11-30
| | | | PiperOrigin-RevId: 177467740
* Hoist function input placeholders out of any control flow context.Gravatar Skye Wanderman-Milne2017-11-30
| | | | | | | | | | | | | | | | Prior to this change, functions that closed over external tensors in a while loop would cause a segfault at runtime. This is because the external tensors are temporarily represented as placeholders in the function body before being replaced by input parameters, and the placeholders would be created directly in the while loop body. This would eventually lead to using the input tensor in the while loop body without an enter node. This wasn't caught by the runtime check because it isn't applied to function bodies. This change adds tests for capturing tensors in a while loop body and in a cond context. Note that the cond test passed without this fix. PiperOrigin-RevId: 177464541
* Implement Python-specific device and colocation logic in import_graph_def ↵Gravatar Skye Wanderman-Milne2017-11-30
| | | | | | with C API enabled. PiperOrigin-RevId: 177462553
* Add support for int32 output types to the Multinomial op.Gravatar Peter Hawkins2017-11-30
| | | | PiperOrigin-RevId: 177444775
* Disable baseline_test in asan.Gravatar Gunhan Gulsoy2017-11-30
| | | | PiperOrigin-RevId: 177423981
* Automated g4 rollback of changelist 177362829Gravatar A. Unique TensorFlower2017-11-30
| | | | PiperOrigin-RevId: 177419730
* Automated g4 rollback of changelist 177375237Gravatar Gunhan Gulsoy2017-11-30
| | | | PiperOrigin-RevId: 177418947
* Always leverage shapes inference now that it can handle fed nodesGravatar Benoit Steiner2017-11-29
| | | | | | conservatively. PiperOrigin-RevId: 177391746
* Add native dilated support for conv2d and its gradients in cudnn v>=6.Gravatar Yangzihao Wang2017-11-29
| | | | PiperOrigin-RevId: 177382431
* Change `tf.contrib.distributions` docstring examples to use `tfd` alias ↵Gravatar Joshua V. Dillon2017-11-29
| | | | | | rather than `ds`, `bs`. PiperOrigin-RevId: 177381853
* Fix more clang-tidy warnings:Gravatar Eli Bendersky2017-11-29
| | | | | | | - Parameter names consistent in function declarations and definitions - Class members naming PiperOrigin-RevId: 177379085
* Using the C API in eager mode for graph functions.Gravatar Alexandre Passos2017-11-29
| | | | PiperOrigin-RevId: 177375237
* Automated g4 rollback of changelist 177353959Gravatar Gunhan Gulsoy2017-11-29
| | | | PiperOrigin-RevId: 177371177
* Updating references to the `tf.data` API to `tf.data` from `Datasets`.Gravatar Jiri Simsa2017-11-29
| | | | PiperOrigin-RevId: 177367024
* Add R1 slice tests.Gravatar A. Unique TensorFlower2017-11-29
| | | | PiperOrigin-RevId: 177362829
* Check when session cannot run because its graph was modifiedGravatar Igor Ganichev2017-11-29
| | | | | | | | | | | | | With current tensorflow code, if user modifies some operation after session.run() was called, this modification will never make it to the C++ runtime and no errors will be raised leading to silent wrong results. This change adds checks for such cases when C API is enabled. We don't change the code path for C API being disabled because C API should be enabled by default soon. PiperOrigin-RevId: 177359630
* Add a way to query a batch scheduler to determine the max task size.Gravatar Christopher Olston2017-11-29
| | | | | | A layer on top of the batcher could use this interface to pre-split large tasks that exceed the max batch size. PiperOrigin-RevId: 177359263
* Simplify const node creation.Gravatar Yao Zhang2017-11-29
| | | | PiperOrigin-RevId: 177357416
* Re-enable Mul hoisting for aggregations other than Add when input shapes match.Gravatar A. Unique TensorFlower2017-11-29
| | | | PiperOrigin-RevId: 177356621
* Added comment/TODO concerning memory use of extract_images_patches.Gravatar A. Unique TensorFlower2017-11-29
| | | | PiperOrigin-RevId: 177354924
* Proper deallocation in the thread-local tape stack.Gravatar Alexandre Passos2017-11-29
| | | | PiperOrigin-RevId: 177354350
* enabling Tensor._set_shape() to work with the C APIGravatar Olivia Nordquist2017-11-29
| | | | PiperOrigin-RevId: 177353959
* Clarify the role of replicate_model_fn.Mode better.Gravatar Igor Saprykin2017-11-29
| | | | PiperOrigin-RevId: 177351409
* Deleted unused method argumentsGravatar Benoit Steiner2017-11-29
| | | | PiperOrigin-RevId: 177350575
* Allow the toolchain defaults to be used instead of hard-coding -Os.Gravatar A. Unique TensorFlower2017-11-29
| | | | | | | For example toolchains with clang may set -Oz which is more analogous to gcc's -Os. -Os for clang is closer to -O2. PiperOrigin-RevId: 177347371
* Made sure the unknown shapes of placeholders always propagate to their fanoutsGravatar Benoit Steiner2017-11-29
| | | | PiperOrigin-RevId: 177344207
* [tf.data] Fix compiler warnings about unused captures in lambda expressions.Gravatar Derek Murray2017-11-29
| | | | PiperOrigin-RevId: 177343020
* Add RandomDataset which generates pseudo random number of type int64.Gravatar Saurabh Saxena2017-11-29
| | | | | | Add tf.contrib.data.shuffle_and_repeat which reshuffles its input on each epoch. Going forward, this will replace reshuffle_each_iteration=true. PiperOrigin-RevId: 177339570
* (Temporarily) call Graph._add_op outside of Operation.__init__ again.Gravatar Skye Wanderman-Milne2017-11-29
| | | | | | | | | | | | | | | | | This change partially undoes my previous commit (https://github.com/tensorflow/tensorflow/commit/f4c18a0eb05e21bae397c9c16527ff8080cae6b8). Without this change, if an op is added that has invalid input shapes and also requires a kernel label, the op will be added to the graph before shape inference is run, but then the shape inference error will prevent the kernel label from being applied. The placer will then complain about the missing label when the graph is run. This is only a problem with the C API disabled. With the C API enabled, shape inference is performed when the TF_Operation is created in Operation.__init__. Thus we can and should move the _add_op call back to Operation.__init__ once the _USE_C_API flag is removed. PiperOrigin-RevId: 177338123
* Go: Bugfix: Make list-of-shape attributes in an operation work.Gravatar Asim Shankar2017-11-29
| | | | | | | | | | | | | By respecting cgo rules on pointers. Without the change to graph.go, the newly added test would fail with: panic: runtime error: cgo argument has Go pointer to Go pointer in the call to the C function TF_SetAttrShapeList. Fixes #14891 PiperOrigin-RevId: 177336663
generated by cgit v1.2.3 (git 2.39.1) at 2024-10-18 16:18:54 +0000