This repository was archived by the owner on Jan 3, 2023. It is now read-only.

WIP: TF threading#5

Open

rsdubtso wants to merge 10000 commits intoNervanaSystems:masterfrom

rsdubtso:master

This pull request is big! We're only showing the most recent 250 commits

Commits on Jul 15, 2019

[Grappler] Generalize Squeeze and Reduce op transformations in GenericLayoutOptimizer.

andyly
authored and
tensorflower-gardener
committed
tensor tracer: minor change in flush function.
tensorflower-gardener
committed
Add spaces to make string replace match
jbedorf
committed
Add support for Python 3 type-annotated assign. Fixes #28849.

Dan Moldovan
authored and
tensorflower-gardener
committed
Do not accumulate loop invariants in while_v2.

saxenasaurabh
authored and
tensorflower-gardener
committed
AveragePrecision@k(e) for an example e is defined as
tensorflower-gardener
committed
Create zeros of correct type when building default gradient for resources.

saxenasaurabh
authored and
tensorflower-gardener
committed
Handle shape inference for "outside compilation to outside compilation data edge" for If/While nodes as well.

er91
authored and
tensorflower-gardener
committed
Use OP_REQUIRES_ASYNC in CollectiveGatherOpKernel.

dubey
authored and
tensorflower-gardener
committed
Adds run_eagerly option for bert.

saberkun
authored and
tensorflower-gardener
committed
[tf.data] Include "(Experimental)ParseExampleDataset" as a passthrough op for rebatching.

Rachel Lim
authored and
tensorflower-gardener
committed
Use declarative rewrite pattern for Gather(V2) legalization

antiagainst
authored and
tensorflower-gardener
committed
Add test where mixed precision is enabled when saving and disabled when loading, and vice versa.

reedwm
authored and
tensorflower-gardener
committed
Merge pull request #30718 from tensorflow:master-numpy
tensorflower-gardener
committed
Merge pull request #30603 from tensorflow:release_1.14.0
tensorflower-gardener
committed
Update select_training_loop to discard inferred default strategy.

qlzh727
authored and
tensorflower-gardener
committed
Preserve input nodes even if they are not used while importing GraphDef to MLIR

smit-hinsu
authored and
tensorflower-gardener
committed
Clif wraps the TPU topology proto target to enable use in clif files.

sudroy
authored and
tensorflower-gardener
committed
Update Eigen to commit 049af2f56331.

ezhulenev
authored and
tensorflower-gardener
committed
Renamed enum from cc3-apache2 to cc-apache.
tensorflower-gardener
committed
Disable failing msan and asan tests until fix is landed.

bmzhao
authored and
tensorflower-gardener
committed
Merge pull request #30586 from siju-samuel:lite_deprecated_tfrun_removal
tensorflower-gardener
committed
Remove --spawn_strategy=standalone and --strategy=Genrule=standalone from bazelrc file
tensorflower-gardener
committed
Refactor OptimizeDatasetOp
feihugis
committed
Automated rollback of commit acab6a20512f29be48b8252db00a388d2c57c29c

allenlavoie
authored and
tensorflower-gardener
committed
Add `--quantize_to_float16` flag to tflite_convert

jdduke
authored and
tensorflower-gardener
committed
TFLite GPU Delegate: Introduce stretch_time operation options.
tensorflower-gardener
committed
Merge pull request #30663 from yongtang:07122019-window-docstring
tensorflower-gardener
committed
Add back a (very hacky) executing_eagerly check to avoid computing second-order gradients where we weren't before cl/256268751

allenlavoie
authored and
tensorflower-gardener
committed
Update TensorFlow API generation so that it uses import statement instead of calling importlib.

yifeif
authored and
tensorflower-gardener
committed
tensor tracer: moving the tf.case that determines the file to flush into the outside_compilation region.
tensorflower-gardener
committed
Metal: 'Add' operation supports scalar.
tensorflower-gardener
committed
Switch FunctionLibraryDefinition to share FunctionDef storage with other instances.

mrry
authored and
tensorflower-gardener
committed
Merge pull request #30723 from feihugis:Refactor_DatasetOps_8
tensorflower-gardener
committed
Minor changes to TFLiteConverter 2.0 documentation.

Nupur Garg
authored and
tensorflower-gardener
committed
Added multi threading support for SnapshotReaderIterator. Currently this doesn't ensure any determinism or ordering. This is especially useful when the data is compressed because the decompression …

rohan100jain
authored and
tensorflower-gardener
committed
Changing test size for topk_op_test_xla_gpu.

bmzhao
authored and
tensorflower-gardener
committed
Make ReachableFunctionLibraryDefinition deterministic in its output by replacing absl::flat_hash_set with std::set.
tensorflower-gardener
committed
[XLA] Expose HloInstruction::CreateBitcast method.

cheshire
authored and
tensorflower-gardener
committed
Minor changes to keras preprocessing layer test utilities

askerry
authored and
tensorflower-gardener
committed

Commits on Jul 16, 2019

Commits on Jul 17, 2019

Transfer the ownership of the calibrated TRT engine to the engine cache, so

aaroey
authored and
tensorflower-gardener
committed
Adding missing argument to tflite_plugin_create_delegate ctypes declaration.
tensorflower-gardener
committed
Internal change

blakehechtman
authored and
tensorflower-gardener
committed
Merge pull request #30170 from DavidNorman:allow-disable-dot-to-multiply
tensorflower-gardener
committed
Compatibility compile fix
tensorflower-gardener
committed
Clean up the way logging is configured and improve error checking. Make failure conditions in micro_interpreter clearer. Reduce tensor arena size in micro_vision so that it builds for Sparkfun Edge.
tensorflower-gardener
committed
[XLA]Fix remat test.

yunxing
authored and
tensorflower-gardener
committed
Do not import contrib.cloud if it is not available.

gunan
authored and
tensorflower-gardener
committed
[TF:XLA] Update resource_operation_safety_analysis file comment.
tensorflower-gardener
committed
[XLA] Teach HloInstruction::OperandElementUse about kGather and fix its handling of kBitcast.

Justin Lebar
authored and
tensorflower-gardener
committed
Fix a Hub export test

allenlavoie
authored and
tensorflower-gardener
committed
Tweaks to 1.x SavedModel loading with Defuns and custom ops

allenlavoie
authored and
tensorflower-gardener
committed
[XLA] Mark div/rem by a constant as not-expensive.

Justin Lebar
authored and
tensorflower-gardener
committed
Significant number of changes to support the Keras v2 loops.
tensorflower-gardener
committed
Quash a -Wshadow-ivar trigger

bbarenblat
authored and
tensorflower-gardener
committed
Support `ReverseSequence` in tflite mlir converter.

haozha111
authored and
tensorflower-gardener
committed
Switch to using ruy::DetectDotprod() for sdot instruction support
tensorflower-gardener
committed
Release python interpreter lock in TFE_TensorHandleToNumpy, since this method may be blocked.

Xiao Yu
authored and
tensorflower-gardener
committed
Unit tests should use file in temporary directory as config file for CLIConfig

lindong28
authored and
tensorflower-gardener
committed
Merge pull request #30669 from Intel-tensorflow:lakshay-ut-fix-selective_registration_header_test
tensorflower-gardener
committed
Simplify for consistency
habernal
authored
compat: Update forward compatibility horizon to 2019-07-17
tensorflower-gardener
committed
Update GraphDef version to 99.
tensorflower-gardener
committed
Removed the deprecated API from the File.
Amit Srivastava
committed
Removed the deprecated API from the File.
Amit Srivastava
committed
Removed the deprecated API from the File.
Amit Srivastava
committed
Removed the deprecated API from the File.
Amit Srivastava
committed
Update MLIR revision.

chsigg
authored and
tensorflower-gardener
committed
Merge pull request #29224 from houtoms:pr_fast_3dy_reduce
tensorflower-gardener
committed
Add tests for transpose convolution with constant inputs
tensorflower-gardener
committed
Fix a compilation error when defining "TFLITE_CUSTOM_OPS_HEADER" macro.

multiverse-tf
authored and
tensorflower-gardener
committed
Merge pull request #30739 from feihugis:Refactor_DatasetOps_9
tensorflower-gardener
committed
Add int16 support to Quant.

jianlijianli
authored and
tensorflower-gardener
committed
Make LinearOperatorDiag tape safe. Also add a standard "check_tape_safe" method in the base test class.

langmore
authored and
tensorflower-gardener
committed
Avoid converting the old-style print statement, because it would otherwise generate a syntax error.

Dan Moldovan
authored and
tensorflower-gardener
committed
Workaround for CI build problems with Arduino nightly

petewarden
authored and
tensorflower-gardener
committed
Add documentation for building/using the TensorFlow Lite AAR

jdduke
authored and
tensorflower-gardener
committed
Update GPU delegate documentation

jdduke
authored and
tensorflower-gardener
committed
Merge pull request #29939 from terryky:fix_tflite_detect_postprocess
tensorflower-gardener
committed
Checks shape compatibility between `initial_value` and `shape` in ResourceVariable's constructor.

wangpengmit
authored and
tensorflower-gardener
committed
[XLA GPU] Factor out helper utils for FileCheck testing into gpu_codegen_test

cheshire
authored and
tensorflower-gardener
committed
Update keras v2 optimizers to reuse coefficients which are shared across all updates, which reduces the total number of ops created by between 5% (for simple optimizers such as SGD and Adagrad) and…

Taylor Robie
authored and
tensorflower-gardener
committed
Export LICENSE.TXT file. Update BUILD file.

Prakalp Srivastava
authored and
tensorflower-gardener
committed
Register GPU kernel for TensorListElementShape

ezhulenev
authored and
tensorflower-gardener
committed
Merge pull request #30760 from yongtang:use_resource
tensorflower-gardener
committed
Add tests for TensorSpec.name

edloper
authored and
tensorflower-gardener
committed
Move IsLocal() into Device and return false only for RemoteDevice.
tensorflower-gardener
committed
Fix control dependencies in Keras distributed TPU loops.

chr1sj0nes
authored and
tensorflower-gardener
committed
Disable test_nonlocal_symbol unit test

lindong28
authored and
tensorflower-gardener
committed
Update all the TPUStrategy examples to use the new v2 APIs, i.e.

rxsang
authored and
tensorflower-gardener
committed
Enable roundtrip of all TF tests through MLIR using the with_mlir_support flag.

Prakalp Srivastava
authored and
tensorflower-gardener
committed
Merge pull request #30053 from yongtang:30040-BinaryCrossEntropy
tensorflower-gardener
committed
Avoid copying a FunctionDef in CapturedFunction.

mrry
authored and
tensorflower-gardener
committed
Improve micro_vision README to include detailed instructions for running on the
tensorflower-gardener
committed
Remove no_windows tab for fixed flaky tests.

Yuefeng Zhou
authored and
tensorflower-gardener
committed
Allow building of MLIR converter in OSS.

aselle
authored and
tensorflower-gardener
committed
Have `model_to_estimator` make a V2 Estimator if TF2 is enabled.

omalleyt12
authored and
tensorflower-gardener
committed
[ROCm] Fix for the broken `--config=rocm` build
deven-amd
committed
Allow specififying the nnapi accelerator in benchmark_model

jdduke
authored and
tensorflower-gardener
committed
Tweak build rules in lite/java

jdduke
authored and
tensorflower-gardener
committed
Add interfaces to modify an HLO module after already being scheduled.
tensorflower-gardener
committed
Remove dead link.
lamberta
authored
Automated rollback of commit 7a2ce469a70e55855b09030a96d2f6eeb825e253

lindong28
authored and
tensorflower-gardener
committed
Disable the test_annotated_assign unit test because the annotation syntax is only recognized by Python 3.6 or later

lindong28
authored and
tensorflower-gardener
committed
Add mixed precision tests for Sequential/subclassed models.

reedwm
authored and
tensorflower-gardener
committed
Automated rollback of commit 18b2f26832dabd864b5a79403f1d92b45e081e9d

fdxmw
authored and
tensorflower-gardener
committed
Delete broadcastable trait TensorFlow and TFLite tests

smit-hinsu
authored and
tensorflower-gardener
committed
[Keras] Use math_ops.divide() instead of math_ops.div() in learning_rate_schedule.py.

mrry
authored and
tensorflower-gardener
committed
Add basic support for float FusedConvBiasActivation in the tensorrt bridge.
tensorflower-gardener
committed
Add experimental implementation of cancelable eager function execution.

mrry
authored and
tensorflower-gardener
committed
Fixes some bugs in how the v2 loop handles batch callbacks when it runs out of data. Also makes it always recreate the validation data iterator between epochs even if validation_steps has been set,…
tensorflower-gardener
committed
Don't distribute Keras models using cloning in Eager mode.

isaprykin
authored and
tensorflower-gardener
committed
Do not infer steps per epoch when auto_sharding is enabled in multi_worker_mode.

Yuefeng Zhou
authored and
tensorflower-gardener
committed
Creating a MultiDeviceIteratorV2 op that parallels IteratorV2 op and provides a python iterator style interface. Also handles deletion of MultiDeviceIteratorResources in the eager and function buil…

rohan100jain
authored and
tensorflower-gardener
committed
Refactor PostprocessLiftedArgs to facilitate adding support for functional If node.

gmagogsfm
authored and
tensorflower-gardener
committed
Multi threading SnapshotWriterIterator in a fashion similar to

rohan100jain
authored and
tensorflower-gardener
committed
Fix toco_python_api check for valid GraphDefInfo in Python 3.

Nupur Garg
authored and
tensorflower-gardener
committed
Add API for checking whether v2 control flow is enabled.

saxenasaurabh
authored and
tensorflower-gardener
committed
Resave deepdream notebook using colab

MarkDaoust
authored and
tensorflower-gardener
committed

Commits on Jul 18, 2019

Improve integration of TPUClusterResolver with connect_to_cluster().

Revan Sopher
authored and
tensorflower-gardener
committed
[XLA] Simplify cudnn convolution autotuning cache key.

timshen91
authored and
tensorflower-gardener
committed
Skip test if tensorflow std server could not be started.

Yuefeng Zhou
authored and
tensorflower-gardener
committed
[XLA] Validate fusion shape vs. the shape of the called computation

cheshire
authored and
tensorflower-gardener
committed
Temporarily disable scoped_allocator_optimizer_test under TSAN.

dubey
authored and
tensorflower-gardener
committed
Add explicit comment for ModifyGraphWithDelegate about delegate lifetime.

karimnosseir
authored and
tensorflower-gardener
committed
Handle zero batch case in BatchNorm only when 'experimental_enable_get_next_as_optional' is set in strategy object.

rxsang
authored and
tensorflower-gardener
committed
Filter Python 3.6+ test in the build file, since skipTest has no effect.

Dan Moldovan
authored and
tensorflower-gardener
committed
Remove Eager Profiler
tensorflower-gardener
committed
Update ops-related pbtxt files.
tensorflower-gardener
committed
Add strided_slice np style to generate_zip tests.

renjie-liu
authored and
tensorflower-gardener
committed
Initial version of HLO translation tools.

authored and
tensorflower-gardener
committed
[TFLite] Defining the "FULL" LSTM op in TFLite dialect.

ashwinmurthy
authored and
tensorflower-gardener
committed
Clarify parse_sequence_example and parse_single_sequence_example documentation.
tensorflower-gardener
committed
compat: Update forward compatibility horizon to 2019-07-18
tensorflower-gardener
committed
Update GraphDef version to 100.
tensorflower-gardener
committed
Add delegate support for QUANTIZE

galarragas
authored and
tensorflower-gardener
committed
Merge pull request #30692 from habernal:patch-1
tensorflower-gardener
committed
Workaround for MSVC 14.0 limitation related to template alias.

chsigg
authored and
tensorflower-gardener
committed
Merge pull request #30765 from yegle:patch-1
tensorflower-gardener
committed
Whitelist additional built-in modules, and alphabetize the list.

Dan Moldovan
authored and
tensorflower-gardener
committed
Copybara fix for proto2 namespace in object detection stage

Sachin Joglekar
authored and
tensorflower-gardener
committed
Added benchmarks for scalar caching in eager mode

superbobry
authored and
tensorflower-gardener
committed
Avoid hard-coded device assignment.
tensorflower-gardener
committed
Merge pull request #30796 from amitsrivastava78:fix_dep
tensorflower-gardener
committed
Ruy: Fix asm instruction syntax
tensorflower-gardener
committed
[Grappler] Update GenericLayoutOptimizer layout selection.

andyly
authored and
tensorflower-gardener
committed
Minor cleanup for the placer_inspection_required_ops_utils.

qlzh727
authored and
tensorflower-gardener
committed
tensorflow::ConvertToEagerTensor now always produces a handle of the requested dtype

superbobry
authored and
tensorflower-gardener
committed
Add force_concurrent_kernels flag to device_tracer.

Andiry
authored and
tensorflower-gardener
committed
Add LRN op to TFLite MLIR Converter

karimnosseir
authored and
tensorflower-gardener
committed
Removing the init_scope from table initialization except for control flow

rohan100jain
authored and
tensorflower-gardener
committed
Better workaround for MSVC 14.0 limitation related to constexpr array.

chsigg
authored and
tensorflower-gardener
committed
Workaround for MSVC 14.0 limitation related to constexpr array.

chsigg
authored and
tensorflower-gardener
committed
Define PLATFORM_WINDOWS when compiling XLA for windows. Tensorflow core headers require it.

chsigg
authored and
tensorflower-gardener
committed
Ensure device_ == nullptr if it is the HostCPU

jaingaurav
authored and
tensorflower-gardener
committed
Merge pull request #30823 from ROCmSoftwarePlatform:google_upstream_rocm_platform_fix_190717
tensorflower-gardener
committed
Distribute blocking threads across requests using exponential distribution

lindong28
authored and
tensorflower-gardener
committed
Internal change

Akshay Modi
authored and
tensorflower-gardener
committed
[XLA] Skip two tests for XLA.

bixia1
authored and
tensorflower-gardener
committed
Use a regular dictionary instead of an object-identity dictionary when printing node names (object-identity was left over from when regular objects were keys)

allenlavoie
authored and
tensorflower-gardener
committed
Fix ordering of module names.

Dan Moldovan
authored and
tensorflower-gardener
committed
Use op type as JSON event name in Profiler
tensorflower-gardener
committed
Merge pull request #30824 from tensorflow:lamberta-patch-1
tensorflower-gardener
committed
Automated rollback of commit 11b33344b3b398db10d370c100e915e7ad023546. Revert #30053.

pavithrasv
authored and
tensorflower-gardener
committed
Merge pull request #30611 from ROCmSoftwarePlatform:google_upstream_rocm_update_190711
tensorflower-gardener
committed
Merge pull request #30500 from ROCmSoftwarePlatform:google_upstream_skip_double_dtyp_subtests
tensorflower-gardener
committed
[XLA] Improve error message in HLO matchers.
tensorflower-gardener
committed
Fix all tests in `training_eager_test` for new codepath.

omalleyt12
authored and
tensorflower-gardener
committed
Simplify backedges tracking in the MLIR graph importer (NFC)

joker-eph
authored and
tensorflower-gardener
committed
Merge pull request #30724 from jbedorf:fix_30440
tensorflower-gardener
committed
[tf.data] Fix forward compatibility issue in rebatch.cc

Rachel Lim
authored and
tensorflower-gardener
committed

Commits on Aug 15, 2019

Temporary change: always link to MKL binary blob
Roman Dubtsov
committed
Disable inter_op manipulation with MKLDNN_TF_THREADING
Roman Dubtsov
committed
PoC for using MKL-DNN with TensorFlow thread pool
Roman Dubtsov
committed