diff options
94 files changed, 890 insertions, 437 deletions
diff --git a/ISSUE_TEMPLATE.md b/ISSUE_TEMPLATE.md index 2bf2c754cf..1a401997c6 100644 --- a/ISSUE_TEMPLATE.md +++ b/ISSUE_TEMPLATE.md @@ -19,6 +19,7 @@ If you open a GitHub issue, here is our policy: - **TensorFlow version (use command below)**: - **Python version**: - **Bazel version (if compiling from source)**: +- **GCC/Compiler version (if compiling from source)**: - **CUDA/cuDNN version**: - **GPU model and memory**: - **Exact command to reproduce**: diff --git a/RELEASE.md b/RELEASE.md index 4a33bce8b2..d8db1f7200 100644 --- a/RELEASE.md +++ b/RELEASE.md @@ -19,6 +19,14 @@ (with GPU and gradient support). * Add a self-check on `import tensorflow` for Windows DLL issues. * Add NCHW support to `tf.depth_to_space` on GPU. +* TensorFlow Debugger (tfdbg): + * Add `eval` command to allow evaluation of arbitrary Python/numpy expressions + in tfdbg command-line interface. See + [Debugging TensorFlow Programs](https://www.tensorflow.org/programmers_guide/debugger) + for more details. + * Usability improvement: The frequently used tensor filter `has_inf_or_nan` is + now added to `Session` wrappers and hooks by default. So there is no need + for clients to call `.add_tensor_filter(tf_debug.has_inf_or_nan)` anymore. * SinhArcsinh (scalar) distribution added to `contrib.distributions`. * Make `GANEstimator` opensource. * `Estimator.export_savedmodel()` now includes all valid serving signatures @@ -60,10 +68,14 @@ * Fix `tf.contrib.distributions.Affine` incorrectly computing log-det-jacobian. * Fix `tf.random_gamma` incorrectly handling non-batch, scalar draws. * Resolved a race condition in TensorForest TreePredictionsV4Op. -* Google Cloud Storage file system and Hadoop file system support are now - default build options. +* Google Cloud Storage file system, Amazon S3 file system, and Hadoop file + system support are now default build options. * Custom op libraries must link against libtensorflow_framework.so (installed at `tf.sysconfig.get_lib()`). +* Change `RunConfig` default behavior to not set a random seed, making random + behavior independently random on distributed workers. We expect this to + generally improve training performance. Models that do rely on determinism + should set a random seed explicitly. ## Breaking Changes to the API * The signature of the `tf.contrib.data.rejection_resample()` function has been @@ -74,6 +86,11 @@ * Remove seldom used and unnecessary `tf.contrib.data.Iterator.dispose_op()`. * Reorder some TFGAN loss functions in a non-backwards compatible way. +## Known Issues +* In Python 3, `Dataset.from_generator()` does not support Unicode strings. + You must convert any strings to bytes objects before yielding them from + the generator. + ## Thanks to our Contributors This release contains contributions from many people at Google, as well as: diff --git a/configure.py b/configure.py index 425eae676c..bc7859fee4 100644 --- a/configure.py +++ b/configure.py @@ -994,6 +994,7 @@ def main(): environ_cp['TF_NEED_HDFS'] = '0' environ_cp['TF_NEED_JEMALLOC'] = '0' environ_cp['TF_NEED_OPENCL'] = '0' + environ_cp['TF_NEED_S3'] = '0' environ_cp['TF_CUDA_CLANG'] = '0' if is_macos(): diff --git a/tensorflow/compiler/aot/tfcompile.bzl b/tensorflow/compiler/aot/tfcompile.bzl index 2adb1dc65e..ee291c12d0 100644 --- a/tensorflow/compiler/aot/tfcompile.bzl +++ b/tensorflow/compiler/aot/tfcompile.bzl @@ -4,7 +4,7 @@ To use from your BUILD file, add the following line to load the macro: -load("//tensorflow/compiler/aot:tfcompile.bzl", "tf_library") +load("@org_tensorflow//tensorflow/compiler/aot:tfcompile.bzl", "tf_library") Then call the macro like this: @@ -16,14 +16,14 @@ tf_library( ) """ -load("//tensorflow:tensorflow.bzl", "if_android", "tf_copts") +load("@org_tensorflow//tensorflow:tensorflow.bzl", "if_android", "tf_copts") def tf_library(name, graph, config, freeze_checkpoint=None, freeze_saver=None, cpp_class=None, gen_test=True, gen_benchmark=True, visibility=None, testonly=None, tfcompile_flags=None, - tfcompile_tool="//tensorflow/compiler/aot:tfcompile", + tfcompile_tool="@org_tensorflow//tensorflow/compiler/aot:tfcompile", include_standard_runtime_deps=True, deps=None, tags=None): """Runs tfcompile to compile a TensorFlow graph into executable code. @@ -121,7 +121,7 @@ def tf_library(name, graph, config, outs=[freeze_file], cmd=("$(location //tensorflow/python/tools:freeze_graph)" + freeze_args), - tools=["//tensorflow/python/tools:freeze_graph"], + tools=["@org_tensorflow//tensorflow/python/tools:freeze_graph"], tags=tags, ) tfcompile_graph = freeze_file @@ -207,22 +207,22 @@ def tf_library(name, graph, config, # These deps are required by all tf_library targets even if # include_standard_runtime_deps is False. Without them, the # generated code will fail to compile. - "//tensorflow/compiler/tf2xla:xla_compiled_cpu_function", - "//tensorflow/core:framework_lite", + "@org_tensorflow//tensorflow/compiler/tf2xla:xla_compiled_cpu_function", + "@org_tensorflow//tensorflow/core:framework_lite", ] + (need_xla_data_proto and [ # If we're generating the program shape, we must depend on the proto. - "//tensorflow/compiler/xla:xla_data_proto", + "@org_tensorflow//tensorflow/compiler/xla:xla_data_proto", ] or []) + (include_standard_runtime_deps and [ # TODO(cwhipkey): only depend on kernel code that the model actually needed. - "//tensorflow/compiler/tf2xla/kernels:index_ops_kernel_argmax_float_1d", - "//tensorflow/compiler/tf2xla/kernels:index_ops_kernel_argmax_float_2d", - "//tensorflow/compiler/xla/service/cpu:cpu_runtime_avx", - "//tensorflow/compiler/xla/service/cpu:cpu_runtime_neon", - "//tensorflow/compiler/xla/service/cpu:cpu_runtime_sse4_1", - "//tensorflow/compiler/xla/service/cpu:runtime_conv2d", - "//tensorflow/compiler/xla/service/cpu:runtime_matmul", - "//tensorflow/compiler/xla/service/cpu:runtime_single_threaded_conv2d", - "//tensorflow/compiler/xla/service/cpu:runtime_single_threaded_matmul", + "@org_tensorflow//tensorflow/compiler/tf2xla/kernels:index_ops_kernel_argmax_float_1d", + "@org_tensorflow//tensorflow/compiler/tf2xla/kernels:index_ops_kernel_argmax_float_2d", + "@org_tensorflow//tensorflow/compiler/xla/service/cpu:cpu_runtime_avx", + "@org_tensorflow//tensorflow/compiler/xla/service/cpu:cpu_runtime_neon", + "@org_tensorflow//tensorflow/compiler/xla/service/cpu:cpu_runtime_sse4_1", + "@org_tensorflow//tensorflow/compiler/xla/service/cpu:runtime_conv2d", + "@org_tensorflow//tensorflow/compiler/xla/service/cpu:runtime_matmul", + "@org_tensorflow//tensorflow/compiler/xla/service/cpu:runtime_single_threaded_conv2d", + "@org_tensorflow//tensorflow/compiler/xla/service/cpu:runtime_single_threaded_matmul", "//third_party/eigen3", ] or []) + (deps or []), tags=tags, @@ -248,7 +248,7 @@ def tf_library(name, graph, config, name=("gen_" + test_name), testonly=1, srcs=[ - "//tensorflow/compiler/aot:test.cc", + "@org_tensorflow//tensorflow/compiler/aot:test.cc", header_file, ], outs=[test_file], @@ -264,13 +264,13 @@ def tf_library(name, graph, config, srcs=[test_file], deps=[ ":" + name, - "//tensorflow/compiler/tf2xla:xla_local_runtime_context", - "//tensorflow/compiler/aot:runtime", - "//tensorflow/compiler/aot:tf_library_test_main", - "//tensorflow/compiler/xla:executable_run_options", + "@org_tensorflow//tensorflow/compiler/tf2xla:xla_local_runtime_context", + "@org_tensorflow//tensorflow/compiler/aot:runtime", + "@org_tensorflow//tensorflow/compiler/aot:tf_library_test_main", + "@org_tensorflow//tensorflow/compiler/xla:executable_run_options", "//third_party/eigen3", - "//tensorflow/core:lib", - "//tensorflow/core:test", + "@org_tensorflow//tensorflow/core:lib", + "@org_tensorflow//tensorflow/core:test", ], tags=tags, ) @@ -278,7 +278,7 @@ def tf_library(name, graph, config, if gen_benchmark: benchmark_name = name + "_benchmark" benchmark_file = benchmark_name + ".cc" - benchmark_main = ("//tensorflow/compiler/aot:" + + benchmark_main = ("@org_tensorflow//tensorflow/compiler/aot:" + "benchmark_main.template") # Rule to rewrite benchmark.cc to produce the benchmark_file. @@ -310,13 +310,13 @@ def tf_library(name, graph, config, linkopts = if_android(["-pie", "-s"]), deps=[ ":" + name, - "//tensorflow/compiler/tf2xla:xla_local_runtime_context", - "//tensorflow/compiler/aot:benchmark", - "//tensorflow/compiler/aot:runtime", - "//tensorflow/compiler/xla:executable_run_options", + "@org_tensorflow//tensorflow/compiler/tf2xla:xla_local_runtime_context", + "@org_tensorflow//tensorflow/compiler/aot:benchmark", + "@org_tensorflow//tensorflow/compiler/aot:runtime", + "@org_tensorflow//tensorflow/compiler/xla:executable_run_options", "//third_party/eigen3", ] + if_android([ - "//tensorflow/compiler/aot:benchmark_extra_android", + "@org_tensorflow//tensorflow/compiler/aot:benchmark_extra_android", ]), tags=tags, ) @@ -326,11 +326,11 @@ def target_llvm_triple(): # TODO(toddw): Add target_triple for other targets. For details see: # http://llvm.org/docs/doxygen/html/Triple_8h_source.html return select({ - "//tensorflow:android_armeabi": "armv5-none-android", - "//tensorflow:android_arm": "armv7-none-android", - "//tensorflow:android_arm64": "aarch64-none-android", - "//tensorflow:android_x86": "i686-none-android", - "//tensorflow:linux_ppc64le": "ppc64le-ibm-linux-gnu", - "//tensorflow:darwin": "x86_64-none-darwin", + "@org_tensorflow//tensorflow:android_armeabi": "armv5-none-android", + "@org_tensorflow//tensorflow:android_arm": "armv7-none-android", + "@org_tensorflow//tensorflow:android_arm64": "aarch64-none-android", + "@org_tensorflow//tensorflow:android_x86": "i686-none-android", + "@org_tensorflow//tensorflow:linux_ppc64le": "ppc64le-ibm-linux-gnu", + "@org_tensorflow//tensorflow:darwin": "x86_64-none-darwin", "//conditions:default": "x86_64-pc-linux", }) diff --git a/tensorflow/contrib/android/java/org/tensorflow/contrib/android/TensorFlowInferenceInterface.java b/tensorflow/contrib/android/java/org/tensorflow/contrib/android/TensorFlowInferenceInterface.java index 80e03f2036..1f423a7a5b 100644 --- a/tensorflow/contrib/android/java/org/tensorflow/contrib/android/TensorFlowInferenceInterface.java +++ b/tensorflow/contrib/android/java/org/tensorflow/contrib/android/TensorFlowInferenceInterface.java @@ -288,6 +288,22 @@ public class TensorFlowInferenceInterface { * as many elements as that of the destination Tensor. If {@link src} has more elements than the * destination has capacity, the copy is truncated. */ + public void feed(String inputName, boolean[] src, long... dims) { + byte[] b = new byte[src.length]; + + for (int i = 0; i < src.length; i++) { + b[i] = src[i] ? (byte) 1 : (byte) 0; + } + + addFeed(inputName, Tensor.create(Boolean.class, dims, ByteBuffer.wrap(b))); + } + + /** + * Given a source array with shape {@link dims} and content {@link src}, copy the contents into + * the input Tensor with name {@link inputName}. The source array {@link src} must have at least + * as many elements as that of the destination Tensor. If {@link src} has more elements than the + * destination has capacity, the copy is truncated. + */ public void feed(String inputName, float[] src, long... dims) { addFeed(inputName, Tensor.create(dims, FloatBuffer.wrap(src))); } diff --git a/tensorflow/contrib/boosted_trees/python/training/functions/gbdt_batch.py b/tensorflow/contrib/boosted_trees/python/training/functions/gbdt_batch.py index f8f4b43a07..5a917ca428 100644 --- a/tensorflow/contrib/boosted_trees/python/training/functions/gbdt_batch.py +++ b/tensorflow/contrib/boosted_trees/python/training/functions/gbdt_batch.py @@ -344,7 +344,7 @@ class GradientBoostedDecisionTreeModel(object): learner_config.num_classes == 2) def _predict_and_return_dict(self, ensemble_handle, ensemble_stamp, mode): - """Runs prediciton and returns a dictionary of the prediction results. + """Runs prediction and returns a dictionary of the prediction results. Args: ensemble_handle: ensemble resource handle. diff --git a/tensorflow/contrib/cmake/tf_tests.cmake b/tensorflow/contrib/cmake/tf_tests.cmake index ac55b9ea92..77d2124914 100644 --- a/tensorflow/contrib/cmake/tf_tests.cmake +++ b/tensorflow/contrib/cmake/tf_tests.cmake @@ -253,7 +253,6 @@ if (tensorflow_BUILD_PYTHON_TESTS) "${tensorflow_source_dir}/tensorflow/python/training/evaluation_test.py" # training tests "${tensorflow_source_dir}/tensorflow/python/training/basic_session_run_hooks_test.py" # Needs tf.contrib fix. - "${tensorflow_source_dir}/tensorflow/python/training/localhost_cluster_performance_test.py" # Needs portpicker. "${tensorflow_source_dir}/tensorflow/python/training/quantize_training_test.py" # Needs quantization ops to be included in windows. "${tensorflow_source_dir}/tensorflow/python/training/supervisor_test.py" # Flaky I/O error on rename. "${tensorflow_source_dir}/tensorflow/python/training/server_lib_test.py" # Test occasionally deadlocks. diff --git a/tensorflow/contrib/eager/README.md b/tensorflow/contrib/eager/README.md index db11dbb0d7..ae4b07799f 100644 --- a/tensorflow/contrib/eager/README.md +++ b/tensorflow/contrib/eager/README.md @@ -35,7 +35,7 @@ print(m) This feature is in early stages and work remains to be done in terms of smooth support for distributed and multi-GPU training and CPU performance. -- [Known issues](https://github.com/tensorflow/tensorflow/issues?q=is%3Aissue%20is%3Aopen%20label%3Aproj%3Aeager) +- [Known issues](https://github.com/tensorflow/tensorflow/issues?q=is%3Aissue%20is%3Aopen%20label%3Acomp%3Aeager) - Feedback is welcome, please consider [filing an issue](https://github.com/tensorflow/tensorflow/issues/new) to provide it. diff --git a/tensorflow/contrib/eager/python/g3doc/guide.md b/tensorflow/contrib/eager/python/g3doc/guide.md index 4ec0ab8275..230fc893bf 100644 --- a/tensorflow/contrib/eager/python/g3doc/guide.md +++ b/tensorflow/contrib/eager/python/g3doc/guide.md @@ -68,9 +68,9 @@ enabled. A significant fraction of the [TensorFlow API](https://www.tensorflow.org/api_docs/python/) consists of numerical operations: -[arithmetic operations](https://www.tensorflow.org/api_docs/python/tf/matmul), -[matrix operations](https://www.tensorflow.org/api_docs/python/tf/matmul), -[linear algebra operations](https://www.tensorflow.org/api_docs/python/tf/linalg), +[arithmetic operations](https://www.tensorflow.org/api_guides/python/math_ops#Arithmetic_Operators), +[matrix operations](https://www.tensorflow.org/api_guides/python/math_ops#Matrix_Math_Functions), +[linear algebra operations](https://www.tensorflow.org/versions/master/api_docs/python/tf/linalg), etc. With eager execution enabled, these operations consume and return @@ -746,7 +746,7 @@ during graph construction. `tf.summary` operations are *not* compatible with eager execution, but an equivalent alternative exists in -[`tf.contrib.summary`](https://www.tensorflow.org/versions/master/api_guides/python/tf/contrib/summary/) +[`tf.contrib.summary`](https://www.tensorflow.org/versions/master/api_docs/python/tf/contrib/summary) that is compatible with both eager execution and graph construction. During model construction simply insert summary operations like @@ -887,7 +887,7 @@ Some differences worth noting: Please give eager execution a spin. This feature is in early stages and is evolving, so we welcome your feedback via issues on GitHub (see [known -issues](https://github.com/tensorflow/tensorflow/labels/eager)). +issues](https://github.com/tensorflow/tensorflow/labels/comp:eager)). You may want to browse through some sample code, including benchmarks for some: diff --git a/tensorflow/contrib/framework/python/framework/tensor_util.py b/tensorflow/contrib/framework/python/framework/tensor_util.py index 92a2a4ff2d..4e6eea8884 100644 --- a/tensorflow/contrib/framework/python/framework/tensor_util.py +++ b/tensorflow/contrib/framework/python/framework/tensor_util.py @@ -77,10 +77,10 @@ def reduce_sum_n(tensors, name=None): return tensors[0] return math_ops.add_n(tensors, name=name_scope) -@deprecated(None, - 'Please switch to tf.confusion_matrix.remove_squeezable_dimensions.' - 'Note that order of the inputs and outputs of labels and ' - 'predictions have also been switched.') +@deprecated( + None, "Please switch to remove_squeezable_dimensions from " + "tf.confusion_matrix. Note that the order of the inputs and outputs of " + "labels and predictions have also been switched.") def remove_squeezable_dimensions(predictions, labels, name=None): """Squeeze last dim if ranks of `predictions` and `labels` differ by 1. diff --git a/tensorflow/contrib/gan/README.md b/tensorflow/contrib/gan/README.md index 5d74df3ef7..3ab8478070 100644 --- a/tensorflow/contrib/gan/README.md +++ b/tensorflow/contrib/gan/README.md @@ -47,7 +47,7 @@ such as the Wasserstein loss, gradient penalty, mutual information penalty, etc * [evaluation](https://www.tensorflow.org/code/tensorflow/contrib/gan/python/eval/python/): Use `Inception Score` or `Frechet Distance` with a pretrained Inception -network to evaluate your unconditional generative model. You can also also use +network to evaluate your unconditional generative model. You can also use your own pretrained classifier for more specific performance numbers, or use other methods for evaluating conditional generative models. diff --git a/tensorflow/contrib/layers/python/layers/layers.py b/tensorflow/contrib/layers/python/layers/layers.py index c429d53cdc..78c1839e51 100644 --- a/tensorflow/contrib/layers/python/layers/layers.py +++ b/tensorflow/contrib/layers/python/layers/layers.py @@ -2008,7 +2008,7 @@ def layer_norm(inputs, Given a tensor `inputs` of rank `R`, moments are calculated and normalization is performed over axes `begin_norm_axis ... R - 1`. Scaling and centering, - if requested, is performed over axes `begin_shift_axis .. R - 1`. + if requested, is performed over axes `begin_params_axis .. R - 1`. By default, `begin_norm_axis = 1` and `begin_params_axis = -1`, meaning that normalization is performed over all but the first axis diff --git a/tensorflow/contrib/makefile/Dockerfile b/tensorflow/contrib/makefile/Dockerfile index 341f22e692..64d571a4ed 100644 --- a/tensorflow/contrib/makefile/Dockerfile +++ b/tensorflow/contrib/makefile/Dockerfile @@ -1,6 +1,6 @@ FROM ubuntu:16.04 -MAINTAINER Gunhan Gulsoy <gunan@google.com> +LABEL maintainer="Gunhan Gulsoy <gunan@google.com>" # Install make build dependencies for TensorFlow. RUN apt-get update diff --git a/tensorflow/contrib/makefile/Makefile b/tensorflow/contrib/makefile/Makefile index b582493131..dba1464653 100644 --- a/tensorflow/contrib/makefile/Makefile +++ b/tensorflow/contrib/makefile/Makefile @@ -11,6 +11,8 @@ # the first for the host (the machine you're compiling on) and the second for # the target (the machine you want the program to run on). +SHELL := /bin/bash + # Host compilation settings # Find where we're running from, so we can store generated files here. @@ -44,6 +46,11 @@ ifdef HEXAGON_LIBS endif endif # HEXAGON_LIBS +# If ANDROID_TYPES is not set assume __ANDROID_TYPES_SLIM__ +ifeq ($(ANDROID_TYPES),) + ANDROID_TYPES := -D__ANDROID_TYPES_SLIM__ +endif + # Try to figure out the host system HOST_OS := ifeq ($(OS),Windows_NT) @@ -58,6 +65,8 @@ else endif endif +HOST_ARCH := $(shell if [[ $(shell uname -m) =~ i[345678]86 ]]; then echo x86_32; else echo $(shell uname -m); fi) + # Where compiled objects are stored. HOST_OBJDIR := $(MAKEFILE_DIR)/gen/host_obj/ HOST_BINDIR := $(MAKEFILE_DIR)/gen/host_bin/ @@ -216,7 +225,7 @@ ifeq ($(TARGET),LINUX) endif # If we're cross-compiling for the Raspberry Pi, use the right gcc. ifeq ($(TARGET),PI) - CXXFLAGS += -D__ANDROID_TYPES_SLIM__ -DRASPBERRY_PI + CXXFLAGS += $(ANDROID_TYPES) -DRASPBERRY_PI LDFLAGS := -Wl,--no-whole-archive LIBS += -ldl -lpthread LIBFLAGS += -Wl,--allow-multiple-definition -Wl,--whole-archive @@ -230,43 +239,93 @@ ifeq ($(TARGET),ANDROID) # NDK_ROOT=/path/to/your/ndk # You need to have an Android version of the protobuf libraries compiled to link # in. The compile_android_protobuf.sh script may help. -# TODO(satok): Support all CPU architectures (Currently only armv7 is supported) - OS_PATH := + ANDROID_HOST_OS_ARCH := ifeq ($(HOST_OS),LINUX) - OS_PATH=linux + ANDROID_HOST_OS_ARCH=linux endif ifeq ($(HOST_OS),OSX) - OS_PATH=darwin + ANDROID_HOST_OS_ARCH=darwin endif ifeq ($(HOST_OS),WINDOWS) $(error "windows is not supported.") endif + ifeq ($(HOST_ARCH),x86_32) + ANDROID_HOST_OS_ARCH := $(ANDROID_HOST_OS_ARCH)-x86 + else + ANDROID_HOST_OS_ARCH := $(ANDROID_HOST_OS_ARCH)-$(HOST_ARCH) + endif + + ifndef ANDROID_ARCH + ANDROID_ARCH := armeabi-v7a + endif + + ifeq ($(ANDROID_ARCH),arm64-v8a) + TOOLCHAIN := aarch64-linux-android-4.9 + SYSROOT_ARCH := arm64 + BIN_PREFIX := aarch64-linux-android + MARCH_OPTION := + endif + ifeq ($(ANDROID_ARCH),armeabi) + TOOLCHAIN := arm-linux-androideabi-4.9 + SYSROOT_ARCH := arm + BIN_PREFIX := arm-linux-androideabi + MARCH_OPTION := + endif + ifeq ($(ANDROID_ARCH),armeabi-v7a) + TOOLCHAIN := arm-linux-androideabi-4.9 + SYSROOT_ARCH := arm + BIN_PREFIX := arm-linux-androideabi + MARCH_OPTION := -march=armv7-a -mfloat-abi=softfp -mfpu=neon + endif + ifeq ($(ANDROID_ARCH),mips) + TOOLCHAIN := mipsel-linux-android-4.9 + SYSROOT_ARCH := mips + BIN_PREFIX := mipsel-linux-android + MARCH_OPTION := + endif + ifeq ($(ANDROID_ARCH),mips64) + TOOLCHAIN := mips64el-linux-android-4.9 + SYSROOT_ARCH := mips64 + BIN_PREFIX := mips64el-linux-android + MARCH_OPTION := + endif + ifeq ($(ANDROID_ARCH),x86) + TOOLCHAIN := x86-4.9 + SYSROOT_ARCH := x86 + BIN_PREFIX := i686-linux-android + MARCH_OPTION := + endif + ifeq ($(ANDROID_ARCH),x86_64) + TOOLCHAIN := x86_64-4.9 + SYSROOT_ARCH := x86_64 + BIN_PREFIX := x86-64-linux-android + MARCH_OPTION := + endif + ifndef NDK_ROOT $(error "NDK_ROOT is not defined.") endif - CXX := $(CC_PREFIX) $(NDK_ROOT)/toolchains/arm-linux-androideabi-4.9/prebuilt/$(OS_PATH)-x86_64/bin/arm-linux-androideabi-g++ - CC := $(CC_PREFIX) $(NDK_ROOT)/toolchains/arm-linux-androideabi-4.9/prebuilt/$(OS_PATH)-x86_64/bin/arm-linux-androideabi-gcc + CXX := $(CC_PREFIX) $(NDK_ROOT)/toolchains/$(TOOLCHAIN)/prebuilt/$(ANDROID_HOST_OS_ARCH)/bin/$(BIN_PREFIX)-g++ + CC := $(CC_PREFIX) $(NDK_ROOT)/toolchains/$(TOOLCHAIN)/prebuilt/$(ANDROID_HOST_OS_ARCH)/bin/$(BIN_PREFIX)-gcc CXXFLAGS +=\ ---sysroot $(NDK_ROOT)/platforms/android-21/arch-arm \ +--sysroot $(NDK_ROOT)/platforms/android-21/arch-$(SYSROOT_ARCH) \ -Wno-narrowing \ -fomit-frame-pointer \ --march=armv7-a \ --mfloat-abi=softfp \ --mfpu=neon \ +$(MARCH_OPTION) \ -fPIE INCLUDES = \ -I$(NDK_ROOT)/sources/android/support/include \ -I$(NDK_ROOT)/sources/cxx-stl/gnu-libstdc++/4.9/include \ --I$(NDK_ROOT)/sources/cxx-stl/gnu-libstdc++/4.9/libs/armeabi/include \ +-I$(NDK_ROOT)/sources/cxx-stl/gnu-libstdc++/4.9/libs/$(ANDROID_ARCH)/include \ -I. \ -I$(MAKEFILE_DIR)/downloads/ \ -I$(MAKEFILE_DIR)/downloads/eigen \ -I$(MAKEFILE_DIR)/downloads/gemmlowp \ -I$(MAKEFILE_DIR)/downloads/nsync/public \ -I$(MAKEFILE_DIR)/downloads/fft2d \ --I$(MAKEFILE_DIR)/gen/protobuf/include \ +-I$(MAKEFILE_DIR)/gen/protobuf_android/$(ANDROID_ARCH)/include \ -I$(PROTOGENDIR) \ -I$(PBTGENDIR) @@ -277,19 +336,20 @@ $(TARGET_NSYNC_LIB) \ -llog \ -lz \ -lm \ --ldl +-ldl \ +-latomic - LD := $(NDK_ROOT)/toolchains/arm-linux-androideabi-4.9/prebuilt/$(OS_PATH)-x86_64/arm-linux-androideabi/bin/ld + LD := $(NDK_ROOT)/toolchains/$(TOOLCHAIN)/prebuilt/$(ANDROID_HOST_OS_ARCH)/$(BIN_PREFIX)/bin/ld LDFLAGS := \ --march=armv7-a \ --L$(MAKEFILE_DIR)/gen/protobuf/lib \ --L$(NDK_ROOT)/sources/cxx-stl/gnu-libstdc++/4.9/libs/armeabi-v7a \ +$(MARCH_OPTION) \ +-L$(MAKEFILE_DIR)/gen/protobuf_android/$(ANDROID_ARCH)/lib \ +-L$(NDK_ROOT)/sources/cxx-stl/gnu-libstdc++/4.9/libs/$(ANDROID_ARCH) \ -fPIE \ -pie \ -v - AR := $(NDK_ROOT)/toolchains/arm-linux-androideabi-4.9/prebuilt/$(OS_PATH)-x86_64/bin/arm-linux-androideabi-ar + AR := $(NDK_ROOT)/toolchains/$(TOOLCHAIN)/prebuilt/$(ANDROID_HOST_OS_ARCH)/bin/$(BIN_PREFIX)-ar ARFLAGS := r LIBFLAGS += -Wl,--allow-multiple-definition -Wl,--whole-archive @@ -313,6 +373,11 @@ $(TARGET_NSYNC_LIB) \ ifdef ENABLE_EXPERIMENTAL_HEXNN_OPS CXXFLAGS += -DENABLE_EXPERIMENTAL_HEXNN_OPS endif + + OBJDIR := $(OBJDIR)android_$(ANDROID_ARCH)/ + LIBDIR := $(LIBDIR)android_$(ANDROID_ARCH)/ + BINDIR := $(BINDIR)android_$(ANDROID_ARCH)/ + DEPDIR := $(DEPDIR)android_$(ANDROID_ARCH)/ endif # ANDROID # LINT.ThenChange(//tensorflow/contrib/android/cmake/CMakeLists.txt) @@ -338,7 +403,7 @@ ifeq ($(TARGET),IOS) -Wno-c++11-narrowing \ -mno-thumb \ -DTF_LEAN_BINARY \ - -D__ANDROID_TYPES_SLIM__ \ + $(ANDROID_TYPES) \ -fno-exceptions \ -isysroot \ ${IPHONEOS_SYSROOT} @@ -362,7 +427,7 @@ ifeq ($(TARGET),IOS) -Wno-c++11-narrowing \ -mno-thumb \ -DTF_LEAN_BINARY \ - -D__ANDROID_TYPES_SLIM__ \ + $(ANDROID_TYPES) \ -fno-exceptions \ -isysroot \ ${IPHONEOS_SYSROOT} @@ -385,7 +450,7 @@ ifeq ($(TARGET),IOS) -DUSE_GEMM_FOR_CONV \ -Wno-c++11-narrowing \ -DTF_LEAN_BINARY \ - -D__ANDROID_TYPES_SLIM__ \ + $(ANDROID_TYPES) \ -fno-exceptions \ -isysroot \ ${IPHONEOS_SYSROOT} @@ -409,7 +474,7 @@ ifeq ($(TARGET),IOS) -DUSE_GEMM_FOR_CONV \ -Wno-c++11-narrowing \ -DTF_LEAN_BINARY \ - -D__ANDROID_TYPES_SLIM__ \ + $(ANDROID_TYPES) \ -fno-exceptions \ -isysroot \ ${IPHONESIMULATOR_SYSROOT} @@ -432,7 +497,7 @@ ifeq ($(TARGET),IOS) -DUSE_GEMM_FOR_CONV \ -Wno-c++11-narrowing \ -DTF_LEAN_BINARY \ - -D__ANDROID_TYPES_SLIM__ \ + $(ANDROID_TYPES) \ -fno-exceptions \ -isysroot \ ${IPHONESIMULATOR_SYSROOT} @@ -655,12 +720,12 @@ clean: # Gets rid of all generated files except protobuf libs generated # before calling make. This allows users not to recompile proto libs everytime. clean_except_protobuf_libs: - find $(MAKEFILE_DIR)/gen -mindepth 1 -maxdepth 1 ! -name "protobuf" ! -name "protobuf-host" -exec rm -r "{}" \; + find $(MAKEFILE_DIR)/gen -mindepth 1 -maxdepth 1 ! -name "protobuf*" -exec rm -r "{}" \; rm -rf tensorflow/core/util/version_info.cc # Gets rid of target files only, leaving the host alone. Also leaves the lib # directory untouched deliberately, so we can persist multiple architectures -# across builds for iOS. +# across builds for iOS and Android. cleantarget: rm -rf $(OBJDIR) rm -rf $(BINDIR) diff --git a/tensorflow/contrib/makefile/build_all_android.sh b/tensorflow/contrib/makefile/build_all_android.sh index 9944f71950..81cb17a311 100755 --- a/tensorflow/contrib/makefile/build_all_android.sh +++ b/tensorflow/contrib/makefile/build_all_android.sh @@ -18,12 +18,15 @@ set -e usage() { - echo "Usage: NDK_ROOT=<path to ndk root> $(basename "$0") [-s:t:Tx:X]" + echo "Usage: NDK_ROOT=<path to ndk root> $(basename "$0") [-Es:t:Tx:a:X]" echo "-E enable experimental hexnn ops" echo "-s [sub_makefiles] sub makefiles separated by white space" echo "-t [build_target] build target for Android makefile [default=all]" echo "-T only build tensorflow" echo "-x [hexagon library path] copy and hexagon libraries in the specified path" + echo "-a [architecture] Architecture of target android [default=armeabi-v7a] \ +(supported architecture list: \ +arm64-v8a armeabi armeabi-v7a mips mips64 x86 x86_64)" exit 1 } @@ -32,13 +35,16 @@ if [[ -z "${NDK_ROOT}" ]]; then exit 1 fi -while getopts "Es:t:Tx:" opt_name; do +ARCH=armeabi-v7a + +while getopts "Es:t:Tx:a:" opt_name; do case "$opt_name" in E) ENABLE_EXPERIMENTAL_HEXNN_OPS="true";; s) SUB_MAKEFILES="${OPTARG}";; t) BUILD_TARGET="${OPTARG}";; T) ONLY_MAKE_TENSORFLOW="true";; x) HEXAGON_LIB_PATH="${OPTARG}";; + a) ARCH="${OPTARG}";; *) usage;; esac done @@ -53,25 +59,23 @@ JOB_COUNT="${JOB_COUNT:-$(get_job_count)}" HEXAGON_DOWNLOAD_PATH="tensorflow/contrib/makefile/downloads/hexagon" +# Remove any old files first. +make -f tensorflow/contrib/makefile/Makefile cleantarget + if [[ "${ONLY_MAKE_TENSORFLOW}" != "true" ]]; then - # Remove any old files first. - make -f tensorflow/contrib/makefile/Makefile clean rm -rf tensorflow/contrib/makefile/downloads # Pull down the required versions of the frameworks we need. tensorflow/contrib/makefile/download_dependencies.sh # Compile protobuf for the target Android device architectures. CC_PREFIX="${CC_PREFIX}" NDK_ROOT="${NDK_ROOT}" \ -tensorflow/contrib/makefile/compile_android_protobuf.sh -c -else - # Only clean files generated by make - make -f tensorflow/contrib/makefile/Makefile clean_except_protobuf_libs +tensorflow/contrib/makefile/compile_android_protobuf.sh -c -a ${ARCH} fi # Compile nsync for the host and the target Android device architecture. # Don't use export var=`something` syntax; it swallows the exit status. HOST_NSYNC_LIB=`tensorflow/contrib/makefile/compile_nsync.sh` TARGET_NSYNC_LIB=`CC_PREFIX="${CC_PREFIX}" NDK_ROOT="${NDK_ROOT}" \ - tensorflow/contrib/makefile/compile_nsync.sh -t android -a armeabi-v7a` + tensorflow/contrib/makefile/compile_nsync.sh -t android -a ${ARCH}` export HOST_NSYNC_LIB TARGET_NSYNC_LIB if [[ ! -z "${HEXAGON_LIB_PATH}" ]]; then @@ -98,7 +102,8 @@ fi if [[ -z "${BUILD_TARGET}" ]]; then make -j"${JOB_COUNT}" -f tensorflow/contrib/makefile/Makefile \ - TARGET=ANDROID NDK_ROOT="${NDK_ROOT}" CC_PREFIX="${CC_PREFIX}" \ + TARGET=ANDROID NDK_ROOT="${NDK_ROOT}" ANDROID_ARCH="${ARCH}" \ + CC_PREFIX="${CC_PREFIX}" \ HOST_NSYNC_LIB="$HOST_NSYNC_LIB" TARGET_NSYNC_LIB="$TARGET_NSYNC_LIB" \ HEXAGON_LIBS="${HEXAGON_LIBS}" HEXAGON_INCLUDE="${HEXAGON_INCLUDE}" \ SUB_MAKEFILES="${SUB_MAKEFILES}" ${EXTRA_MAKE_ARGS[@]} @@ -106,7 +111,8 @@ else # BUILD_TARGET explicitly uncommented to allow multiple targets to be # passed to make in a single build_all_android.sh invocation. make -j"${JOB_COUNT}" -f tensorflow/contrib/makefile/Makefile \ - TARGET=ANDROID NDK_ROOT="${NDK_ROOT}" CC_PREFIX="${CC_PREFIX}" \ + TARGET=ANDROID NDK_ROOT="${NDK_ROOT}" ANDROID_ARCH="${ARCH}" \ + CC_PREFIX="${CC_PREFIX}" \ HOST_NSYNC_LIB="$HOST_NSYNC_LIB" TARGET_NSYNC_LIB="$TARGET_NSYNC_LIB" \ HEXAGON_LIBS="${HEXAGON_LIBS}" HEXAGON_INCLUDE="${HEXAGON_INCLUDE}" \ SUB_MAKEFILES="${SUB_MAKEFILES}" ${EXTRA_MAKE_ARGS[@]} ${BUILD_TARGET} diff --git a/tensorflow/contrib/makefile/compile_android_protobuf.sh b/tensorflow/contrib/makefile/compile_android_protobuf.sh index fadbe271b8..4355e3e597 100755 --- a/tensorflow/contrib/makefile/compile_android_protobuf.sh +++ b/tensorflow/contrib/makefile/compile_android_protobuf.sh @@ -71,10 +71,10 @@ then exit 1 fi -GENDIR="$(pwd)/gen/protobuf" +GENDIR="$(pwd)/gen/protobuf_android" HOST_GENDIR="$(pwd)/gen/protobuf-host" mkdir -p "${GENDIR}" -mkdir -p "${HOST_GENDIR}" +mkdir -p "${GENDIR}/${ARCHITECTURE}" if [[ ! -f "./downloads/protobuf/autogen.sh" ]]; then echo "You need to download dependencies before running this script." 1>&2 @@ -153,7 +153,7 @@ then exit 1 fi -./configure --prefix="${GENDIR}" \ +./configure --prefix="${GENDIR}/${ARCHITECTURE}" \ --host="${bin_prefix}" \ --with-sysroot="${SYSROOT}" \ --disable-shared \ diff --git a/tensorflow/contrib/metrics/python/ops/metric_ops.py b/tensorflow/contrib/metrics/python/ops/metric_ops.py index ca4dcef8de..33377a70c2 100644 --- a/tensorflow/contrib/metrics/python/ops/metric_ops.py +++ b/tensorflow/contrib/metrics/python/ops/metric_ops.py @@ -423,7 +423,8 @@ def streaming_mean_tensor(values, updates_collections=updates_collections, name=name) - +@deprecated(None, "Please switch to tf.metrics.accuracy. Note that the order " + "of the inputs of labels and predictions have been switched.") def streaming_accuracy(predictions, labels, weights=None, @@ -1101,7 +1102,8 @@ def streaming_curve_points(labels=None, return points, update_op - +@deprecated(None, "Please switch to tf.metrics.auc. Note that the order of " + "the inputs of labels and predictions have been switched.") def streaming_auc(predictions, labels, weights=None, @@ -1486,7 +1488,9 @@ def streaming_sensitivity_at_specificity(predictions, updates_collections=updates_collections, name=name) - +@deprecated( + None, "Please switch to tf.metrics.precision_at_thresholds. Note that the " + "order of of the inputs of labels and predictions have been switched.") def streaming_precision_at_thresholds(predictions, labels, thresholds, @@ -1545,7 +1549,9 @@ def streaming_precision_at_thresholds(predictions, updates_collections=updates_collections, name=name) - +@deprecated( + None, "Please switch to tf.metrics.recall_at_thresholds. Note that the " + "order of of the inputs of labels and predictions have been switched.") def streaming_recall_at_thresholds(predictions, labels, thresholds, @@ -1755,8 +1761,8 @@ def _at_k_name(name, k=None, class_id=None): return name -@deprecated('2016-11-08', 'Please use `streaming_sparse_recall_at_k`, ' - 'and reshape labels from [batch_size] to [batch_size, 1].') +@deprecated("2016-11-08", "Please use `streaming_sparse_recall_at_k`, " + "and reshape labels from [batch_size] to [batch_size, 1].") def streaming_recall_at_k(predictions, labels, k, @@ -2389,7 +2395,7 @@ def streaming_sparse_average_precision_at_top_k(top_k_predictions, updates_collections=updates_collections, name=name) - +@deprecated(None, "Please switch to tf.metrics.mean.") def streaming_mean_absolute_error(predictions, labels, weights=None, diff --git a/tensorflow/core/framework/node_def.proto b/tensorflow/core/framework/node_def.proto index 1fd2e50b51..8fcee32e29 100644 --- a/tensorflow/core/framework/node_def.proto +++ b/tensorflow/core/framework/node_def.proto @@ -35,7 +35,7 @@ message NodeDef { // CONSTRAINT ::= ("job:" JOB_NAME) // | ("replica:" [1-9][0-9]*) // | ("task:" [1-9][0-9]*) - // | ("device:" ("gpu" | "cpu") ":" ([1-9][0-9]* | "*") ) + // | ("device:" [A-Za-z]* ":" ([1-9][0-9]* | "*") ) // // Valid values for this string include: // * "/job:worker/replica:0/task:1/device:GPU:3" (full specification) diff --git a/tensorflow/core/kernels/BUILD b/tensorflow/core/kernels/BUILD index 2aef1e3560..1cb7c97be4 100644 --- a/tensorflow/core/kernels/BUILD +++ b/tensorflow/core/kernels/BUILD @@ -1098,7 +1098,7 @@ tf_kernel_library( visibility = [":friends"], deps = [ ":bounds_check", - "//tensorflow/core:framework_lite", + "//tensorflow/core:framework", "//third_party/eigen3", ], ) diff --git a/tensorflow/core/kernels/decode_csv_op.cc b/tensorflow/core/kernels/decode_csv_op.cc index 5e48ae9766..c4555db453 100644 --- a/tensorflow/core/kernels/decode_csv_op.cc +++ b/tensorflow/core/kernels/decode_csv_op.cc @@ -91,9 +91,9 @@ class DecodeCSVOp : public OpKernel { } else { int32 value; OP_REQUIRES(ctx, strings::safe_strto32(fields[f], &value), - errors::InvalidArgument( - "Field ", f, " in record ", i, - " is not a valid int32: ", fields[f])); + errors::InvalidArgument("Field ", f, " in record ", i, + " is not a valid int32: ", + fields[f])); output[f]->flat<int32>()(i) = value; } break; @@ -111,9 +111,9 @@ class DecodeCSVOp : public OpKernel { } else { int64 value; OP_REQUIRES(ctx, strings::safe_strto64(fields[f], &value), - errors::InvalidArgument( - "Field ", f, " in record ", i, - " is not a valid int64: ", fields[f])); + errors::InvalidArgument("Field ", f, " in record ", i, + " is not a valid int64: ", + fields[f])); output[f]->flat<int64>()(i) = value; } break; @@ -130,13 +130,33 @@ class DecodeCSVOp : public OpKernel { } else { float value; OP_REQUIRES(ctx, strings::safe_strtof(fields[f].c_str(), &value), - errors::InvalidArgument( - "Field ", f, " in record ", i, - " is not a valid float: ", fields[f])); + errors::InvalidArgument("Field ", f, " in record ", i, + " is not a valid float: ", + fields[f])); output[f]->flat<float>()(i) = value; } break; } + case DT_DOUBLE: { + // If this field is empty or NA value, check if default is given: + // If yes, use default value; Otherwise report error. + if (fields[f].empty() || fields[f] == na_value_) { + OP_REQUIRES(ctx, record_defaults[f].NumElements() == 1, + errors::InvalidArgument( + "Field ", f, + " is required but missing in record ", i, "!")); + output[f]->flat<double>()(i) = + record_defaults[f].flat<double>()(0); + } else { + double value; + OP_REQUIRES(ctx, strings::safe_strtod(fields[f].c_str(), &value), + errors::InvalidArgument("Field ", f, " in record ", i, + " is not a valid double: ", + fields[f])); + output[f]->flat<double>()(i) = value; + } + break; + } case DT_STRING: { // If this field is empty or NA value, check if default is given: // If yes, use default value; Otherwise report error. @@ -188,10 +208,9 @@ class DecodeCSVOp : public OpKernel { if (!quoted) { while (static_cast<size_t>(current_idx) < input.size() && input[current_idx] != delim_) { - OP_REQUIRES(ctx, - (!use_quote_delim_ || input[current_idx] != '"') && - input[current_idx] != '\n' && - input[current_idx] != '\r', + OP_REQUIRES(ctx, (!use_quote_delim_ || input[current_idx] != '"') && + input[current_idx] != '\n' && + input[current_idx] != '\r', errors::InvalidArgument( "Unquoted fields cannot have quotes/CRLFs inside")); field += input[current_idx]; @@ -219,11 +238,10 @@ class DecodeCSVOp : public OpKernel { } OP_REQUIRES( - ctx, - (static_cast<size_t>(current_idx) < input.size() && - input[current_idx] == '"' && - (static_cast<size_t>(current_idx) == input.size() - 1 || - input[current_idx + 1] == delim_)), + ctx, (static_cast<size_t>(current_idx) < input.size() && + input[current_idx] == '"' && + (static_cast<size_t>(current_idx) == input.size() - 1 || + input[current_idx + 1] == delim_)), errors::InvalidArgument("Quoted field has to end with quote " "followed by delim or end")); diff --git a/tensorflow/core/kernels/gather_functor.cc b/tensorflow/core/kernels/gather_functor.cc index 1b8be9b2ce..dde08b37ea 100644 --- a/tensorflow/core/kernels/gather_functor.cc +++ b/tensorflow/core/kernels/gather_functor.cc @@ -28,7 +28,7 @@ namespace functor { #define DECLARE_GPU_SPECS_INDEX(T, Index) \ template <> \ int64 GatherFunctor<GPUDevice, T, Index>::operator()( \ - const GPUDevice& d, typename TTypes<T, 3>::ConstTensor Tparams, \ + OpKernelContext* ctx, typename TTypes<T, 3>::ConstTensor Tparams, \ typename TTypes<Index>::ConstFlat Tindices, \ typename TTypes<T, 3>::Tensor Tout); \ extern template struct GatherFunctor<GPUDevice, T, Index>; diff --git a/tensorflow/core/kernels/gather_functor.h b/tensorflow/core/kernels/gather_functor.h index dfa1a5f1f9..1e429a037e 100644 --- a/tensorflow/core/kernels/gather_functor.h +++ b/tensorflow/core/kernels/gather_functor.h @@ -23,6 +23,8 @@ limitations under the License. #include "tensorflow/core/kernels/bounds_check.h" #include "tensorflow/core/platform/prefetch.h" #include "tensorflow/core/platform/types.h" +#include "tensorflow/core/framework/op_kernel.h" +#include "tensorflow/core/util/work_sharder.h" namespace tensorflow { typedef Eigen::ThreadPoolDevice CPUDevice; @@ -32,7 +34,8 @@ namespace functor { // Helper method to copy using memcpy. template <typename T, typename Index, typename SliceIndex, SliceIndex static_slice_elems> -SliceIndex HandleCopies(typename TTypes<T, 3>::ConstTensor params, +SliceIndex HandleCopies(OpKernelContext* ctx, + typename TTypes<T, 3>::ConstTensor params, typename TTypes<Index>::ConstFlat indices, SliceIndex slice_elems, typename TTypes<T, 3>::Tensor out) { @@ -47,44 +50,64 @@ SliceIndex HandleCopies(typename TTypes<T, 3>::ConstTensor params, } // Compute slice_bytes here so that static knowledge is available const size_t slice_bytes = slice_elems * sizeof(T); - for (SliceIndex b = 0; b < batch_size; b++) { - for (SliceIndex i = 0; i < indices_size; i++) { - const SliceIndex i_next = i + 1; - const SliceIndex b_next = b + 1; - if (i_next < indices_size) { - port::prefetch<port::PREFETCH_HINT_T0>(¶ms(b, indices(i_next), 0)); - port::prefetch<port::PREFETCH_HINT_T0>(&out(b, i_next, 0)); - } else if (b_next < batch_size) { + auto worker_threads = ctx->device()->tensorflow_cpu_worker_threads(); + mutex mu; + // Store the value of invalidate index for printing error information, it's a shared variable. + SliceIndex result = -1; + auto work = [&] (int64 start, int64 end) { + SliceIndex batch_idx = static_cast<SliceIndex>(start / indices_size); + SliceIndex indices_idx = static_cast<SliceIndex>(start % indices_size); + SliceIndex batch_idx_end = static_cast<SliceIndex>(end / indices_size); + SliceIndex indices_idx_end = static_cast<SliceIndex>(end % indices_size); + + while ((batch_idx < batch_idx_end) || + (batch_idx == batch_idx_end && indices_idx < indices_idx_end)) { + SliceIndex i_next = indices_idx + 1; + SliceIndex b_next = batch_idx + 1; + if ((batch_idx == batch_idx_end && i_next < indices_idx_end) || + (i_next < indices_size)) { + port::prefetch<port::PREFETCH_HINT_T0>(¶ms(batch_idx, indices(i_next), 0)); + port::prefetch<port::PREFETCH_HINT_T0>(&out(batch_idx, i_next, 0)); + b_next = batch_idx; + } else if (b_next <= batch_idx_end) { port::prefetch<port::PREFETCH_HINT_T0>(¶ms(b_next, indices(0), 0)); port::prefetch<port::PREFETCH_HINT_T0>(&out(b_next, 0, 0)); + i_next = 0; + } + const Index index = internal::SubtleMustCopy(indices(indices_idx)); + if (!FastBoundsCheck(index, limit)) { + mutex_lock l(mu); + result = indices_idx; + return; } - // Grab the index and check its validity. An earlier version of the - // code checked it and then grabbed it from memory a second time, which - // was a security risk since it could have changed in between. - const Index index = internal::SubtleMustCopy(indices(i)); - if (!FastBoundsCheck(index, limit)) return i; // Copy using memcpy if possible, otherwise an Eigen loop // TODO(cwhipkey): avoid linking to framework to get Allocator (to improve // ahead-of-time compilation binary size). if (is_simple_type<T>::value) { // Avoid auto-promotion to Index from SliceIndex by casting. - memcpy(out_base + (b * indices_size + i) * slice_elems, - params_base + (b * static_cast<SliceIndex>(limit) + + memcpy(out_base + (batch_idx * indices_size + indices_idx) * slice_elems, + params_base + (batch_idx * static_cast<SliceIndex>(limit) + static_cast<SliceIndex>(index)) * - slice_elems, + slice_elems, slice_bytes); } else { // For non-"simple" types (e.g. strings). - out.template chip<1>(i) = params.template chip<1>(index); + out.template chip<1>(indices_idx) = params.template chip<1>(index); } + indices_idx = i_next; + batch_idx = b_next; } - } - return -1; + }; + + Shard(worker_threads->num_threads, worker_threads->workers, batch_size*indices_size, + slice_elems * sizeof(T), work); + return result; } template <typename T, typename Index> struct GatherFunctorCPU { - int64 operator()(typename TTypes<T, 3>::ConstTensor params, + int64 operator()(OpKernelContext* ctx, + typename TTypes<T, 3>::ConstTensor params, typename TTypes<Index>::ConstFlat indices, typename TTypes<T, 3>::Tensor out) { const int64 N = indices.size(); @@ -94,16 +117,16 @@ struct GatherFunctorCPU { bool use_large = (slice_size > std::numeric_limits<int32>::max() || params.size() > std::numeric_limits<int32>::max() || N > std::numeric_limits<int32>::max()); -#define CALL(elems) \ - do { \ - if (use_large) { \ - bad_i = HandleCopies<T, Index, int64, elems>(params, indices, \ - slice_size, out); \ - } else { \ - const int32 small_slice = static_cast<int32>(slice_size); \ - bad_i = HandleCopies<T, Index, int32, elems>(params, indices, \ - small_slice, out); \ - } \ +#define CALL(elems) \ + do { \ + if (use_large) { \ + bad_i = HandleCopies<T, Index, int64, elems>(ctx, params, indices, \ + slice_size, out); \ + } else { \ + const int32 small_slice = static_cast<int32>(slice_size); \ + bad_i = HandleCopies<T, Index, int32, elems>(ctx, params, indices, \ + small_slice, out); \ + } \ } while (0) if (slice_size == 10) @@ -120,18 +143,18 @@ struct GatherFunctorCPU { template <typename Device, typename T, typename Index> struct GatherFunctor { - int64 operator()(const Device& d, typename TTypes<T, 3>::ConstTensor params, + int64 operator()(OpKernelContext* ctx, typename TTypes<T, 3>::ConstTensor params, typename TTypes<Index>::ConstFlat indices, typename TTypes<T, 3>::Tensor out); }; template <typename T, typename Index> struct GatherFunctor<CPUDevice, T, Index> { - int64 operator()(const CPUDevice& d, + int64 operator()(OpKernelContext* ctx, typename TTypes<T, 3>::ConstTensor params, typename TTypes<Index>::ConstFlat indices, typename TTypes<T, 3>::Tensor out) { - return GatherFunctorCPU<T, Index>()(params, indices, out); + return GatherFunctorCPU<T, Index>()(ctx, params, indices, out); } }; diff --git a/tensorflow/core/kernels/gather_functor_gpu.cu.h b/tensorflow/core/kernels/gather_functor_gpu.cu.h index e2384ef011..a50b51b54b 100644 --- a/tensorflow/core/kernels/gather_functor_gpu.cu.h +++ b/tensorflow/core/kernels/gather_functor_gpu.cu.h @@ -72,10 +72,11 @@ __global__ void GatherOpKernel(const T* params, const Index* indices, T* out, namespace functor { template <typename T, typename Index> struct GatherFunctor<GPUDevice, T, Index> { - int64 operator()(const GPUDevice& d, + int64 operator()(OpKernelContext* ctx, typename TTypes<T, 3>::ConstTensor params, typename TTypes<Index>::ConstFlat indices, typename TTypes<T, 3>::Tensor out) { + const GPUDevice& d = ctx->eigen_gpu_device(); const int64 out_size = out.size(); if (out_size == 0) { // We need a check here since the CPU version does useful error checking diff --git a/tensorflow/core/kernels/gather_op.cc b/tensorflow/core/kernels/gather_op.cc index 7088005e73..239d5d2e99 100644 --- a/tensorflow/core/kernels/gather_op.cc +++ b/tensorflow/core/kernels/gather_op.cc @@ -106,7 +106,7 @@ class GatherOp : public OpKernel { auto out_flat = out->shaped<T, 3>({outer_size, N, inner_size}); functor::GatherFunctor<Device, T, Index> functor; - int64 bad_i = functor(c->eigen_device<Device>(), params_flat, + int64 bad_i = functor(c, params_flat, indices_flat, out_flat); OP_REQUIRES( diff --git a/tensorflow/core/kernels/resource_variable_ops.cc b/tensorflow/core/kernels/resource_variable_ops.cc index 217fb3b781..0ae8a8fdbc 100644 --- a/tensorflow/core/kernels/resource_variable_ops.cc +++ b/tensorflow/core/kernels/resource_variable_ops.cc @@ -464,7 +464,7 @@ class ResourceGatherOp : public OpKernel { auto out_flat = out->shaped<T, 3>({1, N, out->NumElements() / N}); functor::GatherFunctor<Device, T, Index> functor; - int64 bad_i = functor(c->eigen_device<Device>(), params_flat, + int64 bad_i = functor(c, params_flat, indices_flat, out_flat); OP_REQUIRES( diff --git a/tensorflow/core/kernels/tile_ops.cc b/tensorflow/core/kernels/tile_ops.cc index 4c496a12c2..fa5afe6a31 100644 --- a/tensorflow/core/kernels/tile_ops.cc +++ b/tensorflow/core/kernels/tile_ops.cc @@ -248,7 +248,7 @@ TF_CALL_int64(HANDLE_TYPE_NAME_SYCL); #undef HANDLE_CASE // -------------------------------------------------------------------------- -template <typename Device> +template <typename Device, typename Tmultiples> class TileGradientOp : public OpKernel { public: explicit TileGradientOp(OpKernelConstruction* context) : OpKernel(context) {} @@ -273,10 +273,10 @@ class TileGradientOp : public OpKernel { return; } - const gtl::ArraySlice<int32> multiples_array(multiples.flat<int32>().data(), - input_dims); + const gtl::ArraySlice<Tmultiples> multiples_array( + multiples.flat<Tmultiples>().data(), input_dims); TensorShape output_shape; - std::vector<int32> input_dim_size_vec; + std::vector<Tmultiples> input_dim_size_vec; for (int i = 0; i < input_dims; ++i) { OP_REQUIRES( context, multiples_array[i] > 0, @@ -337,19 +337,19 @@ class TileGradientOp : public OpKernel { private: template <DataType DT, int NDIM> void HandleCase(OpKernelContext* context, - const std::vector<int32>& input_dims, - const gtl::ArraySlice<int32>& multiples_array, + const std::vector<Tmultiples>& input_dims, + const gtl::ArraySlice<Tmultiples>& multiples_array, Tensor* result); template <DataType DT, int NDIM> void HandleCaseImpl(OpKernelContext* context, - const std::vector<int32>& input_dims, - const gtl::ArraySlice<int32>& multiples_array, + const std::vector<Tmultiples>& input_dims, + const gtl::ArraySlice<Tmultiples>& multiples_array, Tensor* result) { typedef typename EnumToDataType<DT>::Type T; bool reduction_only = true; - std::vector<int> reduction_dims; + std::vector<Tmultiples> reduction_dims; for (int i = 0; i < NDIM; ++i) { if (input_dims[i] > multiples_array[i] && multiples_array[i] > 1) { @@ -411,7 +411,8 @@ class TileGradientOp : public OpKernel { template <typename T, int NDIM, int REDUCENDIM> void HandleReduce(OpKernelContext* context, - const std::vector<int32>& reduce_dim_in, Tensor* result) { + const std::vector<Tmultiples>& reduce_dim_in, + Tensor* result) { static_assert(NDIM >= REDUCENDIM, "Too many reduced dimensions"); Eigen::DSizes<Eigen::DenseIndex, REDUCENDIM> reduce_dim; Eigen::DSizes<Eigen::DenseIndex, NDIM> reshape_dim; @@ -432,34 +433,41 @@ class TileGradientOp : public OpKernel { TF_DISALLOW_COPY_AND_ASSIGN(TileGradientOp); }; -template <typename Device> +template <typename Device, typename Tmultiples> template <DataType DT, int NDIM> -inline void TileGradientOp<Device>::HandleCase( - OpKernelContext* context, const std::vector<int32>& input_dims, - const gtl::ArraySlice<int32>& multiples_array, Tensor* result) { +inline void TileGradientOp<Device, Tmultiples>::HandleCase( + OpKernelContext* context, const std::vector<Tmultiples>& input_dims, + const gtl::ArraySlice<Tmultiples>& multiples_array, Tensor* result) { LOG(FATAL) << "TileGradientOp: Invalid combination of Device, DT and NDIM: " << MakeTypeIndex<Device>().name() << ", " << DataTypeString(DT) << ", " << NDIM; } -#define HANDLE_CASE(device, T, dtype, ndim) \ +#define HANDLE_CASE(device, T, dtype, Tmultiples, ndim) \ template <> \ template <> \ - void TileGradientOp<device>::HandleCase<dtype, ndim>( \ - OpKernelContext * context, const std::vector<int32>& input_dims, \ - const gtl::ArraySlice<int32>& multiples_array, Tensor* result) { \ + void TileGradientOp<device, Tmultiples>::HandleCase<dtype, ndim>( \ + OpKernelContext * context, const std::vector<Tmultiples>& input_dims, \ + const gtl::ArraySlice<Tmultiples>& multiples_array, Tensor* result) { \ HandleCaseImpl<dtype, ndim>(context, input_dims, multiples_array, result); \ } // 0-D handled specially above -#define HANDLE_CASE_DIM(device, T, dtype) \ - HANDLE_CASE(device, T, dtype, 1); \ - HANDLE_CASE(device, T, dtype, 2); \ - HANDLE_CASE(device, T, dtype, 3); \ - HANDLE_CASE(device, T, dtype, 4); \ - HANDLE_CASE(device, T, dtype, 5); \ - HANDLE_CASE(device, T, dtype, 6); \ - HANDLE_CASE(device, T, dtype, 7); +#define HANDLE_CASE_DIM(device, T, dtype) \ + HANDLE_CASE(device, T, dtype, int32, 1); \ + HANDLE_CASE(device, T, dtype, int32, 2); \ + HANDLE_CASE(device, T, dtype, int32, 3); \ + HANDLE_CASE(device, T, dtype, int32, 4); \ + HANDLE_CASE(device, T, dtype, int32, 5); \ + HANDLE_CASE(device, T, dtype, int32, 6); \ + HANDLE_CASE(device, T, dtype, int32, 7); \ + HANDLE_CASE(device, T, dtype, int64, 1); \ + HANDLE_CASE(device, T, dtype, int64, 2); \ + HANDLE_CASE(device, T, dtype, int64, 3); \ + HANDLE_CASE(device, T, dtype, int64, 4); \ + HANDLE_CASE(device, T, dtype, int64, 5); \ + HANDLE_CASE(device, T, dtype, int64, 6); \ + HANDLE_CASE(device, T, dtype, int64, 7); #define HANDLE_TYPE_NAME_CPU(T) \ HANDLE_CASE_DIM(CPUDevice, T, DataTypeToEnum<T>::value); @@ -514,9 +522,16 @@ REGISTER_KERNEL_BUILDER(Name("Tile") .HostMemory("multiples") .TypeConstraint<int64>("Tmultiples"), TileOp<CPUDevice, int64>); -REGISTER_KERNEL_BUILDER( - Name("TileGrad").Device(DEVICE_CPU).HostMemory("multiples"), - TileGradientOp<CPUDevice>); +REGISTER_KERNEL_BUILDER(Name("TileGrad") + .Device(DEVICE_CPU) + .HostMemory("multiples") + .TypeConstraint<int32>("Tmultiples"), + TileGradientOp<CPUDevice, int32>); +REGISTER_KERNEL_BUILDER(Name("TileGrad") + .Device(DEVICE_CPU) + .HostMemory("multiples") + .TypeConstraint<int64>("Tmultiples"), + TileGradientOp<CPUDevice, int64>); #if GOOGLE_CUDA #define REGISTER_GPU(type) \ @@ -537,7 +552,13 @@ REGISTER_KERNEL_BUILDER( .TypeConstraint<type>("T") \ .TypeConstraint<int32>("Tmultiples") \ .HostMemory("multiples"), \ - TileGradientOp<GPUDevice>); + TileGradientOp<GPUDevice, int32>); \ + REGISTER_KERNEL_BUILDER(Name("TileGrad") \ + .Device(DEVICE_GPU) \ + .TypeConstraint<type>("T") \ + .TypeConstraint<int64>("Tmultiples") \ + .HostMemory("multiples"), \ + TileGradientOp<GPUDevice, int64>); TF_CALL_float(REGISTER_GPU); TF_CALL_double(REGISTER_GPU); @@ -569,7 +590,13 @@ TF_CALL_complex128(REGISTER_GPU) .TypeConstraint<type>("T") \ .TypeConstraint<int32>("Tmultiples") \ .HostMemory("multiples"), \ - TileGradientOp<SYCLDevice>); + TileGradientOp<SYCLDevice, int32>); \ + REGISTER_KERNEL_BUILDER(Name("TileGrad") \ + .Device(DEVICE_SYCL) \ + .TypeConstraint<type>("T") \ + .TypeConstraint<int64>("Tmultiples") \ + .HostMemory("multiples"), \ + TileGradientOp<SYCLDevice, int64>); TF_CALL_float(REGISTER_SYCL); TF_CALL_double(REGISTER_SYCL); diff --git a/tensorflow/core/ops/bitwise_ops.cc b/tensorflow/core/ops/bitwise_ops.cc index 3156162b78..2889953bdb 100644 --- a/tensorflow/core/ops/bitwise_ops.cc +++ b/tensorflow/core/ops/bitwise_ops.cc @@ -56,35 +56,45 @@ representation of that entry. 8- or 16-bit inputs and then aggregate the resulting counts. )doc"); -REGISTER_OP("BitwiseAnd").BINARY_BITWISE().Doc(R"doc( +REGISTER_OP("BitwiseAnd") + .BINARY_BITWISE() + .Doc(R"doc( Elementwise computes the bitwise AND of `x` and `y`. The result will have those bits set, that are set in both `x` and `y`. The computation is performed on the underlying representations of `x` and `y`. )doc"); -REGISTER_OP("BitwiseOr").BINARY_BITWISE().Doc(R"doc( +REGISTER_OP("BitwiseOr") + .BINARY_BITWISE() + .Doc(R"doc( Elementwise computes the bitwise OR of `x` and `y`. The result will have those bits set, that are set in `x`, `y` or both. The computation is performed on the underlying representations of `x` and `y`. )doc"); -REGISTER_OP("BitwiseXor").BINARY_BITWISE().Doc(R"doc( +REGISTER_OP("BitwiseXor") + .BINARY_BITWISE() + .Doc(R"doc( Elementwise computes the bitwise XOR of `x` and `y`. The result will have those bits set, that are different in `x` and `y`. The computation is performed on the underlying representations of `x` and `y`. )doc"); -REGISTER_OP("LeftShift").BINARY_BITWISE().Doc(R"doc( +REGISTER_OP("LeftShift") + .BINARY_BITWISE() + .Doc(R"doc( Elementwise computes the bitwise left-shift of `x` and `y`. If `y` is negative, or greater than or equal to the width of `x` in bits the result is implementation defined. )doc"); -REGISTER_OP("RightShift").BINARY_BITWISE().Doc(R"doc( +REGISTER_OP("RightShift") + .BINARY_BITWISE() + .Doc(R"doc( Elementwise computes the bitwise right-shift of `x` and `y`. Performs a logical shift for unsigned integer types, and an arithmetic shift diff --git a/tensorflow/core/ops/compat/backwards_compatibility_test.cc b/tensorflow/core/ops/compat/backwards_compatibility_test.cc index 6e05ae4be4..add05d6610 100644 --- a/tensorflow/core/ops/compat/backwards_compatibility_test.cc +++ b/tensorflow/core/ops/compat/backwards_compatibility_test.cc @@ -25,9 +25,8 @@ namespace tensorflow { namespace { TEST(BackwardsCompatibilityTest, IsCompatible) { - OpCompatibilityLib compatibility("tensorflow/core/ops", - strings::StrCat("v", TF_MAJOR_VERSION), - nullptr); + OpCompatibilityLib compatibility( + "tensorflow/core/ops", strings::StrCat("v", TF_MAJOR_VERSION), nullptr); Env* env = Env::Default(); int changed_ops = 0; diff --git a/tensorflow/core/ops/data_flow_ops.cc b/tensorflow/core/ops/data_flow_ops.cc index 8e24ea70cb..3b1ed217ce 100644 --- a/tensorflow/core/ops/data_flow_ops.cc +++ b/tensorflow/core/ops/data_flow_ops.cc @@ -2225,7 +2225,6 @@ this op will block until it does. This Op is optimized for performance. )doc"); - REGISTER_OP("StageSize") .Output("size: int32") .Attr("capacity: int >= 0 = 0") @@ -2354,7 +2353,6 @@ REGISTER_OP("MapIncompleteSize") Op returns the number of incomplete elements in the underlying container. )doc"); - REGISTER_OP("MapClear") .Attr("capacity: int >= 0 = 0") .Attr("memory_limit: int >= 0 = 0") @@ -2367,7 +2365,6 @@ REGISTER_OP("MapClear") Op removes all elements in the underlying container. )doc"); - // OrderedMap REGISTER_OP("OrderedMapStage") .Input("key: int64") diff --git a/tensorflow/core/ops/image_ops.cc b/tensorflow/core/ops/image_ops.cc index e9bf29d172..c3f8006415 100644 --- a/tensorflow/core/ops/image_ops.cc +++ b/tensorflow/core/ops/image_ops.cc @@ -925,27 +925,27 @@ use_image_if_no_bounding_boxes: Controls behavior if no bounding boxes supplied. )doc"); REGISTER_OP("SampleDistortedBoundingBoxV2") - .Input("image_size: T") - .Input("bounding_boxes: float") - .Input("min_object_covered: float") - .Output("begin: T") - .Output("size: T") - .Output("bboxes: float") - .Attr("T: {uint8, int8, int16, int32, int64}") - .Attr("seed: int = 0") - .Attr("seed2: int = 0") - .Attr("aspect_ratio_range: list(float) = [0.75, 1.33]") - .Attr("area_range: list(float) = [0.05, 1.0]") - .Attr("max_attempts: int = 100") - .Attr("use_image_if_no_bounding_boxes: bool = false") - .SetIsStateful() - .SetShapeFn([](InferenceContext* c) { - c->set_output(0, c->Vector(3)); - c->set_output(1, c->Vector(3)); - c->set_output(2, c->MakeShape({1, 1, 4})); - return Status::OK(); - }) - .Doc(R"doc( + .Input("image_size: T") + .Input("bounding_boxes: float") + .Input("min_object_covered: float") + .Output("begin: T") + .Output("size: T") + .Output("bboxes: float") + .Attr("T: {uint8, int8, int16, int32, int64}") + .Attr("seed: int = 0") + .Attr("seed2: int = 0") + .Attr("aspect_ratio_range: list(float) = [0.75, 1.33]") + .Attr("area_range: list(float) = [0.05, 1.0]") + .Attr("max_attempts: int = 100") + .Attr("use_image_if_no_bounding_boxes: bool = false") + .SetIsStateful() + .SetShapeFn([](InferenceContext* c) { + c->set_output(0, c->Vector(3)); + c->set_output(1, c->Vector(3)); + c->set_output(2, c->MakeShape({1, 1, 4})); + return Status::OK(); + }) + .Doc(R"doc( Generate a single randomly distorted bounding box for an image. Bounding box annotations are often supplied in addition to ground-truth labels @@ -1236,16 +1236,16 @@ method: A string specifying the interpolation method. Only 'bilinear' is // -------------------------------------------------------------------------- REGISTER_OP("NonMaxSuppression") - .Input("boxes: float") - .Input("scores: float") - .Input("max_output_size: int32") - .Output("selected_indices: int32") - .Attr("iou_threshold: float = 0.5") - .SetShapeFn([](InferenceContext* c) { + .Input("boxes: float") + .Input("scores: float") + .Input("max_output_size: int32") + .Output("selected_indices: int32") + .Attr("iou_threshold: float = 0.5") + .SetShapeFn([](InferenceContext* c) { c->set_output(0, c->Vector(c->UnknownDim())); return Status::OK(); }) - .Doc(R"doc( + .Doc(R"doc( Greedily selects a subset of bounding boxes in descending order of score, pruning away boxes that have high intersection-over-union (IOU) overlap with previously selected boxes. Bounding boxes are supplied as diff --git a/tensorflow/core/ops/linalg_ops.cc b/tensorflow/core/ops/linalg_ops.cc index 76e2149522..4851619f83 100644 --- a/tensorflow/core/ops/linalg_ops.cc +++ b/tensorflow/core/ops/linalg_ops.cc @@ -25,7 +25,6 @@ using shape_inference::ShapeHandle; namespace { - // Return in <out> the result of making the end of <s> a square matrix. Status MakeBatchSquareMatrix(InferenceContext* c, ShapeHandle input, ShapeHandle* out) { diff --git a/tensorflow/core/ops/math_grad_test.cc b/tensorflow/core/ops/math_grad_test.cc index 2b4b35547b..8dcd3e815f 100644 --- a/tensorflow/core/ops/math_grad_test.cc +++ b/tensorflow/core/ops/math_grad_test.cc @@ -385,7 +385,7 @@ class TestOp : public OpKernel { REGISTER_KERNEL_BUILDER(Name("TestOpWithNoGrad").Device(DEVICE_CPU), TestOp); #ifdef TENSORFLOW_USE_SYCL REGISTER_KERNEL_BUILDER(Name("TestOpWithNoGrad").Device(DEVICE_SYCL), TestOp); -#endif // TENSORFLOW_USE_SYCL +#endif // TENSORFLOW_USE_SYCL TEST_F(MathGradTest, Error_Reporting) { auto x = test::AsTensor<float>({-3.f}); @@ -557,11 +557,10 @@ TEST_F(MathGradTest, Acosh) { TEST_F(MathGradTest, Atanh) { auto x = test::AsTensor<float>({-0.3f, -0.2f, -0.1f, 0.1f, 0.2f, 0.3f}, TensorShape({2, 3})); - auto g = [](float x) { - return 1.f / (1.f - x * x); - }; + auto g = [](float x) { return 1.f / (1.f - x * x); }; auto dx = test::AsTensor<float>( - {g(-0.3f), g(-0.2f), g(-0.1f), g(0.1f), g(0.2f), g(0.3f)}, TensorShape({2, 3})); + {g(-0.3f), g(-0.2f), g(-0.1f), g(0.1f), g(0.2f), g(0.3f)}, + TensorShape({2, 3})); auto ans = SymGrad("Atanh", x); test::ExpectClose(ans, dx); } @@ -761,7 +760,7 @@ TEST_F(MathGradTest, Pow) { } } -//TODO{lukeiwanski}: Implement Complex Pow for SYCL +// TODO{lukeiwanski}: Implement Complex Pow for SYCL #ifndef TENSORFLOW_USE_SYCL TEST_F(MathGradTest, ComplexPow) { auto x = test::AsTensor<complex64>({0.f, 2.f, -2.f}, TensorShape({3})); @@ -781,7 +780,7 @@ TEST_F(MathGradTest, ComplexPow) { dy, test::AsTensor<complex64>({h(0.f, 2.f), h(2.f, 2.f), h(-2.f, 2.f)}, TensorShape({3}))); } -#endif // TENSORFLOW_USE_SYCL +#endif // TENSORFLOW_USE_SYCL TEST_F(MathGradTest, Maximum) { auto x = test::AsTensor<float>({-3.f, -2.f, -1.f, 1.f, 2.f, 3.f}, @@ -943,7 +942,7 @@ TEST_F(MathGradTest, MatMul_11) { test::ExpectClose(dy, MatMul(dz, true, x, true)); } -//TODO{lukeiwanski}: Implement BatchMatMul for SYCL +// TODO{lukeiwanski}: Implement BatchMatMul for SYCL #ifndef TENSORFLOW_USE_SYCL TEST_F(MathGradTest, BatchMatMul_00) { auto x = test::AsTensor<float>({1.f, 2.f, 3.f, 4.f, 5.f, 6.f}, @@ -992,7 +991,7 @@ TEST_F(MathGradTest, BatchMatMul_11) { test::ExpectClose(dx, BatchMatMul(y, true, dz, true)); test::ExpectClose(dy, BatchMatMul(dz, true, x, true)); } -#endif // TENSORFLOW_USE_SYCL +#endif // TENSORFLOW_USE_SYCL TEST_F(MathGradTest, Sum_dim0) { auto x = test::AsTensor<float>({-3.f, -2.f, -1.f, 1.f, 2.f, 3.f}, diff --git a/tensorflow/core/ops/math_ops.cc b/tensorflow/core/ops/math_ops.cc index 045b0795ed..7b10af9f44 100644 --- a/tensorflow/core/ops/math_ops.cc +++ b/tensorflow/core/ops/math_ops.cc @@ -235,7 +235,9 @@ value is computed as \\( \sqrt{a^2 + b^2}\\). .Attr("T: {half, float, double, complex64, complex128}") \ .SetShapeFn(shape_inference::UnchangedShape) -REGISTER_OP("Neg").UNARY().Doc(R"doc( +REGISTER_OP("Neg") + .UNARY() + .Doc(R"doc( Computes numerical negative value element-wise. I.e., \\(y = -x\\). )doc"); @@ -258,155 +260,217 @@ is the corresponding input gradient. )doc") .Deprecated(17, "Use ReciprocalGrad"); -REGISTER_OP("Reciprocal").UNARY().Doc(R"doc( +REGISTER_OP("Reciprocal") + .UNARY() + .Doc(R"doc( Computes the reciprocal of x element-wise. I.e., \\(y = 1 / x\\). )doc"); -REGISTER_OP("ReciprocalGrad").UNARY_GRADIENT_COMPLEX().Doc(R"doc( +REGISTER_OP("ReciprocalGrad") + .UNARY_GRADIENT_COMPLEX() + .Doc(R"doc( Computes the gradient for the inverse of `x` wrt its input. Specifically, `grad = -dy * y*y`, where `y = 1/x`, and `dy` is the corresponding input gradient. )doc"); -REGISTER_OP("Square").UNARY().Doc(R"doc( +REGISTER_OP("Square") + .UNARY() + .Doc(R"doc( Computes square of x element-wise. I.e., \\(y = x * x = x^2\\). )doc"); -REGISTER_OP("Sqrt").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Sqrt") + .UNARY_COMPLEX() + .Doc(R"doc( Computes square root of x element-wise. I.e., \\(y = \sqrt{x} = x^{1/2}\\). )doc"); -REGISTER_OP("SqrtGrad").UNARY_GRADIENT_COMPLEX().Doc(R"doc( +REGISTER_OP("SqrtGrad") + .UNARY_GRADIENT_COMPLEX() + .Doc(R"doc( Computes the gradient for the sqrt of `x` wrt its input. Specifically, `grad = dy * 0.5 / y`, where `y = sqrt(x)`, and `dy` is the corresponding input gradient. )doc"); -REGISTER_OP("Rsqrt").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Rsqrt") + .UNARY_COMPLEX() + .Doc(R"doc( Computes reciprocal of square root of x element-wise. I.e., \\(y = 1 / \sqrt{x}\\). )doc"); -REGISTER_OP("Round").UNARY().Doc(R"doc( +REGISTER_OP("Round") + .UNARY() + .Doc(R"doc( Rounds the values of a tensor to the nearest integer, element-wise. Rounds half to even. Also known as bankers rounding. If you want to round according to the current system rounding mode use std::cint. )doc"); -REGISTER_OP("RsqrtGrad").UNARY_GRADIENT_COMPLEX().Doc(R"doc( +REGISTER_OP("RsqrtGrad") + .UNARY_GRADIENT_COMPLEX() + .Doc(R"doc( Computes the gradient for the rsqrt of `x` wrt its input. Specifically, `grad = dy * -0.5 * y^3`, where `y = rsqrt(x)`, and `dy` is the corresponding input gradient. )doc"); -REGISTER_OP("Exp").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Exp") + .UNARY_COMPLEX() + .Doc(R"doc( Computes exponential of x element-wise. \\(y = e^x\\). )doc"); -REGISTER_OP("Expm1").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Expm1") + .UNARY_COMPLEX() + .Doc(R"doc( Computes exponential of x - 1 element-wise. I.e., \\(y = (\exp x) - 1\\). )doc"); -REGISTER_OP("Log").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Log") + .UNARY_COMPLEX() + .Doc(R"doc( Computes natural logarithm of x element-wise. I.e., \\(y = \log_e x\\). )doc"); -REGISTER_OP("Log1p").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Log1p") + .UNARY_COMPLEX() + .Doc(R"doc( Computes natural logarithm of (1 + x) element-wise. I.e., \\(y = \log_e (1 + x)\\). )doc"); -REGISTER_OP("Sinh").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Sinh") + .UNARY_COMPLEX() + .Doc(R"doc( Computes hyperbolic sine of x element-wise. )doc"); -REGISTER_OP("Cosh").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Cosh") + .UNARY_COMPLEX() + .Doc(R"doc( Computes hyperbolic cosine of x element-wise. )doc"); -REGISTER_OP("Tanh").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Tanh") + .UNARY_COMPLEX() + .Doc(R"doc( Computes hyperbolic tangent of `x` element-wise. )doc"); -REGISTER_OP("Asinh").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Asinh") + .UNARY_COMPLEX() + .Doc(R"doc( Computes inverse hyperbolic sine of x element-wise. )doc"); -REGISTER_OP("Acosh").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Acosh") + .UNARY_COMPLEX() + .Doc(R"doc( Computes inverse hyperbolic cosine of x element-wise. )doc"); -REGISTER_OP("Atanh").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Atanh") + .UNARY_COMPLEX() + .Doc(R"doc( Computes inverse hyperbolic tangent of x element-wise. )doc"); -REGISTER_OP("TanhGrad").UNARY_GRADIENT_COMPLEX().Doc(R"doc( +REGISTER_OP("TanhGrad") + .UNARY_GRADIENT_COMPLEX() + .Doc(R"doc( Computes the gradient for the tanh of `x` wrt its input. Specifically, `grad = dy * (1 - y*y)`, where `y = tanh(x)`, and `dy` is the corresponding input gradient. )doc"); -REGISTER_OP("Lgamma").UNARY_REAL().Doc(R"doc( +REGISTER_OP("Lgamma") + .UNARY_REAL() + .Doc(R"doc( Computes the log of the absolute value of `Gamma(x)` element-wise. )doc"); -REGISTER_OP("Digamma").UNARY_REAL().Doc(R"doc( +REGISTER_OP("Digamma") + .UNARY_REAL() + .Doc(R"doc( Computes Psi, the derivative of Lgamma (the log of the absolute value of `Gamma(x)`), element-wise. )doc"); -REGISTER_OP("Erf").UNARY_REAL().Doc(R"doc( +REGISTER_OP("Erf") + .UNARY_REAL() + .Doc(R"doc( Computes the Gauss error function of `x` element-wise. )doc"); -REGISTER_OP("Erfc").UNARY_REAL().Doc(R"doc( +REGISTER_OP("Erfc") + .UNARY_REAL() + .Doc(R"doc( Computes the complementary error function of `x` element-wise. )doc"); -REGISTER_OP("Sigmoid").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Sigmoid") + .UNARY_COMPLEX() + .Doc(R"doc( Computes sigmoid of `x` element-wise. Specifically, `y = 1 / (1 + exp(-x))`. )doc"); -REGISTER_OP("SigmoidGrad").UNARY_GRADIENT_COMPLEX().Doc(R"doc( +REGISTER_OP("SigmoidGrad") + .UNARY_GRADIENT_COMPLEX() + .Doc(R"doc( Computes the gradient of the sigmoid of `x` wrt its input. Specifically, `grad = dy * y * (1 - y)`, where `y = sigmoid(x)`, and `dy` is the corresponding input gradient. )doc"); -REGISTER_OP("Sin").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Sin") + .UNARY_COMPLEX() + .Doc(R"doc( Computes sin of x element-wise. )doc"); -REGISTER_OP("Cos").UNARY_COMPLEX().Doc(R"doc( +REGISTER_OP("Cos") + .UNARY_COMPLEX() + .Doc(R"doc( Computes cos of x element-wise. )doc"); -REGISTER_OP("Tan").UNARY().Doc(R"doc( +REGISTER_OP("Tan") + .UNARY() + .Doc(R"doc( Computes tan of x element-wise. )doc"); -REGISTER_OP("Asin").UNARY().Doc(R"doc( +REGISTER_OP("Asin") + .UNARY() + .Doc(R"doc( Computes asin of x element-wise. )doc"); -REGISTER_OP("Acos").UNARY().Doc(R"doc( +REGISTER_OP("Acos") + .UNARY() + .Doc(R"doc( Computes acos of x element-wise. )doc"); -REGISTER_OP("Atan").UNARY().Doc(R"doc( +REGISTER_OP("Atan") + .UNARY() + .Doc(R"doc( Computes atan of x element-wise. )doc"); @@ -960,28 +1024,36 @@ beta function. .Attr("T: realnumbertype") \ .SetShapeFn(shape_inference::BroadcastBinaryOpShapeFn) -REGISTER_OP("Less").COMPARISON().Doc(R"doc( +REGISTER_OP("Less") + .COMPARISON() + .Doc(R"doc( Returns the truth value of (x < y) element-wise. *NOTE*: `Less` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html) )doc"); -REGISTER_OP("LessEqual").COMPARISON().Doc(R"doc( +REGISTER_OP("LessEqual") + .COMPARISON() + .Doc(R"doc( Returns the truth value of (x <= y) element-wise. *NOTE*: `LessEqual` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html) )doc"); -REGISTER_OP("Greater").COMPARISON().Doc(R"doc( +REGISTER_OP("Greater") + .COMPARISON() + .Doc(R"doc( Returns the truth value of (x > y) element-wise. *NOTE*: `Greater` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html) )doc"); -REGISTER_OP("GreaterEqual").COMPARISON().Doc(R"doc( +REGISTER_OP("GreaterEqual") + .COMPARISON() + .Doc(R"doc( Returns the truth value of (x >= y) element-wise. *NOTE*: `GreaterEqual` supports broadcasting. More about broadcasting @@ -1003,14 +1075,18 @@ Returns the truth value of (x >= y) element-wise. "quint8, qint8, qint32, string, bool, complex128}") \ .SetShapeFn(shape_inference::BroadcastBinaryOpShapeFn) -REGISTER_OP("Equal").EQUALITY_COMPARISON().Doc(R"doc( +REGISTER_OP("Equal") + .EQUALITY_COMPARISON() + .Doc(R"doc( Returns the truth value of (x == y) element-wise. *NOTE*: `Equal` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html) )doc"); -REGISTER_OP("NotEqual").EQUALITY_COMPARISON().Doc(R"doc( +REGISTER_OP("NotEqual") + .EQUALITY_COMPARISON() + .Doc(R"doc( Returns the truth value of (x != y) element-wise. *NOTE*: `NotEqual` supports broadcasting. More about broadcasting @@ -1048,14 +1124,18 @@ Returns the truth value of NOT x element-wise. .SetIsCommutative() \ .SetShapeFn(shape_inference::BroadcastBinaryOpShapeFn) -REGISTER_OP("LogicalAnd").BINARY_LOGICAL().Doc(R"doc( +REGISTER_OP("LogicalAnd") + .BINARY_LOGICAL() + .Doc(R"doc( Returns the truth value of x AND y element-wise. *NOTE*: `LogicalAnd` supports broadcasting. More about broadcasting [here](http://docs.scipy.org/doc/numpy/user/basics.broadcasting.html) )doc"); -REGISTER_OP("LogicalOr").BINARY_LOGICAL().Doc(R"doc( +REGISTER_OP("LogicalOr") + .BINARY_LOGICAL() + .Doc(R"doc( Returns the truth value of x OR y element-wise. *NOTE*: `LogicalOr` supports broadcasting. More about broadcasting @@ -1995,12 +2075,12 @@ Status RangeSize(const Tensor* start_t, const Tensor* limit_t, T limit = limit_t->scalar<T>()(); T delta = delta_t->scalar<T>()(); if (start > limit && delta > 0) { - return errors::InvalidArgument( - "Requires start <= limit when delta > 0: ", start, "/", limit); + return errors::InvalidArgument("Requires start <= limit when delta > 0: ", + start, "/", limit); } if (start < limit && delta < 0) { - return errors::InvalidArgument( - "Requires start >= limit when delta < 0: ", start, "/", limit); + return errors::InvalidArgument("Requires start >= limit when delta < 0: ", + start, "/", limit); } if (delta == 0) { return errors::InvalidArgument("Requires delta != 0"); diff --git a/tensorflow/core/ops/nn_ops.cc b/tensorflow/core/ops/nn_ops.cc index 1d26660a4b..de059a3e7e 100644 --- a/tensorflow/core/ops/nn_ops.cc +++ b/tensorflow/core/ops/nn_ops.cc @@ -2176,9 +2176,9 @@ Status TopKShapeFn(InferenceContext* c) { DimensionHandle last_dim = c->Dim(input, -1); if (c->ValueKnown(last_dim) && c->ValueKnown(k_dim) && c->Value(last_dim) < c->Value(k_dim)) { - return errors::InvalidArgument( - "input must have last dimension >= k = ", c->Value(k_dim), " but is ", - c->Value(last_dim)); + return errors::InvalidArgument("input must have last dimension >= k = ", + c->Value(k_dim), " but is ", + c->Value(last_dim)); } // Replace last_dim with k_dim. @@ -2278,9 +2278,9 @@ REGISTER_OP("NthElement") DimensionHandle last_dim = c->Dim(input, -1); if (c->ValueKnown(last_dim) && c->ValueKnown(n_dim) && c->Value(last_dim) <= c->Value(n_dim)) { - return errors::InvalidArgument( - "Input must have last dimension > n = ", c->Value(n_dim), " but is ", - c->Value(last_dim)); + return errors::InvalidArgument("Input must have last dimension > n = ", + c->Value(n_dim), " but is ", + c->Value(last_dim)); } // Reduce last_dim for output tensor diff --git a/tensorflow/core/ops/nn_ops_test.cc b/tensorflow/core/ops/nn_ops_test.cc index 94ecf4d5db..1b17a7cda6 100644 --- a/tensorflow/core/ops/nn_ops_test.cc +++ b/tensorflow/core/ops/nn_ops_test.cc @@ -95,14 +95,13 @@ TEST(NNOpsTest, NthElement_ShapeFn) { INFER_OK(op, "[?,3,?,21];[]", "[d0_0,d0_1,d0_2]"); INFER_ERROR("Shape must be at least rank 1 but is rank 0", op, "[];[]"); - INFER_ERROR("Input must have last dimension > n = 20 but is 1", op, - "[1];[]"); + INFER_ERROR("Input must have last dimension > n = 20 but is 1", op, "[1];[]"); INFER_ERROR("Input must have last dimension > n = 20 but is 20", op, "[1,2,3,20];[]"); n_t = test::AsScalar<int32>(-1); INFER_ERROR( - "Dimension size, given by scalar input 1, must be non-negative but is -1", - op, "[1,2,3,4];[]"); + "Dimension size, given by scalar input 1, must be non-negative but is -1", + op, "[1,2,3,4];[]"); } TEST(NNOpsTest, BatchNormWithGlobalNormalization_ShapeFn) { @@ -386,9 +385,8 @@ TEST(NNOpsTest, Dilation2DBackpropFilter_ShapeFn) { } TEST(NNOpsTest, MergeBothInputs_ShapeFn) { - for (const char* op_name : - {"ReluGrad", "Relu6Grad", "EluGrad", "SeluGrad", "SoftplusGrad", - "SoftsignGrad"}) { + for (const char* op_name : {"ReluGrad", "Relu6Grad", "EluGrad", "SeluGrad", + "SoftplusGrad", "SoftsignGrad"}) { ShapeInferenceTestOp op(op_name); INFER_OK(op, "?;?", "in0|in1"); diff --git a/tensorflow/core/ops/parsing_ops.cc b/tensorflow/core/ops/parsing_ops.cc index b44ea2e080..40ec792ef8 100644 --- a/tensorflow/core/ops/parsing_ops.cc +++ b/tensorflow/core/ops/parsing_ops.cc @@ -329,7 +329,7 @@ REGISTER_OP("DecodeCSV") .Input("records: string") .Input("record_defaults: OUT_TYPE") .Output("output: OUT_TYPE") - .Attr("OUT_TYPE: list({float,int32,int64,string})") + .Attr("OUT_TYPE: list({float,double,int32,int64,string})") .Attr("field_delim: string = ','") .Attr("use_quote_delim: bool = true") .Attr("na_value: string = ''") diff --git a/tensorflow/core/ops/sparse_ops_test.cc b/tensorflow/core/ops/sparse_ops_test.cc index ea49f1a199..0df3320484 100644 --- a/tensorflow/core/ops/sparse_ops_test.cc +++ b/tensorflow/core/ops/sparse_ops_test.cc @@ -187,8 +187,8 @@ TEST(SparseOpsTest, SparseTensorDenseMatMul_ShapeFn) { // second output dim comes from b, depending on adjoint_b value. INFER_OK(op, "?;?;?;?", "[?,?]"); - INFER_OK(op, "?;?;?;[?,?]", "[?,d3_1]"); // use d3_1, !adjoint_b. - INFER_OK(op, "?;?;?;[1,2]", "[?,d3_1]"); // use d3_1, !adjoint_b. + INFER_OK(op, "?;?;?;[?,?]", "[?,d3_1]"); // use d3_1, !adjoint_b. + INFER_OK(op, "?;?;?;[1,2]", "[?,d3_1]"); // use d3_1, !adjoint_b. INFER_OK(op, "?;?;[2];[1,2]", "[?,d3_1]"); // use d3_1, !adjoint_b. set_adjoints(false, true); diff --git a/tensorflow/core/ops/stateless_random_ops.cc b/tensorflow/core/ops/stateless_random_ops.cc index b222b5b241..7c00fdb99f 100644 --- a/tensorflow/core/ops/stateless_random_ops.cc +++ b/tensorflow/core/ops/stateless_random_ops.cc @@ -45,7 +45,8 @@ static Status StatelessShape(shape_inference::InferenceContext* context) { .SetShapeFn(StatelessShape) // This op is exposed through contrib/stateless only. The interface may change. -REGISTER_STATELESS_OP("StatelessRandomUniform").Doc(R"doc( +REGISTER_STATELESS_OP("StatelessRandomUniform") + .Doc(R"doc( Outputs deterministic pseudorandom random values from a uniform distribution. The generated values follow a uniform distribution in the range `[0, 1)`. The @@ -60,7 +61,8 @@ output: Random values with specified shape. )doc"); // This op is exposed through contrib/stateless only. The interface may change. -REGISTER_STATELESS_OP("StatelessRandomNormal").Doc(R"doc( +REGISTER_STATELESS_OP("StatelessRandomNormal") + .Doc(R"doc( Outputs deterministic pseudorandom values from a normal distribution. The generated values will have mean 0 and standard deviation 1. @@ -74,7 +76,8 @@ output: Random values with specified shape. )doc"); // This op is exposed through contrib/stateless only. The interface may change. -REGISTER_STATELESS_OP("StatelessTruncatedNormal").Doc(R"doc( +REGISTER_STATELESS_OP("StatelessTruncatedNormal") + .Doc(R"doc( Outputs deterministic pseudorandom values from a truncated normal distribution. The generated values follow a normal distribution with mean 0 and standard diff --git a/tensorflow/core/public/version.h b/tensorflow/core/public/version.h index bd590be460..1bf9c93101 100644 --- a/tensorflow/core/public/version.h +++ b/tensorflow/core/public/version.h @@ -24,7 +24,7 @@ limitations under the License. // TF_VERSION_SUFFIX is non-empty for pre-releases (e.g. "-alpha", "-alpha.1", // "-beta", "-rc", "-rc.1") -#define TF_VERSION_SUFFIX "-rc0" +#define TF_VERSION_SUFFIX "-rc1" #define TF_STR_HELPER(x) #x #define TF_STR(x) TF_STR_HELPER(x) @@ -117,5 +117,7 @@ extern const char* tf_compiler_version(); // The git commit designator when tensorflow was built // If no git repository, this will be "internal". extern const char* tf_git_version(); +// Value of the _GLIBCXX_USE_CXX11_ABI flag, or 0 if it's not set. +extern const int tf_cxx11_abi_flag(); #endif // TENSORFLOW_CORE_PUBLIC_VERSION_H_ diff --git a/tensorflow/docs_src/api_guides/cc/guide.md b/tensorflow/docs_src/api_guides/cc/guide.md index f30bf3797e..81fb1e1fda 100644 --- a/tensorflow/docs_src/api_guides/cc/guide.md +++ b/tensorflow/docs_src/api_guides/cc/guide.md @@ -1,4 +1,12 @@ # C++ API + +Note: By default [tensorflow.org](http://tensorflow.org) shows docs for the +most recent stable version. The instructions in this doc require building from +source. You will probably want to build from the `master` version of tensorflow. +You should, as a result, be sure you are following the +[`master` version of this doc](https://www.tensorflow.org/versions/master/api_guides/cc/guide), +in case there have been any changes. + [TOC] TensorFlow's C++ API provides mechanisms for constructing and executing a data @@ -48,7 +56,9 @@ TensorFlow `BUILD` file in the same directory with the following contents: ```python -cc_binary( +load("//tensorflow:tensorflow.bzl", "tf_cc_binary") + +tf_cc_binary( name = "example", srcs = ["example.cc"], deps = [ @@ -59,8 +69,10 @@ cc_binary( ) ``` -You should be able to build and run the example using the following command -(be sure to run `./configure` in your build sandbox first): +Use `tf_cc_binary` rather than Bazel's native `cc_binary` to link in necessary +symbols from `libtensorflow_framework.so`. You should be able to build and run +the example using the following command (be sure to run `./configure` in your +build sandbox first): ```shell bazel run -c opt //tensorflow/cc/example:example diff --git a/tensorflow/docs_src/community/documentation.md b/tensorflow/docs_src/community/documentation.md index 655506b098..77d4e0caec 100644 --- a/tensorflow/docs_src/community/documentation.md +++ b/tensorflow/docs_src/community/documentation.md @@ -1,6 +1,6 @@ # Writing TensorFlow Documentation -We welcome contributions to the Tensorflow documentation from the community. +We welcome contributions to the TensorFlow documentation from the community. This document explains how you can contribute to that documentation. In particular, this document explains the following: @@ -8,28 +8,30 @@ particular, this document explains the following: * How to make conformant edits. * How to build and test your documentation changes before you submit them. -You can view Tensorflow documentation on https://www.tensorflow.org, and you -can view and edit the raw files on Github. We're publishing our docs on Github -so everybody can contribute. Whatever gets checked in tensorflow/docs_src will -be published soon after on https://www.tensorflow.org. +You can view TensorFlow documentation on https://www.tensorflow.org, and you +can view and edit the raw files on +[GitHub](https://www.tensorflow.org/code/tensorflow/docs_src/). +We're publishing our docs on GitHub so everybody can contribute. Whatever gets +checked in to `tensorflow/docs_src` will be published soon after on +https://www.tensorflow.org. Republishing TensorFlow documentation in different forms is absolutely allowed, but we are unlikely to accept other documentation formats (or the tooling to generate them) into our repository. If you do choose to republish our documentation in another form, please be sure to include: -* The version of the API this represents (i.e. r1.0, master, etc.) +* The version of the API this represents (for example, r1.0, master, etc.) * The commit or version from which the documentation was generated * Where to get the latest documentation (that is, https://www.tensorflow.org) * The Apache 2.0 license. -## A Note on Versions +## A note on versions tensorflow.org, at root, shows documentation for the latest stable binary. This is the documentation you should be reading if you are using `pip` to install TensorFlow. -However, most developers will contribute documentation into the master Github +However, most developers will contribute documentation into the master GitHub branch, which is published, occasionally, at [tensorflow.org/versions/master](https://www.tensorflow.org/versions/master). @@ -49,8 +51,9 @@ in the code: To modify the reference documentation, you edit the appropriate code comments. Non-reference documentation (for example, the TensorFlow installation guides) is -authored by humans. This documentation is located in the `tensorflow/docs_src` -directory. Each subdirectory of `docs_src` contains a set of related Tensorflow +authored by humans. This documentation is located in the +[`tensorflow/docs_src`](https://www.tensorflow.org/code/tensorflow/docs_src/) +directory. Each subdirectory of `docs_src` contains a set of related TensorFlow documentation. For example, the TensorFlow installation guides are all in the `docs_src/install` directory. @@ -183,7 +186,7 @@ documentation in the `/tmp/tfdocs` dir: Note: You must set `src_dir` and `output_dir` to absolute file paths. -## Generating Python API Documentation +## Generating Python API documentation Ops, classes, and utility functions are defined in Python modules, such as `image_ops.py`. Python modules contain a module docstring. For example: @@ -216,7 +219,7 @@ the following: Only top level modules (currently just `tf` and `tfdbg`) need to be manually added to the generate script. -### Sealing Modules +### Sealing modules Because the doc generator walks all visible symbols, and descends into anything it finds, it will document any accidentally exposed symbols. If a module only @@ -242,7 +245,7 @@ following options for dealing with them: We'll discuss these options in detail below. -#### Private Symbols and Imports +#### Private symbols and imports The easiest way to conform to the API sealing expectations is to make non-public symbols private (by prepending an underscore _). The doc generator respects @@ -288,7 +291,7 @@ are public. All `@@`s will eventually be removed. If you see them, however, please do not randomly delete them as they are still in use by some of our systems. -#### Traversal Blacklist +#### Traversal blacklist If all else fails, you may add entries to the traversal blacklist in `generate_lib.py.` **Almost all entries in this list are an abuse of its @@ -311,7 +314,7 @@ flags, ...) included for platform abstraction can be documented without documenting their interior. Its use beyond this purpose is a shortcut that may be acceptable for contrib, but not for core tensorflow. -## Op Documentation Style Guide +## Op documentation style guide Long, descriptive module-level documentation for modules should go in the API Guides in `docs_src/api_guides/python`. @@ -334,7 +337,7 @@ is [here](https://daringfireball.net/projects/markdown/). You are allowed to use [MathJax](https://www.mathjax.org) notation for equations (see above for restrictions). -### Writing About Code +### Writing about code Put backticks around these things when they're used in text: @@ -375,7 +378,7 @@ Two notes about backticks for code samples in Markdown: However, do NOT indent four spaces and use backticks simultaneously. Use one or the other. -### Tensor Dimensions +### Tensor dimensions When you're talking about a tensor in general, don't capitalize the word tensor. When you're talking about the specific object that's provided to an op as an @@ -500,7 +503,7 @@ def foo(x, y, name="bar"): """ ``` -## Description of the Docstring Sections +## Description of the docstring sections This section details each of the elements in docstrings. diff --git a/tensorflow/docs_src/community/welcome.md b/tensorflow/docs_src/community/welcome.md index c4f78051f0..33740de5d5 100644 --- a/tensorflow/docs_src/community/welcome.md +++ b/tensorflow/docs_src/community/welcome.md @@ -20,7 +20,6 @@ The TensorFlow community has created many great projects around TensorFlow, incl * [Machine Learning with TensorFlow (Book & Code)](http://tensorflowbook.com) * [@jtoy's awesome "Awesome TensorFlow" list of awesome things](https://github.com/jtoy/awesome-tensorflow) * [TensorFlow tutorials](https://github.com/pkmital/tensorflow_tutorials) -* [Scikit Flow - Simplified Interface for TensorFlow](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/learn/python/learn) * [Caffe to TensorFlow model converter](https://github.com/ethereon/caffe-tensorflow) * [Bitfusion's` GPU-enabled AWS EC2 TensorFlow AMI](https://github.com/bitfusionio/amis/tree/master/awsmrkt-bfboost-ubuntu14-cuda75-tensorflow) ([Launch AMI](https://aws.amazon.com/marketplace/pp/B01EYKBEQ0)) * [Rust language bindings](https://github.com/google/tensorflow-rust) diff --git a/tensorflow/docs_src/install/install_c.md b/tensorflow/docs_src/install/install_c.md index 70f756b194..3a153e8114 100644 --- a/tensorflow/docs_src/install/install_c.md +++ b/tensorflow/docs_src/install/install_c.md @@ -38,7 +38,7 @@ enable TensorFlow for C: OS="linux" # Change to "darwin" for macOS TARGET_DIRECTORY="/usr/local" curl -L \ - "https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-${TF_TYPE}-${OS}-x86_64-1.4.0-rc0.tar.gz" | + "https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-${TF_TYPE}-${OS}-x86_64-1.4.0-rc1.tar.gz" | sudo tar -C $TARGET_DIRECTORY -xz The `tar` command extracts the TensorFlow C library into the `lib` diff --git a/tensorflow/docs_src/install/install_go.md b/tensorflow/docs_src/install/install_go.md index eca2ecc5ac..df43255896 100644 --- a/tensorflow/docs_src/install/install_go.md +++ b/tensorflow/docs_src/install/install_go.md @@ -38,7 +38,7 @@ steps to install this library and enable TensorFlow for Go: TF_TYPE="cpu" # Change to "gpu" for GPU support TARGET_DIRECTORY='/usr/local' curl -L \ - "https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-${TF_TYPE}-$(go env GOOS)-x86_64-1.4.0-rc0.tar.gz" | + "https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-${TF_TYPE}-$(go env GOOS)-x86_64-1.4.0-rc1.tar.gz" | sudo tar -C $TARGET_DIRECTORY -xz The `tar` command extracts the TensorFlow C library into the `lib` diff --git a/tensorflow/docs_src/install/install_java.md b/tensorflow/docs_src/install/install_java.md index 8eaec3712a..f7f2c3cdc7 100644 --- a/tensorflow/docs_src/install/install_java.md +++ b/tensorflow/docs_src/install/install_java.md @@ -36,7 +36,7 @@ following to the project's `pom.xml` to use the TensorFlow Java APIs: <dependency> <groupId>org.tensorflow</groupId> <artifactId>tensorflow</artifactId> - <version>1.4.0-rc0</version> + <version>1.4.0-rc1</version> </dependency> ``` @@ -65,7 +65,7 @@ As an example, these steps will create a Maven project that uses TensorFlow: <dependency> <groupId>org.tensorflow</groupId> <artifactId>tensorflow</artifactId> - <version>1.4.0-rc0</version> + <version>1.4.0-rc1</version> </dependency> </dependencies> </project> @@ -124,7 +124,7 @@ refer to the simpler instructions above instead. Take the following steps to install TensorFlow for Java on Linux or macOS: 1. Download - [libtensorflow.jar](https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-1.4.0-rc0.jar), + [libtensorflow.jar](https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-1.4.0-rc1.jar), which is the TensorFlow Java Archive (JAR). 2. Decide whether you will run TensorFlow for Java on CPU(s) only or with @@ -143,7 +143,7 @@ Take the following steps to install TensorFlow for Java on Linux or macOS: OS=$(uname -s | tr '[:upper:]' '[:lower:]') mkdir -p ./jni curl -L \ - "https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow_jni-${TF_TYPE}-${OS}-x86_64-1.4.0-rc0.tar.gz" | + "https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow_jni-${TF_TYPE}-${OS}-x86_64-1.4.0-rc1.tar.gz" | tar -xz -C ./jni ### Install on Windows @@ -151,10 +151,10 @@ Take the following steps to install TensorFlow for Java on Linux or macOS: Take the following steps to install TensorFlow for Java on Windows: 1. Download - [libtensorflow.jar](https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-1.4.0-rc0.jar), + [libtensorflow.jar](https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-1.4.0-rc1.jar), which is the TensorFlow Java Archive (JAR). 2. Download the following Java Native Interface (JNI) file appropriate for - [TensorFlow for Java on Windows](https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow_jni-cpu-windows-x86_64-1.4.0-rc0.zip). + [TensorFlow for Java on Windows](https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow_jni-cpu-windows-x86_64-1.4.0-rc1.zip). 3. Extract this .zip file. @@ -202,7 +202,7 @@ must be part of your `classpath`. For example, you can include the downloaded `.jar` in your `classpath` by using the `-cp` compilation flag as follows: -<pre><b>javac -cp libtensorflow-1.4.0-rc0.jar HelloTF.java</b></pre> +<pre><b>javac -cp libtensorflow-1.4.0-rc1.jar HelloTF.java</b></pre> ### Running @@ -216,11 +216,11 @@ two files are available to the JVM: For example, the following command line executes the `HelloTF` program on Linux and macOS X: -<pre><b>java -cp libtensorflow-1.4.0-rc0.jar:. -Djava.library.path=./jni HelloTF</b></pre> +<pre><b>java -cp libtensorflow-1.4.0-rc1.jar:. -Djava.library.path=./jni HelloTF</b></pre> And the following command line executes the `HelloTF` program on Windows: -<pre><b>java -cp libtensorflow-1.4.0-rc0.jar;. -Djava.library.path=jni HelloTF</b></pre> +<pre><b>java -cp libtensorflow-1.4.0-rc1.jar;. -Djava.library.path=jni HelloTF</b></pre> If the program prints <tt>Hello from <i>version</i></tt>, you've successfully installed TensorFlow for Java and are ready to use the API. If the program diff --git a/tensorflow/docs_src/install/install_linux.md b/tensorflow/docs_src/install/install_linux.md index 2b321e7dcb..414ab7b1f7 100644 --- a/tensorflow/docs_src/install/install_linux.md +++ b/tensorflow/docs_src/install/install_linux.md @@ -81,22 +81,22 @@ TensorFlow with GPU support, but only if you do the following: You must pick the mechanism by which you install TensorFlow. The supported choices are as follows: - * [virtualenv](#InstallingVirtualenv) + * [Virtualenv](#InstallingVirtualenv) * ["native" pip](#InstallingNativePip) * [Docker](#InstallingDocker) * [Anaconda](#InstallingAnaconda) * installing from sources, which is documented in [a separate guide](https://www.tensorflow.org/install/install_sources). -**We recommend the virtualenv installation.** +**We recommend the Virtualenv installation.** [Virtualenv](https://virtualenv.pypa.io/en/stable/) is a virtual Python environment isolated from other Python development, incapable of interfering with or being affected by other Python programs -on the same machine. During the virtualenv installation process, +on the same machine. During the Virtualenv installation process, you will install not only TensorFlow but also all the packages that TensorFlow requires. (This is actually pretty easy.) To start working with TensorFlow, you simply need to "activate" the -virtual environment. All in all, virtualenv provides a safe and +virtual environment. All in all, Virtualenv provides a safe and reliable mechanism for installing and running TensorFlow. Native pip installs TensorFlow directly on your system without going @@ -125,26 +125,26 @@ Use that package at your own risk. <a name="InstallingVirtualenv"></a> -## Installing with virtualenv +## Installing with Virtualenv Take the following steps to install TensorFlow with Virtualenv: - 1. Install pip and virtualenv by issuing one of the following commands: + 1. Install pip and Virtualenv by issuing one of the following commands: <pre>$ <b>sudo apt-get install python-pip python-dev python-virtualenv</b> # for Python 2.7 $ <b>sudo apt-get install python3-pip python3-dev python-virtualenv</b> # for Python 3.n</pre> - 2. Create a virtualenv environment by issuing one of the following commands: + 2. Create a Virtualenv environment by issuing one of the following commands: <pre>$ <b>virtualenv --system-site-packages</b> <i>targetDirectory</i> # for Python 2.7 $ <b>virtualenv --system-site-packages -p python3</b> <i>targetDirectory</i> # for Python 3.n</pre> where <code><em>targetDirectory</em></code> specifies the top of the - virtualenv tree. Our instructions assume that + Virtualenv tree. Our instructions assume that <code><em>targetDirectory</em></code> is `~/tensorflow`, but you may choose any directory. - 3. Activate the virtualenv environment by issuing one of the following + 3. Activate the Virtualenv environment by issuing one of the following commands: <pre>$ <b>source ~/tensorflow/bin/activate</b> # bash, sh, ksh, or zsh @@ -160,18 +160,18 @@ Take the following steps to install TensorFlow with Virtualenv: <pre>(tensorflow)$ <b>easy_install -U pip</b></pre> 5. Issue one of the following commands to install TensorFlow in the active - virtualenv environment: + Virtualenv environment: <pre>(tensorflow)$ <b>pip install --upgrade tensorflow</b> # for Python 2.7 (tensorflow)$ <b>pip3 install --upgrade tensorflow</b> # for Python 3.n (tensorflow)$ <b>pip install --upgrade tensorflow-gpu</b> # for Python 2.7 and GPU (tensorflow)$ <b>pip3 install --upgrade tensorflow-gpu</b> # for Python 3.n and GPU</pre> - If the preceding command succeeds, skip Step 6. If the preceding + If the above command succeeds, skip Step 6. If the preceding command fails, perform Step 6. 6. (Optional) If Step 5 failed (typically because you invoked a pip version - lower than 8.1), install TensorFlow in the active virtualenv environment + lower than 8.1), install TensorFlow in the active Virtualenv environment by issuing a command of the following format: <pre>(tensorflow)$ <b>pip install --upgrade</b> <i>tfBinaryURL</i> # Python 2.7 @@ -185,10 +185,10 @@ Take the following steps to install TensorFlow with Virtualenv: [here](#the_url_of_the_tensorflow_python_package). For example, if you are installing TensorFlow for Linux, Python 3.4, and CPU-only support, issue the following command to install TensorFlow in the active - virtualenv environment: + Virtualenv environment: <pre>(tensorflow)$ <b>pip3 install --upgrade \ - https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc0-cp34-cp34m-linux_x86_64.whl</b></pre> + https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc1-cp34-cp34m-linux_x86_64.whl</b></pre> If you encounter installation problems, see [Common Installation Problems](#common_installation_problems). @@ -199,14 +199,14 @@ If you encounter installation problems, see After installing TensorFlow, [validate the installation](#ValidateYourInstallation). -Note that you must activate the virtualenv environment each time you -use TensorFlow. If the virtualenv environment is not currently active, +Note that you must activate the Virtualenv environment each time you +use TensorFlow. If the Virtualenv environment is not currently active, invoke one of the following commands: <pre> $ <b>source ~/tensorflow/bin/activate</b> # bash, sh, ksh, or zsh $ <b>source ~/tensorflow/bin/activate.csh</b> # csh or tcsh</pre> -When the virtualenv environment is active, you may run +When the Virtualenv environment is active, you may run TensorFlow programs from this shell. Your prompt will become the following to indicate that your tensorflow environment is active: @@ -293,7 +293,7 @@ take the following steps: <pre> $ <b>sudo pip3 install --upgrade \ - https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc0-cp34-cp34m-linux_x86_64.whl</b> + https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc1-cp34-cp34m-linux_x86_64.whl</b> </pre> If this step fails, see @@ -480,7 +480,7 @@ Take the following steps to install TensorFlow in an Anaconda environment: <pre> (tensorflow)$ <b>pip install --ignore-installed --upgrade \ - https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc0-cp34-cp34m-linux_x86_64.whl</b></pre> + https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc1-cp34-cp34m-linux_x86_64.whl</b></pre> <a name="ValidateYourInstallation"></a> @@ -494,11 +494,11 @@ To validate your TensorFlow installation, do the following: ### Prepare your environment -If you installed on native pip, virtualenv, or Anaconda, then +If you installed on native pip, Virtualenv, or Anaconda, then do the following: 1. Start a terminal. - 2. If you installed with virtualenv or Anaconda, activate your container. + 2. If you installed with Virtualenv or Anaconda, activate your container. 3. If you installed TensorFlow source code, navigate to any directory *except* one containing TensorFlow source code. @@ -648,14 +648,14 @@ This section documents the relevant values for Linux installations. CPU only: <pre> -https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc0-cp27-none-linux_x86_64.whl +https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc1-cp27-none-linux_x86_64.whl </pre> GPU support: <pre> -https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.4.0rc0-cp27-none-linux_x86_64.whl +https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.4.0rc1-cp27-none-linux_x86_64.whl </pre> Note that GPU support requires the NVIDIA hardware and software described in @@ -667,14 +667,14 @@ Note that GPU support requires the NVIDIA hardware and software described in CPU only: <pre> -https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc0-cp34-cp34m-linux_x86_64.whl +https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc1-cp34-cp34m-linux_x86_64.whl </pre> GPU support: <pre> -https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.4.0rc0-cp34-cp34m-linux_x86_64.whl +https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.4.0rc1-cp34-cp34m-linux_x86_64.whl </pre> Note that GPU support requires the NVIDIA hardware and software described in @@ -686,14 +686,14 @@ Note that GPU support requires the NVIDIA hardware and software described in CPU only: <pre> -https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc0-cp35-cp35m-linux_x86_64.whl +https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc1-cp35-cp35m-linux_x86_64.whl </pre> GPU support: <pre> -https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.4.0rc0-cp35-cp35m-linux_x86_64.whl +https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.4.0rc1-cp35-cp35m-linux_x86_64.whl </pre> @@ -705,14 +705,14 @@ Note that GPU support requires the NVIDIA hardware and software described in CPU only: <pre> -https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc0-cp36-cp36m-linux_x86_64.whl +https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.4.0rc1-cp36-cp36m-linux_x86_64.whl </pre> GPU support: <pre> -https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.4.0rc0-cp36-cp36m-linux_x86_64.whl +https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.4.0rc1-cp36-cp36m-linux_x86_64.whl </pre> diff --git a/tensorflow/docs_src/install/install_mac.md b/tensorflow/docs_src/install/install_mac.md index d799298b8b..9a95710bfa 100644 --- a/tensorflow/docs_src/install/install_mac.md +++ b/tensorflow/docs_src/install/install_mac.md @@ -13,21 +13,21 @@ Note: As of version 1.2, TensorFlow no longer provides GPU support on macOS. You must pick the mechanism by which you install TensorFlow. The supported choices are as follows: - * virtualenv + * Virtualenv * "native" pip * Docker * installing from sources, which is documented in [a separate guide](https://www.tensorflow.org/install/install_sources). -**We recommend the virtualenv installation.** +**We recommend the Virtualenv installation.** [Virtualenv](https://virtualenv.pypa.io/en/stable) is a virtual Python environment isolated from other Python development, incapable of interfering with or being affected by other Python programs -on the same machine. During the virtualenv installation process, +on the same machine. During the Virtualenv installation process, you will install not only TensorFlow but also all the packages that TensorFlow requires. (This is actually pretty easy.) To start working with TensorFlow, you simply need to "activate" the -virtual environment. All in all, virtualenv provides a safe and +virtual environment. All in all, Virtualenv provides a safe and reliable mechanism for installing and running TensorFlow. Native pip installs TensorFlow directly on your system without going through @@ -53,30 +53,30 @@ However, within Anaconda, we recommend installing TensorFlow with the That is, the TensorFlow team neither tests nor maintains the conda package. Use that package at your own risk. -## Installing with virtualenv +## Installing with Virtualenv Take the following steps to install TensorFlow with Virtualenv: 1. Start a terminal (a shell). You'll perform all subsequent steps in this shell. - 2. Install pip and virtualenv by issuing the following commands: + 2. Install pip and Virtualenv by issuing the following commands: <pre> $ <b>sudo easy_install pip</b> $ <b>pip install --upgrade virtualenv</b> </pre> - 3. Create a virtualenv environment by issuing a command of one + 3. Create a Virtualenv environment by issuing a command of one of the following formats: <pre> $ <b>virtualenv --system-site-packages</b> <i>targetDirectory</i> # for Python 2.7 $ <b>virtualenv --system-site-packages -p python3</b> <i>targetDirectory</i> # for Python 3.n </pre> - where <i>targetDirectory</i> identifies the top of the virtualenv tree. + where <i>targetDirectory</i> identifies the top of the Virtualenv tree. Our instructions assume that <i>targetDirectory</i> is `~/tensorflow`, but you may choose any directory. - 4. Activate the virtualenv environment by issuing one of the + 4. Activate the Virtualenv environment by issuing one of the following commands: <pre>$ <b>source ~/tensorflow/bin/activate</b> # If using bash, sh, ksh, or zsh @@ -98,7 +98,7 @@ Take the following steps to install TensorFlow with Virtualenv: 7. Optional. If Step 6 failed (typically because you invoked a pip version lower than 8.1), install TensorFlow in the active - virtualenv environment by issuing a command of the following format: + Virtualenv environment by issuing a command of the following format: <pre> $ <b>pip install --upgrade</b> <i>tfBinaryURL</i> # Python 2.7 $ <b>pip3 install --upgrade</b> <i>tfBinaryURL</i> # Python 3.n </pre> @@ -114,7 +114,7 @@ Take the following steps to install TensorFlow with Virtualenv: TensorFlow in the active Virtualenv is as follows: <pre> $ <b>pip3 install --upgrade \ - https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.4.0rc0-py2-none-any.whl</b></pre> + https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.4.0rc1-py2-none-any.whl</b></pre> If you encounter installation problems, see [Common Installation Problems](#common-installation-problems). @@ -126,8 +126,8 @@ After installing TensorFlow, [validate your installation](#ValidateYourInstallation) to confirm that the installation worked properly. -Note that you must activate the virtualenv environment each time you -use TensorFlow in a new shell. If the virtualenv environment is not +Note that you must activate the Virtualenv environment each time you +use TensorFlow in a new shell. If the Virtualenv environment is not currently active (that is, the prompt is not `(tensorflow)`, invoke one of the following commands: @@ -139,7 +139,7 @@ tensorflow environment is active: <pre> (tensorflow)$ </pre> -When the virtualenv environment is active, you may run +When the Virtualenv environment is active, you may run TensorFlow programs from this shell. When you are done using TensorFlow, you may deactivate the @@ -235,7 +235,7 @@ take the following steps: issue the following command: <pre> $ <b>sudo pip3 install --upgrade \ - https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.4.0rc0-py2-none-any.whl</b> </pre> + https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.4.0rc1-py2-none-any.whl</b> </pre> If the preceding command fails, see [installation problems](#common-installation-problems). @@ -344,7 +344,7 @@ Take the following steps to install TensorFlow in an Anaconda environment: TensorFlow for Python 2.7: <pre> (tensorflow)$ <b>pip install --ignore-installed --upgrade \ - https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.4.0rc0-py2-none-any.whl</b></pre> + https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.4.0rc1-py2-none-any.whl</b></pre> <a name="ValidateYourInstallation"></a> @@ -358,11 +358,11 @@ To validate your TensorFlow installation, do the following: ### Prepare your environment -If you installed on native pip, virtualenv, or Anaconda, then +If you installed on native pip, Virtualenv, or Anaconda, then do the following: 1. Start a terminal. - 2. If you installed with virtualenv or Anaconda, activate your container. + 2. If you installed with Virtualenv or Anaconda, activate your container. 3. If you installed TensorFlow source code, navigate to any directory *except* one containing TensorFlow source code. @@ -517,7 +517,7 @@ This section documents the relevant values for Mac OS installations. <pre> -https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.4.0rc0-py2-none-any.whl +https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.4.0rc1-py2-none-any.whl </pre> @@ -525,7 +525,7 @@ https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.4.0rc0-py2-none-a <pre> -https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.4.0rc0-py3-none-any.whl +https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.4.0rc1-py3-none-any.whl </pre> diff --git a/tensorflow/docs_src/install/install_sources.md b/tensorflow/docs_src/install/install_sources.md index 28bc5f5159..6d0dcdcd4a 100644 --- a/tensorflow/docs_src/install/install_sources.md +++ b/tensorflow/docs_src/install/install_sources.md @@ -355,10 +355,10 @@ Invoke `pip install` to install that pip package. The filename of the `.whl` file depends on your platform. For example, the following command will install the pip package -for TensorFlow 1.4.0rc0 on Linux: +for TensorFlow 1.4.0rc1 on Linux: <pre> -$ <b>sudo pip install /tmp/tensorflow_pkg/tensorflow-1.4.0rc0-py2-none-any.whl</b> +$ <b>sudo pip install /tmp/tensorflow_pkg/tensorflow-1.4.0rc1-py2-none-any.whl</b> </pre> ## Validate your installation @@ -447,8 +447,8 @@ Stack Overflow and specify the `tensorflow` tag. **Linux** <table> <tr><th>Version:</th><th>CPU/GPU:</th><th>Python Version:</th><th>Compiler:</th><th>Build Tools:</th><th>cuDNN:</th><th>CUDA:</th></tr> -<tr><td>tensorflow-1.4.0rc0</td><td>CPU</td><td>2.7, 3.3-3.6</td><td>GCC 4.8</td><td>Bazel 0.4.5</td><td>N/A</td><td>N/A</td></tr> -<tr><td>tensorflow_gpu-1.4.0rc0</td><td>GPU</td><td>2.7, 3.3-3.6</td><td>GCC 4.8</td><td>Bazel 0.4.5</td><td>6</td><td>8</td></tr> +<tr><td>tensorflow-1.4.0rc1</td><td>CPU</td><td>2.7, 3.3-3.6</td><td>GCC 4.8</td><td>Bazel 0.4.5</td><td>N/A</td><td>N/A</td></tr> +<tr><td>tensorflow_gpu-1.4.0rc1</td><td>GPU</td><td>2.7, 3.3-3.6</td><td>GCC 4.8</td><td>Bazel 0.4.5</td><td>6</td><td>8</td></tr> <tr><td>tensorflow-1.2.0</td><td>CPU</td><td>2.7, 3.3-3.6</td><td>GCC 4.8</td><td>Bazel 0.4.5</td><td>N/A</td><td>N/A</td></tr> <tr><td>tensorflow_gpu-1.2.0</td><td>GPU</td><td>2.7, 3.3-3.6</td><td>GCC 4.8</td><td>Bazel 0.4.5</td><td>5.1</td><td>8</td></tr> <tr><td>tensorflow-1.1.0</td><td>CPU</td><td>2.7, 3.3-3.6</td><td>GCC 4.8</td><td>Bazel 0.4.2</td><td>N/A</td><td>N/A</td></tr> @@ -460,19 +460,19 @@ Stack Overflow and specify the `tensorflow` tag. **Mac** <table> <tr><th>Version:</th><th>CPU/GPU:</th><th>Python Version:</th><th>Compiler:</th><th>Build Tools:</th><th>cuDNN:</th><th>CUDA:</th></tr> -<tr><td>tensorflow-1.4.0rc0</td><td>CPU</td><td>2.7, 3.3-3.6</td><td>Clang from xcode</td><td>Bazel 0.4.5</td><td>N/A</td><td>N/A</td></tr> -<tr><td>ttensorflow-1.2.0</td><td>CPU</td><td>2.7, 3.3-3.6</td><td>Clang from xcode</td><td>Bazel 0.4.5</td><td>N/A</td><td>N/A</td></tr> -<tr><td>ttensorflow-1.1.0</td><td>CPU</td><td>2.7, 3.3-3.6</td><td>Clang from xcode</td><td>Bazel 0.4.2</td><td>N/A</td><td>N/A</td></tr> -<tr><td>ttensorflow_gpu-1.1.0</td><td>GPU</td><td>2.7, 3.3-3.6</td><td>Clang from xcode</td><td>Bazel 0.4.2</td><td>5.1</td><td>8</td></tr> -<tr><td>ttensorflow-1.0.0</td><td>CPU</td><td>2.7, 3.3-3.6</td><td>Clang from xcode</td><td>Bazel 0.4.2</td><td>N/A</td><td>N/A</td></tr> -<tr><td>ttensorflow_gpu-1.0.0</td><td>GPU</td><td>2.7, 3.3-3.6</td><td>Clang from xcode</td><td>Bazel 0.4.2</td><td>5.1</td><td>8</td></tr> +<tr><td>tensorflow-1.4.0rc1</td><td>CPU</td><td>2.7, 3.3-3.6</td><td>Clang from xcode</td><td>Bazel 0.4.5</td><td>N/A</td><td>N/A</td></tr> +<tr><td>tensorflow-1.2.0</td><td>CPU</td><td>2.7, 3.3-3.6</td><td>Clang from xcode</td><td>Bazel 0.4.5</td><td>N/A</td><td>N/A</td></tr> +<tr><td>tensorflow-1.1.0</td><td>CPU</td><td>2.7, 3.3-3.6</td><td>Clang from xcode</td><td>Bazel 0.4.2</td><td>N/A</td><td>N/A</td></tr> +<tr><td>tensorflow_gpu-1.1.0</td><td>GPU</td><td>2.7, 3.3-3.6</td><td>Clang from xcode</td><td>Bazel 0.4.2</td><td>5.1</td><td>8</td></tr> +<tr><td>tensorflow-1.0.0</td><td>CPU</td><td>2.7, 3.3-3.6</td><td>Clang from xcode</td><td>Bazel 0.4.2</td><td>N/A</td><td>N/A</td></tr> +<tr><td>tensorflow_gpu-1.0.0</td><td>GPU</td><td>2.7, 3.3-3.6</td><td>Clang from xcode</td><td>Bazel 0.4.2</td><td>5.1</td><td>8</td></tr> </table> **Windows** <table> <tr><th>Version:</th><th>CPU/GPU:</th><th>Python Version:</th><th>Compiler:</th><th>Build Tools:</th><th>cuDNN:</th><th>CUDA:</th></tr> -<tr><td>tensorflow-1.4.0rc0</td><td>CPU</td><td>3.5-3.6</td><td>MSVC 2015 update 3</td><td>Cmake v3.6.3</td><td>N/A</td><td>N/A</td></tr> -<tr><td>tensorflow_gpu-1.4.0rc0</td><td>GPU</td><td>3.5-3.6</td><td>MSVC 2015 update 3</td><td>Cmake v3.6.3</td><td>6</td><td>8</td></tr> +<tr><td>tensorflow-1.4.0rc1</td><td>CPU</td><td>3.5-3.6</td><td>MSVC 2015 update 3</td><td>Cmake v3.6.3</td><td>N/A</td><td>N/A</td></tr> +<tr><td>tensorflow_gpu-1.4.0rc1</td><td>GPU</td><td>3.5-3.6</td><td>MSVC 2015 update 3</td><td>Cmake v3.6.3</td><td>6</td><td>8</td></tr> <tr><td>tensorflow-1.2.0</td><td>CPU</td><td>3.5-3.6</td><td>MSVC 2015 update 3</td><td>Cmake v3.6.3</td><td>N/A</td><td>N/A</td></tr> <tr><td>tensorflow_gpu-1.2.0</td><td>GPU</td><td>3.5-3.6</td><td>MSVC 2015 update 3</td><td>Cmake v3.6.3</td><td>5.1</td><td>8</td></tr> <tr><td>tensorflow-1.1.0</td><td>CPU</td><td>3.5</td><td>MSVC 2015 update 3</td><td>Cmake v3.6.3</td><td>N/A</td><td>N/A</td></tr> diff --git a/tensorflow/docs_src/programmers_guide/graphs.md b/tensorflow/docs_src/programmers_guide/graphs.md index 10f53fe8f2..984058297f 100644 --- a/tensorflow/docs_src/programmers_guide/graphs.md +++ b/tensorflow/docs_src/programmers_guide/graphs.md @@ -89,7 +89,7 @@ to all API functions in the same context. For example: * Executing `v = tf.Variable(0)` adds to the graph a @{tf.Operation} that will store a writeable tensor value that persists between @{tf.Session.run} calls. The @{tf.Variable} object wraps this operation, and can be used [like a - tensor](#tensor-like-objects), which will read the current value of the + tensor](#tensor-like_objects), which will read the current value of the stored value. The @{tf.Variable} object also has methods such as @{tf.Variable.assign$`assign`} and @{tf.Variable.assign_add$`assign_add`} that create @{tf.Operation} objects that, when executed, update the stored value. @@ -100,7 +100,7 @@ to all API functions in the same context. For example: when run, will apply those gradients to a set of variables. Most programs rely solely on the default graph. However, -see [Dealing with multiple graphs](#dealing-with-multiple-graphs) for more +see [Dealing with multiple graphs](#programming_with_multiple_graphs) for more advanced use cases. High-level APIs such as the @{tf.estimator.Estimator} API manage the default graph on your behalf, and--for example--may create different graphs for training and evaluation. @@ -329,7 +329,7 @@ described below. * **`graph`.** By default, a new @{tf.Session} will be bound to---and only able to run operations in---the current default graph. If you are using multiple graphs in your program (see [Programming with multiple - graphs](programming-with-multiple-graphs) for more details), you can specify + graphs](#programming_with_multiple_graphs) for more details), you can specify an explicit @{tf.Graph} when you construct the session. * **`config`.** This argument allows you to specify a @{tf.ConfigProto} that diff --git a/tensorflow/docs_src/programmers_guide/tensors.md b/tensorflow/docs_src/programmers_guide/tensors.md index cc4181e75e..d6f80430cd 100644 --- a/tensorflow/docs_src/programmers_guide/tensors.md +++ b/tensorflow/docs_src/programmers_guide/tensors.md @@ -197,7 +197,7 @@ For example, here is how to make a vector of zeros with the same size as the number of columns in a given matrix: ``` python -zeros = tf.zeros(tf.shape(my_matrix)[1]) +zeros = tf.zeros(my_matrix.shape[1]) ``` ### Changing the shape of a `tf.Tensor` diff --git a/tensorflow/examples/get_started/regression/test.py b/tensorflow/examples/get_started/regression/test.py index 652b44f543..0b1477ad96 100644 --- a/tensorflow/examples/get_started/regression/test.py +++ b/tensorflow/examples/get_started/regression/test.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================== -"""A simple smoke test that runs these examples for 1 training iteraton.""" +"""A simple smoke test that runs these examples for 1 training iteration.""" from __future__ import absolute_import from __future__ import division diff --git a/tensorflow/examples/tutorials/mnist/fully_connected_feed.py b/tensorflow/examples/tutorials/mnist/fully_connected_feed.py index af89c8c77b..35ca1b2f7f 100644 --- a/tensorflow/examples/tutorials/mnist/fully_connected_feed.py +++ b/tensorflow/examples/tutorials/mnist/fully_connected_feed.py @@ -109,7 +109,7 @@ def do_eval(sess, labels_placeholder) true_count += sess.run(eval_correct, feed_dict=feed_dict) precision = float(true_count) / num_examples - print(' Num examples: %d Num correct: %d Precision @ 1: %0.04f' % + print('Num examples: %d Num correct: %d Precision @ 1: %0.04f' % (num_examples, true_count, precision)) diff --git a/tensorflow/examples/tutorials/mnist/mnist_deep.py b/tensorflow/examples/tutorials/mnist/mnist_deep.py index 4b5b50400a..a4dbab5123 100644 --- a/tensorflow/examples/tutorials/mnist/mnist_deep.py +++ b/tensorflow/examples/tutorials/mnist/mnist_deep.py @@ -82,7 +82,7 @@ def deepnn(x): W_fc1 = weight_variable([7 * 7 * 64, 1024]) b_fc1 = bias_variable([1024]) - h_pool2_flat = tf.reshape(h_pool2, [-1, 7*7*64]) + h_pool2_flat = tf.reshape(h_pool2, [-1, 7 * 7 * 64]) h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1) + b_fc1) # Dropout - controls the complexity of the model, prevents co-adaptation of diff --git a/tensorflow/examples/udacity/Dockerfile b/tensorflow/examples/udacity/Dockerfile index 3d48ced41b..3ca58566c1 100644 --- a/tensorflow/examples/udacity/Dockerfile +++ b/tensorflow/examples/udacity/Dockerfile @@ -1,5 +1,5 @@ FROM gcr.io/tensorflow/tensorflow:latest -MAINTAINER Vincent Vanhoucke <vanhoucke@google.com> +LABEL maintainer="Vincent Vanhoucke <vanhoucke@google.com>" # Pillow needs libjpeg by default as of 3.0. RUN apt-get update && apt-get install -y --no-install-recommends \ diff --git a/tensorflow/python/__init__.py b/tensorflow/python/__init__.py index 8d9c5de9ad..af34aca3e3 100644 --- a/tensorflow/python/__init__.py +++ b/tensorflow/python/__init__.py @@ -262,6 +262,7 @@ _allowed_symbols.extend([ 'VERSION', 'GIT_VERSION', 'COMPILER_VERSION', + 'CXX11_ABI_FLAG', ]) # Remove all extra symbols that don't have a docstring or are not explicitly @@ -280,6 +281,7 @@ _exported_dunders = set([ '__version__', '__git_version__', '__compiler_version__', + '__cxx11_abi_flag__', ]) # Expose symbols minus dunders, unless they are whitelisted above. diff --git a/tensorflow/python/client/tf_session.i b/tensorflow/python/client/tf_session.i index a8d92a40a5..f45bc13602 100644 --- a/tensorflow/python/client/tf_session.i +++ b/tensorflow/python/client/tf_session.i @@ -45,6 +45,9 @@ tensorflow::ImportNumpy(); // Compiler %constant const char* __compiler_version__ = tf_compiler_version(); +// _GLIBCXX_USE_CXX11_ABI flag value +%constant const int __cxx11_abi_flag__ = tf_cxx11_abi_flag(); + // Release the Python GIL for the duration of most methods. %exception { Py_BEGIN_ALLOW_THREADS; diff --git a/tensorflow/python/estimator/run_config.py b/tensorflow/python/estimator/run_config.py index 372f01dc82..d71964d2ec 100644 --- a/tensorflow/python/estimator/run_config.py +++ b/tensorflow/python/estimator/run_config.py @@ -528,6 +528,7 @@ class RunConfig(object): """Returns a new instance of `RunConfig` replacing specified properties. Only the properties in the following list are allowed to be replaced: + - `model_dir`. - `tf_random_seed`, - `save_summary_steps`, diff --git a/tensorflow/python/framework/versions.py b/tensorflow/python/framework/versions.py index f4b01635dc..81529e2b1e 100644 --- a/tensorflow/python/framework/versions.py +++ b/tensorflow/python/framework/versions.py @@ -24,10 +24,12 @@ from tensorflow.python import pywrap_tensorflow __version__ = pywrap_tensorflow.__version__ __git_version__ = pywrap_tensorflow.__git_version__ __compiler_version__ = pywrap_tensorflow.__compiler_version__ +__cxx11_abi_flag__ = pywrap_tensorflow.__cxx11_abi_flag__ VERSION = __version__ GIT_VERSION = __git_version__ COMPILER_VERSION = __compiler_version__ +CXX11_ABI_FLAG = __cxx11_abi_flag__ GRAPH_DEF_VERSION = pywrap_tensorflow.GRAPH_DEF_VERSION GRAPH_DEF_VERSION_MIN_CONSUMER = ( @@ -39,7 +41,9 @@ __all__ = [ "__version__", "__git_version__", "__compiler_version__", + "__cxx11_abi_flag__", "COMPILER_VERSION", + "CXX11_ABI_FLAG", "GIT_VERSION", "GRAPH_DEF_VERSION", "GRAPH_DEF_VERSION_MIN_CONSUMER", diff --git a/tensorflow/python/kernel_tests/decode_csv_op_test.py b/tensorflow/python/kernel_tests/decode_csv_op_test.py index 7d9e57c8e5..fec52fa9cc 100644 --- a/tensorflow/python/kernel_tests/decode_csv_op_test.py +++ b/tensorflow/python/kernel_tests/decode_csv_op_test.py @@ -34,7 +34,7 @@ class DecodeCSVOpTest(test.TestCase): out = sess.run(decode) for i, field in enumerate(out): - if field.dtype == np.float32: + if field.dtype == np.float32 or field.dtype == np.float64: self.assertAllClose(field, expected_out[i]) else: self.assertAllEqual(field, expected_out[i]) @@ -85,6 +85,17 @@ class DecodeCSVOpTest(test.TestCase): self._test(args, expected_out) + def testDouble(self): + args = { + "records": ["1.0", "-1.79e+308", '"1.79e+308"'], + "record_defaults": [np.array( + [], dtype=np.double)], + } + + expected_out = [[1.0, -1.79e+308, 1.79e+308]] + + self._test(args, expected_out) + def testInt64(self): args = { "records": ["1", "2", '"2147483648"'], diff --git a/tensorflow/python/ops/array_ops.py b/tensorflow/python/ops/array_ops.py index 2133a00ff6..8f9828e8cf 100644 --- a/tensorflow/python/ops/array_ops.py +++ b/tensorflow/python/ops/array_ops.py @@ -336,8 +336,8 @@ def size(input, name=None, out_type=dtypes.int32): # pylint: disable=redefined-builtin """Returns the size of a tensor. - This operation returns an integer representing the number of elements in - `input`. + Returns a 0-D `Tensor` representing the number of elements in `input` + of type `out_type`. Defaults to tf.int32. For example: @@ -349,11 +349,15 @@ def size(input, name=None, out_type=dtypes.int32): Args: input: A `Tensor` or `SparseTensor`. name: A name for the operation (optional). - out_type: (Optional) The specified output type of the operation - (`int32` or `int64`). Defaults to tf.int32. + out_type: (Optional) The specified non-quantized numeric output type + of the operation. Defaults to `tf.int32`. Returns: - A `Tensor` of type `out_type`. Defaults to tf.int32. + A `Tensor` of type `out_type`. Defaults to `tf.int32`. + + @compatibility(numpy) + Equivalent to np.size() + @end_compatibility """ return size_internal(input, name, optimize=True, out_type=out_type) @@ -366,11 +370,11 @@ def size_internal(input, name=None, optimize=True, out_type=dtypes.int32): input: A `Tensor` or `SparseTensor`. name: A name for the operation (optional). optimize: if true, encode the size as a constant when possible. - out_type: (Optional) The specified output type of the operation - (`int32` or `int64`). Defaults to tf.int32. + out_type: (Optional) The specified non-quantized numeric output type + of the operation. Defaults to `tf.int32`. Returns: - A `Tensor` of type `out_type`. + A `Tensor` of type `out_type`. Defaults to `tf.int32`. """ with ops.name_scope(name, "Size", [input]) as name: if isinstance(input, (sparse_tensor.SparseTensor, diff --git a/tensorflow/python/ops/lookup_ops.py b/tensorflow/python/ops/lookup_ops.py index 7f00344be2..fa58ffc37e 100644 --- a/tensorflow/python/ops/lookup_ops.py +++ b/tensorflow/python/ops/lookup_ops.py @@ -1011,7 +1011,7 @@ def index_table_from_tensor(vocabulary_list, Args: vocabulary_list: A 1-D `Tensor` that specifies the mapping of keys to - indices. Thetype of this object must be castable to `dtype`. + indices. The type of this object must be castable to `dtype`. num_oov_buckets: The number of out-of-vocabulary buckets. default_value: The value to use for out-of-vocabulary feature values. Defaults to -1. diff --git a/tensorflow/python/ops/parsing_ops.py b/tensorflow/python/ops/parsing_ops.py index ea7132791c..14aef01dec 100644 --- a/tensorflow/python/ops/parsing_ops.py +++ b/tensorflow/python/ops/parsing_ops.py @@ -1183,7 +1183,7 @@ def decode_csv(records, record_defaults, field_delim=",", Each string is a record/row in the csv and all records should have the same format. record_defaults: A list of `Tensor` objects with specific types. - Acceptable types are `float32`, `int32`, `int64`, `string`. + Acceptable types are `float32`, `float64`, `int32`, `int64`, `string`. One tensor per column of the input record, with either a scalar default value for that column or empty if the column is required. field_delim: An optional `string`. Defaults to `","`. diff --git a/tensorflow/python/ops/variable_scope.py b/tensorflow/python/ops/variable_scope.py index 197e5abcc9..92fa928eed 100644 --- a/tensorflow/python/ops/variable_scope.py +++ b/tensorflow/python/ops/variable_scope.py @@ -590,7 +590,7 @@ class _VariableStore(object): if reuse is True: raise ValueError("PartitionedVariable %s does not exist, or was not " "created with tf.get_variable(). Did you mean to set " - "reuse=None in VarScope?" % name) + "reuse=False or reuse=tf.AUTO_REUSE in VarScope?" % name) slice_dim, slice_shape = _compute_slice_dim_and_shape( shape.as_list(), partitions) diff --git a/tensorflow/python/ops/variables.py b/tensorflow/python/ops/variables.py index fdd0666403..f906b7b3c4 100644 --- a/tensorflow/python/ops/variables.py +++ b/tensorflow/python/ops/variables.py @@ -197,6 +197,14 @@ class Variable(object): ValueError: If the initial value is not specified, or does not have a shape and `validate_shape` is `True`. RuntimeError: If eager execution is enabled. + + @compatibility(eager) + `tf.Variable` is not compatible with eager execution. Use + `tfe.Variable` instead which is compatable with both eager execution + and graph construction. See [the TensorFlow Eager Execution + guide](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/eager/python/g3doc/guide.md#variables-and-optimizers) + for details on how variables work in eager execution. + @end_compatibility """ if not context.in_graph_mode(): raise RuntimeError("tf.Variable not supported in Eager mode. " diff --git a/tensorflow/python/platform/sysconfig.py b/tensorflow/python/platform/sysconfig.py index 1e3be40933..57635fb4d9 100644 --- a/tensorflow/python/platform/sysconfig.py +++ b/tensorflow/python/platform/sysconfig.py @@ -17,6 +17,8 @@ @@get_include @@get_lib +@@get_compile_flags +@@get_link_flags """ from __future__ import absolute_import from __future__ import division @@ -24,6 +26,7 @@ from __future__ import print_function import os.path as _os_path +from tensorflow.python.framework.versions import CXX11_ABI_FLAG as _CXX11_ABI_FLAG from tensorflow.python.util.all_util import remove_undocumented @@ -51,5 +54,30 @@ def get_lib(): import tensorflow as tf return _os_path.join(_os_path.dirname(tf.__file__)) + +def get_compile_flags(): + """Get the compilation flags for custom operators. + + Returns: + The compilation flags. + """ + flags = [] + flags.append('-I%s' % get_include()) + flags.append('-I%s/external/nsync/public' % get_include()) + flags.append('-D_GLIBCXX_USE_CXX11_ABI=%d' % _CXX11_ABI_FLAG) + return flags + + +def get_link_flags(): + """Get the link flags for custom operators. + + Returns: + The link flags. + """ + flags = [] + flags.append('-L%s' % get_lib()) + flags.append('-ltensorflow_framework') + return flags + _allowed_symbols = [] remove_undocumented(__name__, _allowed_symbols) diff --git a/tensorflow/python/pywrap_tensorflow.py b/tensorflow/python/pywrap_tensorflow.py index 000ed8df8b..91373fa544 100644 --- a/tensorflow/python/pywrap_tensorflow.py +++ b/tensorflow/python/pywrap_tensorflow.py @@ -59,6 +59,7 @@ try: from tensorflow.python.pywrap_tensorflow_internal import __version__ from tensorflow.python.pywrap_tensorflow_internal import __git_version__ from tensorflow.python.pywrap_tensorflow_internal import __compiler_version__ + from tensorflow.python.pywrap_tensorflow_internal import __cxx11_abi_flag__ if _use_dlopen_global_flags: pywrap_dlopen_global_flags.reset_dlopen_flags() diff --git a/tensorflow/python/training/evaluation.py b/tensorflow/python/training/evaluation.py index fdcb9c2e90..b36444a14c 100644 --- a/tensorflow/python/training/evaluation.py +++ b/tensorflow/python/training/evaluation.py @@ -19,6 +19,7 @@ from __future__ import division from __future__ import print_function import time +import math from tensorflow.python.framework import dtypes from tensorflow.python.framework import ops @@ -91,6 +92,9 @@ class _StopAfterNEvalsHook(session_run_hook.SessionRunHook): self._num_evals = num_evals self._evals_completed = None self._log_progress = log_progress + # Reduce logging frequency if there are 20 or more evaluations. + self._log_frequency = (1 if (num_evals is None or num_evals < 20) + else math.floor(num_evals / 10.)) def _set_evals_completed_tensor(self, updated_eval_step): self._evals_completed = updated_eval_step @@ -106,7 +110,9 @@ class _StopAfterNEvalsHook(session_run_hook.SessionRunHook): if self._num_evals is None: logging.info('Evaluation [%d]', evals_completed) else: - logging.info('Evaluation [%d/%d]', evals_completed, self._num_evals) + if ((evals_completed % self._log_frequency) == 0 or + (self._num_evals == evals_completed)): + logging.info('Evaluation [%d/%d]', evals_completed, self._num_evals) if self._num_evals is not None and evals_completed >= self._num_evals: run_context.request_stop() diff --git a/tensorflow/stream_executor/cuda/cuda_diagnostics.cc b/tensorflow/stream_executor/cuda/cuda_diagnostics.cc index bf81b9c0ad..00506fa54b 100644 --- a/tensorflow/stream_executor/cuda/cuda_diagnostics.cc +++ b/tensorflow/stream_executor/cuda/cuda_diagnostics.cc @@ -76,10 +76,10 @@ string DriverVersionStatusToString(port::StatusOr<DriverVersion> version) { port::StatusOr<DriverVersion> StringToDriverVersion(const string &value) { std::vector<string> pieces = port::Split(value, '.'); - if (pieces.size() != 2 && pieces.size() != 3) { + if (pieces.size() < 2 || pieces.size() > 4) { return port::Status{ port::error::INVALID_ARGUMENT, - port::Printf("expected %%d.%%d or %%d.%%d.%%d form for driver version; got \"%s\"", + port::Printf("expected %%d.%%d, %%d.%%d.%%d, or %%d.%%d.%%d.%%d form for driver version; got \"%s\"", value.c_str())}; } diff --git a/tensorflow/tools/api/golden/tensorflow.pbtxt b/tensorflow/tools/api/golden/tensorflow.pbtxt index f61f82e43e..bf7bc6a7c1 100644 --- a/tensorflow/tools/api/golden/tensorflow.pbtxt +++ b/tensorflow/tools/api/golden/tensorflow.pbtxt @@ -17,6 +17,10 @@ tf_module { mtype: "<type \'str\'>" } member { + name: "CXX11_ABI_FLAG" + mtype: "<type \'int\'>" + } + member { name: "ConditionalAccumulator" mtype: "<type \'type\'>" } diff --git a/tensorflow/tools/api/golden/tensorflow.sysconfig.pbtxt b/tensorflow/tools/api/golden/tensorflow.sysconfig.pbtxt index 02dec04b9c..2f00aeac25 100644 --- a/tensorflow/tools/api/golden/tensorflow.sysconfig.pbtxt +++ b/tensorflow/tools/api/golden/tensorflow.sysconfig.pbtxt @@ -1,6 +1,10 @@ path: "tensorflow.sysconfig" tf_module { member_method { + name: "get_compile_flags" + argspec: "args=[], varargs=None, keywords=None, defaults=None" + } + member_method { name: "get_include" argspec: "args=[], varargs=None, keywords=None, defaults=None" } @@ -8,4 +12,8 @@ tf_module { name: "get_lib" argspec: "args=[], varargs=None, keywords=None, defaults=None" } + member_method { + name: "get_link_flags" + argspec: "args=[], varargs=None, keywords=None, defaults=None" + } } diff --git a/tensorflow/tools/ci_build/Dockerfile.android b/tensorflow/tools/ci_build/Dockerfile.android index facff47621..99a69d7b43 100644 --- a/tensorflow/tools/ci_build/Dockerfile.android +++ b/tensorflow/tools/ci_build/Dockerfile.android @@ -1,6 +1,6 @@ FROM ubuntu:14.04 -MAINTAINER Jan Prach <jendap@google.com> +LABEL maintainer="Jan Prach <jendap@google.com>" # Copy and run the install scripts. COPY install/*.sh /install/ diff --git a/tensorflow/tools/ci_build/Dockerfile.cmake b/tensorflow/tools/ci_build/Dockerfile.cmake index 9013dc012d..37ba24d65a 100644 --- a/tensorflow/tools/ci_build/Dockerfile.cmake +++ b/tensorflow/tools/ci_build/Dockerfile.cmake @@ -14,7 +14,7 @@ # ============================================================================== FROM ubuntu:16.04 -MAINTAINER Shanqing Cai <cais@google.com> +LABEL maintainer="Shanqing Cai <cais@google.com>" # Copy and run the install scripts. COPY install/*.sh /install/ diff --git a/tensorflow/tools/ci_build/Dockerfile.cpu b/tensorflow/tools/ci_build/Dockerfile.cpu index 206108930a..57a854a9df 100644 --- a/tensorflow/tools/ci_build/Dockerfile.cpu +++ b/tensorflow/tools/ci_build/Dockerfile.cpu @@ -1,6 +1,6 @@ FROM ubuntu:14.04 -MAINTAINER Jan Prach <jendap@google.com> +LABEL maintainer="Jan Prach <jendap@google.com>" # Copy and run the install scripts. COPY install/*.sh /install/ diff --git a/tensorflow/tools/ci_build/Dockerfile.debian.jessie.cpu b/tensorflow/tools/ci_build/Dockerfile.debian.jessie.cpu index b914f51918..eb9d0d4dd0 100644 --- a/tensorflow/tools/ci_build/Dockerfile.debian.jessie.cpu +++ b/tensorflow/tools/ci_build/Dockerfile.debian.jessie.cpu @@ -1,6 +1,6 @@ FROM debian:jessie -MAINTAINER Jan Prach <jendap@google.com> +LABEL maintainer="Jan Prach <jendap@google.com>" # Copy and run the install scripts. COPY install/*.sh /install/ diff --git a/tensorflow/tools/ci_build/Dockerfile.gpu b/tensorflow/tools/ci_build/Dockerfile.gpu index 5d18295f68..2d46ccb6b1 100644 --- a/tensorflow/tools/ci_build/Dockerfile.gpu +++ b/tensorflow/tools/ci_build/Dockerfile.gpu @@ -1,6 +1,6 @@ FROM nvidia/cuda:8.0-cudnn6-devel-ubuntu14.04 -MAINTAINER Jan Prach <jendap@google.com> +LABEL maintainer="Jan Prach <jendap@google.com>" # In the Ubuntu 14.04 images, cudnn is placed in system paths. Move them to # /usr/local/cuda diff --git a/tensorflow/tools/ci_build/Dockerfile.gpu_clang b/tensorflow/tools/ci_build/Dockerfile.gpu_clang index c4342d17f5..0ecd8c75e0 100644 --- a/tensorflow/tools/ci_build/Dockerfile.gpu_clang +++ b/tensorflow/tools/ci_build/Dockerfile.gpu_clang @@ -1,6 +1,6 @@ FROM nvidia/cuda:8.0-cudnn6-devel-ubuntu14.04 -MAINTAINER Ilya Biryukov <ibiryukov@google.com> +LABEL maintainer="Ilya Biryukov <ibiryukov@google.com>" # In the Ubuntu 14.04 images, cudnn is placed in system paths. Move them to # /usr/local/cuda diff --git a/tensorflow/tools/ci_build/Dockerfile.hadoop b/tensorflow/tools/ci_build/Dockerfile.hadoop index 489493c26e..6010aedb33 100644 --- a/tensorflow/tools/ci_build/Dockerfile.hadoop +++ b/tensorflow/tools/ci_build/Dockerfile.hadoop @@ -1,6 +1,6 @@ FROM ubuntu:14.04 -MAINTAINER Jonathan Hseu <jhseu@google.com> +LABEL maintainer="Jonathan Hseu <jhseu@google.com>" # Copy and run the install scripts. COPY install/*.sh /install/ diff --git a/tensorflow/tools/ci_build/Dockerfile.pi b/tensorflow/tools/ci_build/Dockerfile.pi index 2fddd6a2c0..75ef30d32b 100644 --- a/tensorflow/tools/ci_build/Dockerfile.pi +++ b/tensorflow/tools/ci_build/Dockerfile.pi @@ -1,6 +1,6 @@ FROM ubuntu:14.04 -MAINTAINER Jan Prach <jendap@google.com> +LABEL maintainer="Jan Prach <jendap@google.com>" # Copy and run the install scripts. COPY install/*.sh /install/ diff --git a/tensorflow/tools/ci_build/Dockerfile.pi-python3 b/tensorflow/tools/ci_build/Dockerfile.pi-python3 index 18b131ea19..b1c648ba30 100644 --- a/tensorflow/tools/ci_build/Dockerfile.pi-python3 +++ b/tensorflow/tools/ci_build/Dockerfile.pi-python3 @@ -1,6 +1,6 @@ FROM ubuntu:14.04 -MAINTAINER Jan Prach <jendap@google.com> +LABEL maintainer="Jan Prach <jendap@google.com>" # Copy and run the install scripts. COPY install/*.sh /install/ diff --git a/tensorflow/tools/ci_build/README.md b/tensorflow/tools/ci_build/README.md index acef833909..202fcb9101 100644 --- a/tensorflow/tools/ci_build/README.md +++ b/tensorflow/tools/ci_build/README.md @@ -50,7 +50,7 @@ and tests. Click on **Details** to see the results from Jenkins or the internal CI system. Results from Jenkins are displayed in the Jenkins UI. For more information, -see the [Jenkns documentation](https://jenkins.io/doc/). +see the [Jenkins documentation](https://jenkins.io/doc/). Results from the internal CI system are displayed in the Build Status UI. In this UI, to see the logs for a failed build: diff --git a/tensorflow/tools/ci_build/ci_sanity.sh b/tensorflow/tools/ci_build/ci_sanity.sh index 26053de4e9..f1c207f9b6 100755 --- a/tensorflow/tools/ci_build/ci_sanity.sh +++ b/tensorflow/tools/ci_build/ci_sanity.sh @@ -426,6 +426,72 @@ do_code_link_check() { tensorflow/tools/ci_build/code_link_check.sh } +# List .h|.cc files changed in the last non-merge git commit that still exist, +# i.e., not removed. +# Usage: get_clang_files_to_check [--incremental] +get_clang_files_to_check() { + if [[ "$1" == "--incremental" ]]; then + CHANGED_CLANG_FILES=$(get_changed_files_in_last_non_merge_git_commit | \ + grep '.*\.h$\|.*\.cc$') + + # Do not include files removed in the last non-merge commit. + CLANG_FILES="" + for CLANG_FILE in ${CHANGED_CLANG_FILES}; do + if [[ -f "${CLANG_FILE}" ]]; then + CLANG_FILES="${CLANG_FILES} ${CLANG_FILE}" + fi + done + + echo "${CLANG_FILES}" + else + find tensorflow -name '*.h' -o -name '*.cc' + fi +} + +do_clang_format_check() { + if [[ $# != "0" ]] && [[ $# != "1" ]]; then + echo "Invalid syntax when invoking do_clang_format_check" + echo "Usage: do_clang_format_check [--incremental]" + return 1 + fi + + if [[ "$1" == "--incremental" ]]; then + CLANG_SRC_FILES=$(get_clang_files_to_check --incremental) + + if [[ -z "${CLANG_SRC_FILES}" ]]; then + echo "do_clang_format_check will NOT run due to --incremental flag and "\ +"due to the absence of .h or .cc code changes in the last commit." + return 0 + fi + elif [[ -z "$1" ]]; then + # TODO (yongtang): Always pass --incremental until all files have + # been sanitized gradually. Then this --incremental could be removed. + CLANG_SRC_FILES=$(get_clang_files_to_check --incremental) + else + echo "Invalid syntax for invoking do_clang_format_check" + echo "Usage: do_clang_format_check [--incremental]" + return 1 + fi + + CLANG_FORMAT=${CLANG_FORMAT:-clang-format-3.8} + + success=1 + for filename in $CLANG_SRC_FILES; do + $CLANG_FORMAT --style=google $filename | diff $filename - > /dev/null + if [ ! $? -eq 0 ]; then + success=0 + echo File $filename is not properly formatted with "clang-format "\ +"--style=google" + fi + done + + if [ $success == 0 ]; then + echo Clang format check fails. + exit 1 + fi + echo Clang format check success. +} + do_check_load_py_test() { BUILD_CMD="bazel build ${BAZEL_FLAGS} //tensorflow/tools/pip_package:check_load_py_test" ${BUILD_CMD} diff --git a/tensorflow/tools/ci_build/install/install_deb_packages.sh b/tensorflow/tools/ci_build/install/install_deb_packages.sh index da1f2199d0..4ab307c925 100755 --- a/tensorflow/tools/ci_build/install/install_deb_packages.sh +++ b/tensorflow/tools/ci_build/install/install_deb_packages.sh @@ -28,6 +28,7 @@ if [[ "$1" != "" ]] && [[ "$1" != "--without_cmake" ]]; then fi # Install dependencies from ubuntu deb repository. +apt-key adv --keyserver keyserver.ubuntu.com --recv 084ECFC5828AB726 apt-get update if [[ "$ubuntu_version" == "14" ]]; then @@ -41,6 +42,7 @@ apt-get install -y --no-install-recommends \ autoconf \ automake \ build-essential \ + clang-format-3.8 \ curl \ ffmpeg \ git \ diff --git a/tensorflow/tools/ci_build/update_version.py b/tensorflow/tools/ci_build/update_version.py index c7841f35aa..d2a63e5d66 100755 --- a/tensorflow/tools/ci_build/update_version.py +++ b/tensorflow/tools/ci_build/update_version.py @@ -17,7 +17,7 @@ # Automatically update TensorFlow version in source files # # Usage: -# ./tensorflow/tools/ci_build/update_version.py --version 1.4.0-rc0 +# ./tensorflow/tools/ci_build/update_version.py --version 1.4.0-rc1 # ./tensorflow/tools/ci_build/update_version.py --nightly # """Update version of TensorFlow script.""" diff --git a/tensorflow/tools/dist_test/Dockerfile b/tensorflow/tools/dist_test/Dockerfile index cd64e2c518..2a7605bbc9 100644 --- a/tensorflow/tools/dist_test/Dockerfile +++ b/tensorflow/tools/dist_test/Dockerfile @@ -20,7 +20,7 @@ FROM ubuntu:16.04 -MAINTAINER Shanqing Cai <cais@google.com> +LABEL maintainer="Shanqing Cai <cais@google.com>" RUN apt-get update RUN apt-get install -y \ diff --git a/tensorflow/tools/dist_test/Dockerfile.local b/tensorflow/tools/dist_test/Dockerfile.local index 7a896ab611..795aeee1b5 100644 --- a/tensorflow/tools/dist_test/Dockerfile.local +++ b/tensorflow/tools/dist_test/Dockerfile.local @@ -19,7 +19,7 @@ FROM ubuntu:16.04 -MAINTAINER Shanqing Cai <cais@google.com> +LABEL maintainer="Shanqing Cai <cais@google.com>" # Pick up some TF dependencies. RUN apt-get update && apt-get install -y \ diff --git a/tensorflow/tools/dist_test/local/Dockerfile b/tensorflow/tools/dist_test/local/Dockerfile index 96846f6564..383c3c2f4c 100644 --- a/tensorflow/tools/dist_test/local/Dockerfile +++ b/tensorflow/tools/dist_test/local/Dockerfile @@ -1,6 +1,6 @@ FROM jpetazzo/dind -MAINTAINER Shanqing Cai <cais@google.com> +LABEL maintainer="Shanqing Cai <cais@google.com>" RUN apt-get update diff --git a/tensorflow/tools/dist_test/server/Dockerfile b/tensorflow/tools/dist_test/server/Dockerfile index fabc8a7105..1359428f11 100644 --- a/tensorflow/tools/dist_test/server/Dockerfile +++ b/tensorflow/tools/dist_test/server/Dockerfile @@ -19,7 +19,7 @@ FROM ubuntu:16.04 -MAINTAINER Shanqing Cai <cais@google.com> +LABEL maintainer="Shanqing Cai <cais@google.com>" # Pick up some TF dependencies RUN apt-get update && apt-get install -y \ diff --git a/tensorflow/tools/dist_test/server/Dockerfile.test b/tensorflow/tools/dist_test/server/Dockerfile.test index 908af8af9b..ce7e783a1a 100644 --- a/tensorflow/tools/dist_test/server/Dockerfile.test +++ b/tensorflow/tools/dist_test/server/Dockerfile.test @@ -19,7 +19,7 @@ FROM ubuntu:16.04 -MAINTAINER Shanqing Cai <cais@google.com> +LABEL maintainer="Shanqing Cai <cais@google.com>" # Pick up some TF dependencies RUN apt-get update && apt-get install -y \ diff --git a/tensorflow/tools/docker/Dockerfile.devel-gpu-cuda9-cudnn7 b/tensorflow/tools/docker/Dockerfile.devel-gpu-cuda9-cudnn7 index 4558bc5293..64ebc4607a 100644 --- a/tensorflow/tools/docker/Dockerfile.devel-gpu-cuda9-cudnn7 +++ b/tensorflow/tools/docker/Dockerfile.devel-gpu-cuda9-cudnn7 @@ -1,6 +1,6 @@ FROM nvidia/cuda:9.0-cudnn7-devel-ubuntu16.04 -MAINTAINER Gunhan Gulsoy <gunan@google.com> +LABEL maintainer="Gunhan Gulsoy <gunan@google.com>" # It is possible to override these for releases. ARG TF_BRANCH=master diff --git a/tensorflow/tools/gcs_test/Dockerfile b/tensorflow/tools/gcs_test/Dockerfile index 5af753226f..69b554047b 100644 --- a/tensorflow/tools/gcs_test/Dockerfile +++ b/tensorflow/tools/gcs_test/Dockerfile @@ -1,6 +1,6 @@ FROM ubuntu:16.04 -MAINTAINER Shanqing Cai <cais@google.com> +LABEL maintainer="Shanqing Cai <cais@google.com>" RUN apt-get update RUN apt-get install -y \ diff --git a/tensorflow/tools/git/gen_git_source.py b/tensorflow/tools/git/gen_git_source.py index a7f8b5bb5f..0307d2a0eb 100755 --- a/tensorflow/tools/git/gen_git_source.py +++ b/tensorflow/tools/git/gen_git_source.py @@ -170,8 +170,16 @@ def write_version_info(filename, git_version): if b"\"" in git_version or b"\\" in git_version: git_version = "git_version_is_invalid" # do not cause build to fail! contents = """/* Generated by gen_git_source.py */ +#include <string> const char* tf_git_version() {return "%s";} const char* tf_compiler_version() {return __VERSION__;} +const int tf_cxx11_abi_flag() { +#ifdef _GLIBCXX_USE_CXX11_ABI + return _GLIBCXX_USE_CXX11_ABI; +#else + return 0; +#endif +} """ % git_version open(filename, "w").write(contents) diff --git a/tensorflow/tools/git/gen_git_source.sh b/tensorflow/tools/git/gen_git_source.sh index 977fe16333..788f9e6e57 100755 --- a/tensorflow/tools/git/gen_git_source.sh +++ b/tensorflow/tools/git/gen_git_source.sh @@ -26,7 +26,15 @@ if [[ $? != 0 ]]; then fi cat <<EOF > ${OUTPUT_FILENAME} +#include <string> const char* tf_git_version() {return "${GIT_VERSION}";} const char* tf_compiler_version() {return __VERSION__;} +const int tf_cxx11_abi_flag() { +#ifdef _GLIBCXX_USE_CXX11_ABI + return _GLIBCXX_USE_CXX11_ABI; +#else + return 0; +#endif +} EOF diff --git a/tensorflow/tools/pip_package/setup.py b/tensorflow/tools/pip_package/setup.py index 4f0de8f768..071b3a2a18 100644 --- a/tensorflow/tools/pip_package/setup.py +++ b/tensorflow/tools/pip_package/setup.py @@ -29,7 +29,7 @@ from setuptools.dist import Distribution # This version string is semver compatible, but incompatible with pip. # For pip, we will remove all '-' characters from this string, and use the # result for pip. -_VERSION = '1.4.0-rc0' +_VERSION = '1.4.0-rc1' REQUIRED_PACKAGES = [ 'enum34 >= 1.1.6', diff --git a/third_party/aws.BUILD b/third_party/aws.BUILD index 9d8e7946cd..bc6a2fd8cc 100644 --- a/third_party/aws.BUILD +++ b/third_party/aws.BUILD @@ -18,6 +18,9 @@ cc_library( "@%ws%//tensorflow:darwin": glob([ "aws-cpp-sdk-core/source/platform/linux-shared/*.cpp", ]), + "@%ws%//tensorflow:linux_ppc64le": glob([ + "aws-cpp-sdk-core/source/platform/linux-shared/*.cpp", + ]), "//conditions:default": [], }) + glob([ "aws-cpp-sdk-core/include/**/*.h", @@ -57,6 +60,11 @@ cc_library( "ENABLE_CURL_CLIENT", "ENABLE_NO_ENCRYPTION", ], + "@%ws%//tensorflow:linux_ppc64le": [ + "PLATFORM_LINUX", + "ENABLE_CURL_CLIENT", + "ENABLE_NO_ENCRYPTION", + ], "//conditions:default": [], }), includes = [ |