Merge remote-tracking branch 'upstream/master' into upstream

.bazelrc

@@ -134,7 +134,7 @@ build --experimental_link_static_libraries_once=false
 # Prevent regressions on those two incompatible changes
 # TODO: remove those flags when they are flipped in the default Bazel version TF uses.
 build --incompatible_enforce_config_setting_visibility
-# TODO: also enable this flag after fixing the visbility violations
+# TODO: also enable this flag after fixing the visibility violations
 # build --incompatible_config_setting_private_default_visibility
 
 # Default options should come above this line.
@@ -249,7 +249,7 @@ build:cuda_clang --repo_env=TF_CUDA_COMPUTE_CAPABILITIES="sm_50,sm_60,sm_70,sm_8
 build:cuda_clang_official --config=cuda_clang
 build:cuda_clang_official --action_env=TF_CUDA_VERSION="12"
 build:cuda_clang_official --action_env=TF_CUDNN_VERSION="8"
-build:cuda_clang_official --action_env=CUDA_TOOLKIT_PATH="/usr/local/cuda-12.2"
+build:cuda_clang_official --action_env=CUDA_TOOLKIT_PATH="/usr/local/cuda-12.3"
 build:cuda_clang_official --action_env=GCC_HOST_COMPILER_PATH="/dt9/usr/bin/gcc"
 build:cuda_clang_official --action_env=CLANG_CUDA_COMPILER_PATH="/usr/lib/llvm-17/bin/clang"
 build:cuda_clang_official --action_env=LD_LIBRARY_PATH="/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64"
@@ -307,7 +307,7 @@ build:macos --copt=-w
 build:windows --copt=/W0
 build:windows --host_copt=/W0
 
-# Suppress most C++ complier warnings to reduce log size but allow
+# Suppress most C++ compiler warnings to reduce log size but allow
 # for specific warnings to still be present.
 build:linux --copt="-Wno-all"
 build:linux --copt="-Wno-extra"
@@ -588,7 +588,7 @@ test:release_base --test_size_filters=small,medium
 # Target the AVX instruction set
 build:release_linux_base --config=avx_linux
 
-# Disable clang extention that rejects type definitions within offsetof.
+# Disable clang extension that rejects type definitions within offsetof.
 # This was added in clang-16 by https://reviews.llvm.org/D133574.
 # Can be removed once upb is updated, since a type definition is used within
 # offset of in the current version of ubp.
@@ -791,6 +791,14 @@ test:macos_arm64_pycpp_test_filters --test_tag_filters=-no_oss,-oss_excluded,-os
 test:macos_arm64_pycpp_test_filters --build_tag_filters=-no_oss,-oss_excluded,-oss_serial,-no_oss_py39,-no_oss_py310,-nomac,-no_mac,-mac_excluded,-v1only,-gpu,-tpu,-benchmark-test,-no_mac_arm64,-no_aarch64
 test:macos_arm64_pycpp_test_filters --test_lang_filters=cc,py --test_size_filters=small,medium
 test:macos_arm64_pycpp_test --config=macos_arm64_pycpp_test_filters -- //tensorflow/... -//tensorflow/python/integration_testing/... -//tensorflow/compiler/tf2tensorrt/... -//tensorflow/core/tpu/... -//tensorflow/lite/... -//tensorflow/tools/toolchains/... -//tensorflow/go/... -//tensorflow/java/... -//tensorflow/compiler/aot/... -//tensorflow/core/kernels/image:resize_bicubic_op_test
+# MACOS X86 PYCPP
+test:macos_x86_pycpp_test_filters --test_tag_filters=-no_oss,-oss_excluded,-oss_serial,-no_oss_py38,-no_oss_py39,-no_oss_py310,-nomac,-no_mac,-mac_excluded,-v1only,-gpu,-tpu,-benchmark-test
+test:macos_x86_pycpp_test_filters --build_tag_filters=-no_oss,-oss_excluded,-oss_serial,-no_oss_py38,-no_oss_py39,-no_oss_py310,-nomac,-no_mac,-mac_excluded,-v1only,-gpu,-tpu,-benchmark-test
+test:macos_x86_pycpp_test_filters --keep_going --test_lang_filters=cc,py --test_size_filters=small,medium
+test:macos_x86_pycpp_test --config=macos_x86_pycpp_test_filters -- //tensorflow/... -//tensorflow/compiler/tf2tensorrt/... -//tensorflow/core/tpu/... -//tensorflow/go/... -//tensorflow/java/... -//tensorflow/python/integration_testing/... -//tensorflow/tools/toolchains/... -//tensorflow/lite/... -//tensorflow/compiler/aot/...
+# CROSS-COMPILE MACOS X86 PYCPP
+test:cross_compile_macos_x86_pycpp_test --config=macos_x86_pycpp_test
+test:cross_compile_macos_x86_pycpp_test -//tensorflow/core/kernels:quantized_conv_ops_test -//tensorflow/core/kernels:quantized_matmul_op_test -//tensorflow/python/ops:quantized_conv_ops_test -//tensorflow/tools/graph_transforms:transforms_test -//tensorflow/python/tools:aot_compiled_test
 # END TF TEST SUITE OPTIONS
 
 # START CROSS-COMPILE CONFIGS
@@ -836,7 +844,7 @@ build:cross_compile_macos_x86 --action_env  MACOSX_DEPLOYMENT_TARGET=10.15
 build:cross_compile_macos_x86 --platforms=//tensorflow/tools/toolchains/cross_compile/config:darwin_x86_64
 build:cross_compile_macos_x86 --cpu=darwin
 build:cross_compile_macos_x86 --crosstool_top=//tensorflow/tools/toolchains/cross_compile/cc:cross_compile_toolchain_suite
-# When RBE cross-compiling for macOS, we need to explictly register the
+# When RBE cross-compiling for macOS, we need to explicitly register the
 # toolchain. Otherwise, oddly, RBE complains that a "docker container must be
 # specified".
 build:cross_compile_macos_x86 --extra_toolchains=//tensorflow/tools/toolchains/cross_compile/config:macos-x86-cross-compile-cc-toolchain
@@ -848,5 +856,7 @@ build:cross_compile_macos_x86 --platform_mappings=tensorflow/tools/toolchains/cr
 build:rbe_cross_compile_macos_x86 --config=cross_compile_macos_x86
 build:rbe_cross_compile_macos_x86 --config=rbe_cross_compile_base
 test:rbe_cross_compile_macos_x86 --config=rbe_cross_compile_base
+# Increase the test timeout as tests often take longer on mac.
+test:rbe_cross_compile_macos_x86 --test_timeout=300,450,1200,3600
 # END MACOS CROSS-COMPILE CONFIGS
 # END CROSS-COMPILE CONFIGS

CONTRIBUTING.md

@@ -44,7 +44,7 @@ Before sending your pull requests, make sure you do the following:
 -   If a change is needed, the contributor is requested to make the suggested
     change.
 -   You make the change and submit it for the review again.
--   This cycle repeats itself untill the PR gets approved.
+-   This cycle repeats itself until the PR gets approved.
 -   Note: As a friendly reminder, we may reach out to you if the PR is awaiting
     your response for more than 2 weeks.
 

ci/official/README.md

@@ -15,40 +15,72 @@ from the `envs` directory that is filled with configuration options:
 
 -   Nightly jobs (Run nightly on the `nightly` branch)
     -   Uses `wheel.sh`, `libtensorflow.sh`, `code_check_full.sh`
-    -   Uses `envs/nightly_...`
 -   Continuous jobs (Run on every GitHub commit)
     -   Uses `pycpp.sh`
-    -   Uses `envs/continuous_...`
 -   Presubmit jobs (Run on every GitHub PR)
     -   Uses `pycpp.sh`, `code_check_changed_files.sh`
-    -   Also uses `envs/continuous_...`
 
 These "env" files match up with an environment matrix that roughly covers:
 
+-   Different Python versions
 -   Linux, MacOS, and Windows machines (these pool definitions are internal)
 -   x86 and arm64
 -   CPU-only, or with NVIDIA CUDA support (Linux only), or with TPUs
--   Different Python versions
 
 ## How to Test Your Changes to TensorFlow
 
-**Important for Googlers only**: Google employees can use an internal-only tool
-to make Kokoro test changes without having to run locally. Search for "MLCI"
-internally.
+You may check how your changes will affect TensorFlow by:
+
+1. Creating a PR and observing the presubmit test results
+2. Running the CI scripts locally, as explained below
+3. **Google employees only**: Google employees can use an internal-only tool
+called "MLCI" that makes testing more convenient: it can execute any full CI job
+against a pending change. Search for "MLCI" internally to find it.
+
+You may invoke a CI script of your choice by following these instructions:
 
 ```bash
-# Required: create a file to hold your settings
 cd tensorflow-git-dir
-export TFCI=$(mktemp)
 
-# Required: choose an environment (env) to copy.
+# Here is a single-line example of running a script on Linux to build the
+# GPU version of TensorFlow for Python 3.12, using the public TF bazel cache and
+# a local build cache:
+TFCI=py312,linux_x86_cuda,multicache ci/official/wheel.sh
+
+# First, set your TFCI variable to choose the environment settings.
+#   TFCI is a comma-separated list of filenames from the envs directory, which
+#   are all settings for the scripts. TF's CI jobs are all made of a combination
+#   of these env files.
+#
 #   If you've clicked on a test result from our CI (via a dashboard or GitHub link),
 #   click to "Invocation Details" and find BUILD_CONFIG, which will contain a TFCI
 #   value in the "env_vars" list that you can choose to copy that environment.
-echo >>$TFCI source ci/official/envs/nightly_linux_x86_cpu_py311
-
-# Required: Reset settings for local execution
-echo >>$TFCI source ci/official/envs/local_default
+#      Ex. 1: TFCI=py311,linux_x86_cuda,nightly_upload  (nightly job)
+#      Ex. 2: TFCI=py39,linux_x86,rbe                   (continuous job)
+#   Non-Googlers should replace "nightly_upload" or "rbe" with "multicache".
+#   Googlers should replace "nightly_upload" with "multicache" or "rbe", if
+#     you have set up your system to use RBE (see further below).
+#
+# Here is how to choose your TFCI value:
+# 1. A Python version must come first, because other scripts reference it.
+#      Ex. py39  -- Python 3.9
+#      Ex. py310 -- Python 3.10
+#      Ex. py311 -- Python 3.11
+#      Ex. py312 -- Python 3.12
+# 2. Choose the platform, which corresponds to the version of TensorFlow to
+#    build. This should also match the system you're using--you cannot build
+#    the TF MacOS package from Linux.
+#      Ex. linux_x86        -- x86_64 Linux platform
+#      Ex. linux_x86_cuda   -- x86_64 Linux platform, with Nvidia CUDA support
+#      Ex. macos_arm64      -- arm64 MacOS platform
+# 3. Add modifiers. Some modifiers for local execution are:
+#      Ex. multicache -- Use a local cache combined with TF's public cache
+#      Ex. rbe        -- Use RBE for faster builds (Googlers only; see below)
+#      Ex. no_docker  -- Disable docker on enabled platforms
+#    See full examples below for more details on these. Some other modifiers are:
+#      Ex. versions_upload -- for TF official release versions
+#      Ex. nightly_upload -- for TF nightly official builds; changes version numbers
+#      Ex. no_upload      -- Disable all uploads, usually for temporary CI issues
 
 # Recommended: use a local+remote cache.
 #
@@ -62,19 +94,7 @@ echo >>$TFCI source ci/official/envs/local_default
 #      or tests passing incorrectly.
 #    - Automatic LLVM updates are known to extend build time even with
 #      the cache; this is unavoidable.
-echo >>$TFCI source ci/official/envs/local_multicache
-
-# Advanced: Use Remote Build Execution (RBE) (internal developers only)
-#
-#   RBE dramatically speeds up builds and testing. It also gives you a
-#   public URL to share your build results with collaborators. However,
-#   it is only available to a limited set of internal TensorFlow developers.
-#
-#   RBE is incompatible with local caching, so you must remove
-#   ci/official/envs/local_multicache from your $TFCI file.
-#
-# To use RBE, you must first run `gcloud auth application-default login`, then:
-# echo >>$TFCI source ci/official/envs/local_rbe
+export TFCI=py311,linux_x86,multicache
 
 # Recommended: Configure Docker. (Linux only)
 #
@@ -98,7 +118,19 @@ echo >>$TFCI source ci/official/envs/local_multicache
 #   to transfer ownership to your user.
 #
 # Docker is enabled by default on Linux. You may disable it if you prefer:
-# echo >>$TFCI source ci/official/envs/local_nodocker
+# export TFCI=py311,linux_x86,no_docker
+
+# Advanced: Use Remote Build Execution (RBE) (internal developers only)
+#
+#   RBE dramatically speeds up builds and testing. It also gives you a
+#   public URL to share your build results with collaborators. However,
+#   it is only available to a limited set of internal TensorFlow developers.
+#
+#   RBE is incompatible with local caching, so you must remove
+#   ci/official/envs/local_multicache from your $TFCI file.
+#
+# To use RBE, you must first run `gcloud auth application-default login`, then:
+export TFCI=py311,linux_x86,rbe
 
 # Finally: Run your script of choice.
 #   If you've clicked on a test result from our CI (via a dashboard or GitHub link),
@@ -127,9 +159,9 @@ The top-level scripts and utility scripts should be fairly well-documented. Here
 is a brief explanation of how they tie together:
 
 1.  `envs/*` are lists of variables made with bash syntax. A user must set a
-    `TFCI` env param pointing to one of the `env` files.
+    `TFCI` env param pointing to a list of `env` files.
 2.  `utilities/setup.sh`, initialized by all top-level scripts, reads and sets
-    values from that `TFCI` path.
+    values from those `TFCI` paths.
     -   `set -a` / `set -o allexport` exports the variables from `env` files so
         all scripts can use them.
     -   `utilities/setup_docker.sh` creates a container called `tf` with all

ci/official/any.sh

@@ -38,18 +38,17 @@
 #       ...
 set -euxo pipefail
 cd "$(dirname "$0")/../../"  # tensorflow/
-REQUESTED_TFCI="$TFCI"
-export TFCI=$(mktemp)
-echo >>$TFCI "source $REQUESTED_TFCI"
-echo >>$TFCI "source ci/official/envs/disable_all_uploads"
+# Any request that includes "nightly_upload" should just use the
+# local multi-cache instead.
+export TFCI="$(echo $TFCI | sed 's/,nightly_upload/,multicache/')"
 if [[ -n "${TF_ANY_EXTRA_ENV:-}" ]]; then
-  echo >>$TFCI "source $TF_ANY_EXTRA_ENV"
+  export TFCI="$TFCI,$TF_ANY_EXTRA_ENV"
 fi
 if [[ -n "${TF_ANY_SCRIPT:-}" ]]; then
   "$TF_ANY_SCRIPT"
 elif [[ -n "${TF_ANY_TARGETS:-}" ]]; then
   source "${BASH_SOURCE%/*}/utilities/setup.sh"
-  tfrun bazel $TFCI_BAZEL_BAZELRC_ARGS "${TF_ANY_MODE:-test}" $TFCI_BAZEL_COMMON_ARGS $TF_ANY_TARGETS
+  tfrun bazel "${TF_ANY_MODE:-test}" $TFCI_BAZEL_COMMON_ARGS $TF_ANY_TARGETS
 else
   echo 'Looks like $TF_ANY_TARGETS are $TF_ANY_SCRIPT are both empty. That is an error.'
   exit 1

ci/official/bisect.sh

@@ -34,10 +34,6 @@
 #   export TF_ANY_MODE=test
 set -euxo pipefail
 cd "$(dirname "$0")/../../"  # tensorflow/
-REQUESTED_TFCI="$TFCI"
-export TFCI=$(mktemp)
-echo >>$TFCI "source $REQUESTED_TFCI"
-echo >>$TFCI "source ci/official/envs/disable_all_uploads"
-echo >>$TFCI "source ci/official/envs/local_multicache"
+export TFCI="$(echo $TFCI | sed 's/,nightly_upload/,multicache/')"
 git bisect start "$TF_BISECT_BAD" "$TF_BISECT_GOOD"
 git bisect run $TF_BISECT_SCRIPT

ci/official/containers/linux_arm64/Dockerfile

@@ -23,7 +23,7 @@ COPY apt.conf /etc/apt/
 RUN /build_patchelf.sh
 
 ################################################################################
-FROM nvidia/cuda:12.2.0-devel-ubuntu20.04 as devel
+FROM nvidia/cuda:12.3.1-devel-ubuntu20.04 as devel
 ################################################################################
 COPY --from=builder /dt10 /dt10
 COPY --from=builder /patchelf/patchelf_0.14.3-1_arm64.deb /patchelf/patchelf_0.14.3-1_arm64.deb

ci/official/containers/linux_arm64/cuda.packages.txt

@@ -1,3 +1,8 @@
 # CuDNN: https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html#ubuntu-network-installation
-libcudnn8=8.9.4.25-1+cuda12.2
-libcudnn8-dev=8.9.4.25-1+cuda12.2
+libcudnn8=8.9.6.50-1+cuda12.2
+libcudnn8-dev=8.9.6.50-1+cuda12.2
+
+# This can be removed once NVIDIA publishes a cuda-12.3.2 Docker image.
+# For now it ensures that we install at least version 12.3.107 of PTXAS,
+# since 12.3.103 has a bug.
+cuda-compiler-12-3=12.3.2-1

ci/official/containers/linux_arm64/devel.usertools/wheel_verification.bats

@@ -71,9 +71,3 @@ teardown_file() {
     source /tf/venv/bin/activate
     python3 -c 'import sys; import tensorflow as tf; sys.exit(0 if "keras" in tf.keras.__name__ else 1)'
 }
-
-# Is this still useful?
-@test "TensorFlow has Estimator" {
-    source /tf/venv/bin/activate
-    python3 -c 'import sys; import tensorflow as tf; sys.exit(0 if "_v2.estimator" in tf.estimator.__name__ else 1)'
-}

ci/official/envs/ci_default

@@ -1,3 +1,17 @@
+# Copyright 2023 The TensorFlow Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
 # Note: this file gets sourced in utilities/setup.sh, which has "set -u"
 # (error on undefined variables). This ensures that (a) every TFCI variable
 # has an explicit default value, and (b) no script can accidentally use a
@@ -9,41 +23,45 @@
 # You may also get an overview, e.g.:
 #   cd ci/official
 #   grep -o '^TFCI\w*' envs/ci_default | xargs -n 1 -I{} bash -c "echo; echo {}; grep -R -H --exclude-dir=envs --color=always '{}'"
-
+TFCI_ARTIFACT_FINAL_GCS_ENABLE=
+TFCI_ARTIFACT_FINAL_GCS_SA_PATH=
+TFCI_ARTIFACT_FINAL_GCS_URI=
+TFCI_ARTIFACT_FINAL_PYPI_ARGS=
+TFCI_ARTIFACT_FINAL_PYPI_ENABLE=
+TFCI_ARTIFACT_LATEST_GCS_URI=
+TFCI_ARTIFACT_STAGING_GCS_ENABLE=
+TFCI_ARTIFACT_STAGING_GCS_URI=
 TFCI_BAZEL_BAZELRC_ARGS=
 TFCI_BAZEL_COMMON_ARGS=
 TFCI_BAZEL_TARGET_SELECTING_CONFIG_PREFIX=
 TFCI_BUILD_PIP_PACKAGE_ARGS=
 TFCI_DOCKER_ARGS=
-TFCI_DOCKER_ENABLE=1
+TFCI_DOCKER_ENABLE=
 TFCI_DOCKER_IMAGE=
-TFCI_DOCKER_PULL_ENABLE=1
+TFCI_DOCKER_PULL_ENABLE=
 TFCI_DOCKER_REBUILD_ARGS=
-TFCI_DOCKER_REBUILD_ENABLE=0
-TFCI_DOCKER_REBUILD_UPLOAD_ENABLE=0
-TFCI_INDEX_HTML_ENABLE=1
+TFCI_DOCKER_REBUILD_ENABLE=
+TFCI_DOCKER_REBUILD_UPLOAD_ENABLE=
+TFCI_GIT_DIR=
+TFCI_INDEX_HTML_ENABLE=
 TFCI_LIB_SUFFIX=
 TFCI_MACOS_BAZEL_TEST_DIR_ENABLE=
 TFCI_MACOS_BAZEL_TEST_DIR_PATH=
+TFCI_MACOS_CROSS_COMPILE_ENABLE=
+TFCI_MACOS_CROSS_COMPILE_SDK_DEST=
+TFCI_MACOS_CROSS_COMPILE_SDK_SOURCE=
 TFCI_MACOS_INSTALL_BAZELISK_ENABLE=
 TFCI_MACOS_INSTALL_BAZELISK_URL=
 TFCI_MACOS_PYENV_INSTALL_ENABLE=
+TFCI_MACOS_TWINE_INSTALL_ENABLE=
 TFCI_MACOS_UPGRADE_PYENV_ENABLE=
 TFCI_NIGHTLY_UPDATE_VERSION_ENABLE=
 TFCI_NVIDIA_SMI_ENABLE=
-TFCI_OUTPUT_DIR=build_output
+TFCI_OUTPUT_DIR=
 TFCI_PYTHON_VERIFY_PIP_INSTALL_ARGS=
 TFCI_PYTHON_VERSION=
-TFCI_UPLOAD_LIB_ENABLE=
-TFCI_UPLOAD_LIB_LATEST_ENABLE=
-TFCI_UPLOAD_LIB_LATEST_URI=
-TFCI_UPLOAD_LIB_URI=
-TFCI_UPLOAD_WHL_GCS_ENABLE=
-TFCI_UPLOAD_WHL_GCS_URI=
-TFCI_UPLOAD_WHL_PYPI_ARGS=
-TFCI_UPLOAD_WHL_PYPI_ENABLE=
-TFCI_WHL_AUDIT_ENABLE=1
+TFCI_WHL_AUDIT_ENABLE=
 TFCI_WHL_AUDIT_PLAT=
-TFCI_WHL_BAZEL_TEST_ENABLE=1
+TFCI_WHL_BAZEL_TEST_ENABLE=
 TFCI_WHL_SIZE_LIMIT=
-TFCI_WHL_SIZE_LIMIT_ENABLE=1
+TFCI_WHL_SIZE_LIMIT_ENABLE=