* Avoid CPU migrations by pinning your benchmarks to specific CPU cores. On Linux you can use `taskset`.
* Fix the CPU frequency to avoid Turbo Boost from kicking in and skewing your results. On Linux you can use `cpufreq-set` and the
  `performance` CPU governor.
* Vary the problem input size with `@Param`.

if [[ "$TFCI_WHL_NUMPY_VERSION" == 1 ]]; then
  # Uninstall tf nightly wheel built with numpy1.
  "$python" -m pip uninstall -y tf_nightly_numpy1
  # Install tf nightly cpu wheel built with numpy2.x from PyPI in numpy1.x env.
  "$python" -m pip install tf-nightly-cpu
  if [[ "$TFCI_WHL_IMPORT_TEST_ENABLE" == "1" ]]; then
    "$python" -c 'import tensorflow as tf; t1=tf.constant([1,2,3,4]); t2=tf.constant([5,6,7,8]); print(tf.add(t1,t2).shape)'

TFCI_BAZEL_TARGET_SELECTING_CONFIG_PREFIX=linux_cuda
TFCI_BUILD_PIP_PACKAGE_WHEEL_NAME_ARG="--repo_env=WHEEL_NAME=tensorflow"
TFCI_DOCKER_ARGS="--gpus all"
TFCI_LIB_SUFFIX="-gpu-linux-x86_64"
# TODO: Set back to 610M once the wheel size is fixed.

TFCI_DOCKER_IMAGE=us-docker.pkg.dev/ml-oss-artifacts-published/ml-public-container/ml-build:latest
TFCI_DOCKER_PULL_ENABLE=1
TFCI_DOCKER_REBUILD_ARGS="--target=devel ci/official/containers/ml_build"
TFCI_INDEX_HTML_ENABLE=1
TFCI_LIB_SUFFIX="-cpu-linux-x86_64"
TFCI_OUTPUT_DIR=build_output
TFCI_WHL_AUDIT_ENABLE=1
TFCI_WHL_AUDIT_PLAT=manylinux_2_27_x86_64
TFCI_WHL_BAZEL_TEST_ENABLE=1
TFCI_WHL_SIZE_LIMIT=300M

```bash
tensorflow/tools/ci_build/ci_build.sh CPU tensorflow/tools/ci_build/ci_sanity.sh
```

This will catch most license, Python coding style and BUILD file issues that
may exist in your changes.

#### Running unit tests

There are two ways to run TensorFlow unit tests.

1.  Using tools and libraries installed directly on your system.

    Refer to the

* Use local disk for ConfigMap volume instead of tmpfs ([#25306](https://github.com/kubernetes/kubernetes/pull/25306), [@pmorie](https://github.com/pmorie))
* Alpha support for scheduling pods on machines with NVIDIA GPUs whose kubelets use the `--experimental-nvidia-gpus` flag, using the alpha.kubernetes.io/nvidia-gpu resource  ([#24836](https://github.com/kubernetes/kubernetes/pull/24836), [@therc](https://github.com/therc))

TFCI_BAZEL_TARGET_SELECTING_CONFIG_PREFIX=macos_arm64
TFCI_BUILD_PIP_PACKAGE_WHEEL_NAME_ARG="--repo_env=WHEEL_NAME=tensorflow"
TFCI_INDEX_HTML_ENABLE=1
TFCI_LIB_SUFFIX="-cpu-darwin-arm64"
TFCI_MACOS_BAZEL_TEST_DIR_ENABLE=1
TFCI_MACOS_BAZEL_TEST_DIR_PATH="/Volumes/BuildData/bazel_output"
TFCI_OUTPUT_DIR=build_output
TFCI_WHL_BAZEL_TEST_ENABLE=1
TFCI_WHL_SIZE_LIMIT=245M

    public boolean hasPrimaryKey() {
        return false;
    }

    @Override
    public boolean hasCompoundPrimaryKey() {
        return false;
    }

    @Override
    protected UniqueInfo cpui() {
        return null;
    }

    // ===================================================================================
    //                                                                           Type Name

streaming compression due to its stability and performance.

This algorithm is specifically optimized for machine generated content.
Write throughput is typically at least 500MB/s per CPU core,
and scales with the number of available CPU cores.
Decompression speed is typically at least 1GB/s.

This means that in cases where raw IO is below these numbers
compression will not only reduce disk usage but also help increase system throughput.

mkdir -p ~/.gradle/init.d && cp -v $WORKSPACE/.ci/init.gradle ~/.gradle/init.d
if [ "$(uname -m)" = "arm64" ] || [ "$(uname -m)" = "aarch64" ]; then
   MAX_WORKERS=16
elif [ -f /proc/cpuinfo ]; then
   MAX_WORKERS=`grep '^cpu\scores' /proc/cpuinfo  | uniq | sed 's/\s\+//g' |  cut -d':' -f 2`
else
   if [[ "$OSTYPE" == "darwin"* ]]; then
      MAX_WORKERS=`sysctl -n hw.physicalcpu | sed 's/\s\+//g'`
   else
      echo "Unsupported OS Type: $OSTYPE"