let summary = "Rewrite graph for embedding pipelining";
  let constructor = "TFDevice::CreateEmbeddingPipeliningPass()";
    let description = [{
    For architectures that support accelerated embedding lookups, this pass will
    rewrite the graph to use pipelining for better device utilization.
  }];
}

def EmbeddingSequencingPass : Pass<"tf-embedding-sequencing", "mlir::ModuleOp"> {

#include "tensorflow/core/common_runtime/local_device.h"
#include "tensorflow/core/framework/device_base.h"

namespace tensorflow {

// tensorflow::PjRtBaseDevice replaces the deprecated tensorflow::XlaDevice.
// This accelerator agnostic device is mainly used to store metadata.
class PjRtBaseDevice : public LocalDevice {
 public:
  // Stores metadata about the PjRtBaseDevice.
  class Metadata {
   public:

        uses: actions/setup-java@v4
        with:
          distribution: 'zulu'
          java-version: 17

      - name: Enable KVM group perms
        # https://github.blog/changelog/2023-02-23-hardware-accelerated-android-virtualization-on-actions-windows-and-linux-larger-hosted-runners/
        run: |
          echo 'KERNEL=="kvm", GROUP="kvm", MODE="0666", OPTIONS+="static_node=kvm"' | sudo tee /etc/udev/rules.d/99-kvm4all.rules

# is a request for 2 SCSI formatted and mounted SSDs and 1 NVMe block device SSD.
NODE_LOCAL_SSDS_EXT=${NODE_LOCAL_SSDS_EXT:-}
# Accelerators to be attached to each node. Format "type=<accelerator-type>,count=<accelerator-count>"
# More information on available GPUs here - https://cloud.google.com/compute/docs/gpus/
NODE_ACCELERATORS=${NODE_ACCELERATORS:-""}
export REGISTER_MASTER_KUBELET=${REGISTER_MASTER:-true}

- Introduced a new alpha feature gate, `SELinuxMount`, which can now be enabled to accelerate SELinux relabeling. ([#123157](https://github.com/kubernetes/kubernetes/pull/123157), [@jsafrane](https://github.com/jsafrane))

  let summary = "Lifting resource operations out of device computation";
  let description = [{
    This pass lifts resource variable operations outside of device computation.
    This is useful because a lot of accelerator devices can not interact with
    resource variables directly..

    Here is a simple example in TensorFlow where a device doubles the value of a
    TensorFlow resource variable and returns new value:

    ```mlir

        "//learning/brain/mlir/...",
        "//third_party/iree/...",
        "//third_party/odml/infra/...",
        "//tensorflow/compiler/mlir/...",
        "//tensorflow/lite/python/...",
        "//waymo/accelerator/alpine/tools/...",
        "//waymo/ml/compiler/mlir/...",
        # Allow visibility from the mlir language server.
        "//learning/brain/mlir/mlir_lsp_server/...",

fi

if [[ -n "${NODE_ACCELERATORS}" ]]; then
    if [[ "${NODE_ACCELERATORS}" =~ .*type=([a-zA-Z0-9-]+).* ]]; then
        NON_MASTER_NODE_LABELS="${NON_MASTER_NODE_LABELS},cloud.google.com/gke-accelerator=${BASH_REMATCH[1]}"
    fi
fi

# List of the set of feature gates recognized by the GCP CCM

        "//third_party/py/guesslang/...",
        "//third_party/py/keras/...",
        "//third_party/py/tf_keras/...",
        "//third_party/yggdrasil_decision_forests/...",
        "//waymo/accelerator/...",
        "//waymo/ml/cn/...",
        "//waymo/ml/models/...",
    ],
)

package_group(
    name = "ndarray_tensor_allow_list",
    packages = [
        "//third_party/py/courier/...",

    *   Fix GPU delegate crash with C++17.
    *   Add 5D support to TFLite `strided_slice`.
    *   Fix error in delegation of `DEPTH_TO_SPACE` to `NNAPI` causing op not to
        be accelerated.
    *   Fix segmentation fault when running a model with LSTM nodes using
        `NNAPI` Delegate
    *   Fix `NNAPI` delegate failure when an operand for Maximum/Minimum
        operation is a scalar.