Merged commit includes the following changes:

504684855  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Remove legacy references from `ops.py`.

    This is done to eventually remove the lazy loads in `indexed_slices.py`.

--
504682660  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Add missing reference symbol.

--
504682467  by A. Unique TensorFlower<gardener@tensorflow.org>:

    #tf-data-service Add a function to parse split file names.

--
504680829  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Update XNNPACK README to document feature parity for FP16 operators

--
504677404  by A. Unique TensorFlower<gardener@tensorflow.org>:
    Automated rollback of changelist 504103068.

504676634  by A. Unique TensorFlower<gardener@tensorflow.org>:
    Automated rollback of changelist 504656381.

504672100  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Delete compiler/xla/stream_executor/lib/initialize.h, update users to depend on stream_executor/platform/intialize.h directly

--
504669543  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Refactor transforms to make buildozer object injectable.

--
504667726  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Update all CI dependencies to non vulnerable versions.

    This MUST land before branch cut.

--
504661218  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Update TFRT dependency to use revision
    http://github.com/tensorflow/runtime/commit/c653281a1a23c0c3d41536a983c7d10fcc5b1fbf.

--
504660212  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Change TraceMeArg to keep a reference to an AlphaNum constructed on the stack rather than constructing it by value.

    The eliminates a lifetime safety bug where AlphaNum uses a default constructed argument as a string buffer that was a dangling reference before.

    Also add a LIFETIME_BOUND attribute to ensure this is used correctly.

--
504660069  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Add kIsOpenSource to tsl/platform/test.h

--
504659330  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Update usages of tensorflow::Status::SetPayload.

--
504656381  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Remove the scrubbing of constants in tpu_cluster_formation.

    Instead, just ignore tf.Const attributes, in CollectAndGroupClusterOps. This
    is functionally equivalent (for this pass), but a bit more clean since it
    doesn't leave modified tf.Const ops behind.

--
504654874  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Add missing dependency for the Mac build.

--
504653641  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Internal visibility change

--
504648821  by A. Unique TensorFlower<gardener@tensorflow.org>:
    Automated rollback of changelist 504006190.
BEGIN_PUBLIC
Rollback of PR #59418: [aarch64] enable acl linking for cpu xla for mkl_aarch64_threadpool config
Reason: The configuration is still not very well-tested.
END_PUBLIC

504645490  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Fix input validation for GRU ops.

--
504633571  by A. Unique TensorFlower<gardener@tensorflow.org>:

    [SE] Replace uses of stream_executor::port::error with tsl::error.

    This lets us remove stream_executor/lib/error.h!

    No functional change; these are aliases of one another.

--
504633188  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Update tf.tpu.experimental.initialize_tpu_system to disable running on eager mode.

--
504618714  by A. Unique TensorFlower<gardener@tensorflow.org>:

    [XLA:GPU] Update hlo graph dumper to additionally dump the gemm epilogue from the GemmBackendConfig

--
504618661  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Fix ODR violation from zen_util.h

--
504616623  by A. Unique TensorFlower<gardener@tensorflow.org>:

    [tf.data] Collect the iterator's memory usage metrics for /tfdataz page.

--
504612989  by A. Unique TensorFlower<gardener@tensorflow.org>:

    Internal change

PiperOrigin-RevId: 504684855

.bazelrc

@@ -242,7 +242,6 @@ build:mkl_aarch64 -c opt
 # Config setting to build oneDNN with Compute Library for the Arm Architecture (ACL).
 # with Eigen threadpool support
 build:mkl_aarch64_threadpool --define=build_with_mkl_aarch64=true
-build:mkl_aarch64_threadpool --define=build_with_acl=true
 build:mkl_aarch64_threadpool -c opt
 
 # This config refers to building CUDA op kernels with nvcc.

tensorflow/compiler/jit/BUILD

@@ -411,6 +411,7 @@ cc_library(
         ":internal",
         # We reuse VariableInfo in TFRT's implementation of TpuExecuteOp.
         "//learning/brain/tfrt/tf_tpu:__pkg__",
+        "//learning/brain/tfrt/tpu_plugin:__pkg__",
         "//learning/brain/tfrt/tpu_common:__pkg__",
         "//tensorflow/core/common_runtime/next_pluggable_device:__pkg__",
     ],

tensorflow/compiler/xla/backends/interpreter/BUILD

@@ -138,7 +138,7 @@ cc_library(
         ":executor",
         ":platform_id",
         "//tensorflow/compiler/xla/stream_executor",
-        "//tensorflow/compiler/xla/stream_executor/lib",
+        "//tensorflow/compiler/xla/stream_executor/platform",
         "//tensorflow/tsl/platform:status",
         "@com_google_absl//absl/strings:str_format",
     ],

tensorflow/compiler/xla/backends/interpreter/platform.cc

@@ -21,9 +21,9 @@ limitations under the License.
 #include "absl/strings/str_format.h"
 #include "tensorflow/compiler/xla/backends/interpreter/executor.h"
 #include "tensorflow/compiler/xla/stream_executor/device_options.h"
-#include "tensorflow/compiler/xla/stream_executor/lib/initialize.h"
 #include "tensorflow/compiler/xla/stream_executor/multi_platform_manager.h"
 #include "tensorflow/compiler/xla/stream_executor/platform.h"
+#include "tensorflow/compiler/xla/stream_executor/platform/initialize.h"
 #include "tensorflow/tsl/platform/status.h"
 
 namespace stream_executor {

tensorflow/compiler/xla/hlo/evaluator/hlo_evaluator.cc

@@ -222,7 +222,7 @@ Status MakeEvalErrorDueToParamOrInfeed(const HloInstruction& eval_instruction) {
   absl::little_endian::Store32(
       const_cast<char*>(error_payload.data()),
       static_cast<uint32_t>(EvalErrorDetail::kDynamicValueDependence));
-  error.SetPayload(kEvalErrorDetailUrl, error_payload);
+  error.SetPayload(kEvalErrorDetailUrl, absl::Cord(error_payload));
   return error;
 }
 

tensorflow/compiler/xla/mlir/tools/mlir_replay/BUILD

@@ -9,6 +9,7 @@ xla_cc_binary(
         "//tensorflow/compiler/xla:debug_options_flags",
         "//tensorflow/compiler/xla/mlir/runtime/ir:rt",
         "//tensorflow/compiler/xla/mlir/tools/mlir_replay/public:compiler_trace_proto_cc",
+        "//tensorflow/compiler/xla/mlir/tools/mlir_replay/public:compiler_trace_proto_cc_impl",
         "//tensorflow/compiler/xla/mlir/tools/mlir_replay/public:execution_trace_proto_cc",
         "//tensorflow/compiler/xla/mlir_hlo:gml_st",
         "//tensorflow/compiler/xla/mlir_hlo:hlo_dialect_registration",

tensorflow/compiler/xla/service/hlo_graph_dumper.cc

@@ -1251,6 +1251,10 @@ ExtractGemmBackendConfigProps(const gpu::GemmBackendConfig& config,
   if (config.algorithm_case() == gpu::GemmBackendConfig::kSelectedAlgorithm) {
     props.emplace_back("algorithm", StrCat(config.selected_algorithm()));
   }
+  if (config.epilogue() != gpu::GemmBackendConfig::DEFAULT) {
+    props.emplace_back(
+        "epilogue", gpu::GemmBackendConfig::Epilogue_Name(config.epilogue()));
+  }
   return props;
 }
 

tensorflow/compiler/xla/stream_executor/BUILD

@@ -239,6 +239,7 @@ cc_library(
         ":stream_executor_headers",
         "//tensorflow/compiler/xla/stream_executor/lib",
         "//tensorflow/compiler/xla/stream_executor/platform",
+        "//tensorflow/tsl/platform:errors",
         "//tensorflow/tsl/platform:status",
         "//tensorflow/tsl/platform:statusor",
         "@com_google_absl//absl/strings",
@@ -450,6 +451,7 @@ tsl_gpu_library(
         "//tensorflow/compiler/xla/stream_executor/lib",
         "//tensorflow/compiler/xla/stream_executor/platform",
         "//tensorflow/tsl/platform:env",
+        "//tensorflow/tsl/platform:errors",
         "//tensorflow/tsl/platform:logging",
         "//tensorflow/tsl/platform:stacktrace",
         "//tensorflow/tsl/platform:status",
@@ -661,7 +663,7 @@ cc_library(
         ":platform",
         ":plugin",
         ":stream_executor_headers",
-        "//tensorflow/compiler/xla/stream_executor/lib",
+        "//tensorflow/tsl/platform:errors",
         "//tensorflow/tsl/platform:status",
         "//tensorflow/tsl/platform:statusor",
         "@com_google_absl//absl/base:core_headers",

tensorflow/compiler/xla/stream_executor/cuda/cuda_blas.cc

@@ -63,7 +63,7 @@ limitations under the License.
 #include "tensorflow/compiler/xla/stream_executor/gpu/gpu_stream.h"
 #include "tensorflow/compiler/xla/stream_executor/gpu/gpu_timer.h"
 #include "tensorflow/compiler/xla/stream_executor/gpu/gpu_types.h"
-#include "tensorflow/compiler/xla/stream_executor/lib/initialize.h"
+#include "tensorflow/compiler/xla/stream_executor/platform/initialize.h"
 #include "tensorflow/compiler/xla/stream_executor/platform/logging.h"
 #include "tensorflow/compiler/xla/stream_executor/platform/port.h"
 #include "tensorflow/compiler/xla/stream_executor/plugin_registry.h"