"SparseSoftmaxCrossEntropyWithLogits", "LogSoftmax"}},
          {"REDUCEWINDOWPW", {"BiasAddGrad", "LRN", "LRNGrad"}},
          {"BN",
           {"FusedBatchNorm", "FusedBatchNormV2", "FusedBatchNormV3",
            "_FusedBatchNormEx", "FusedBatchNormGrad", "FusedBatchNormGradV2",
            "FusedBatchNormGradV3"}},
          {"Conv", {"_FusedConv2D"}},

func.func @testFusedBatchNormToBatchNormV3(%arg0: tensor<8x8x8x8xf32>, %arg1: tensor<8xf32>, %arg2: tensor<8xf32>, %arg3: tensor<8xf32>, %arg4: tensor<8xf32>) -> (tensor<8x8x8x8xf32>) {
  // CHECK: "tf.FusedBatchNormV3"
  // CHECK-SAME: device = "/job:localhost/replica:0/task:0/device:GPU:0"

        support generators and Sequences.
*   `tf.lite`
    *   Legalization for `NMS` ops in TFLite.
    *   add `narrow_range` and `axis` to `quantize_v2` and `dequantize` ops.
    *   Added support for `FusedBatchNormV3` in converter.
    *   Add an `errno`-like field to `NNAPI` delegate for detecting `NNAPI`
        errors for fallback behaviour.
    *   Refactors `NNAPI` Delegate to support detailed reason why an operation

    // TF_LayoutSensitiveInterface:
    StringRef GetOptimalLayout(const RuntimeDevices& devices);
    LogicalResult UpdateDataFormat(StringRef data_format);
  }];
}

def TF_FusedBatchNormV3Op : TF_Op<"FusedBatchNormV3", [Pure, TF_FoldOperandsTransposeInterface, TF_LayoutSensitiveInterface]> {
  let summary = "Batch normalization.";

  let description = [{
Note that the size of 4D Tensors are defined by either "NHWC" or "NCHW".