// type.
	identityTypeMap = generateReversedMap(oidTagMap)

	// The OID for the SAN extension (See
	// http://www.alvestrand.no/objectid/2.5.29.17.html).
	oidSubjectAlternativeName = asn1.ObjectIdentifier{2, 5, 29, 17}
)

// Identity is an object holding both the encoded identifier bytes as well as
// the type of the identity.
type Identity struct {
	Type  IdentityType
	Value []byte

    if (user->getDialect()->getNamespace() != "tf") {
      return failure();
    }
    // Identity ops of returning values might be helpful for some other
    // compilers, so avoid removing these Identity ops.
    if (user->hasTrait<OpTrait::IsTerminator>()) {
      return failure();
    }
  }

  rewriter.replaceOp(identity, identity.getInput());
  return success();
}

}  // namespace quant

  // CHECK-NOT: tf.Identity
  %5 = "tf.Identity"(%4) {T = f32, _output_shapes = ["tfshape$dim { size: 3 } dim { size: 3 }"], device = "/device:CPU:0"} : (tensor<3x3xf32>) -> tensor<3x3xf32>
  // CHECK-NOT: tf.IdentityN

  attributes  {tf.entry_function = {inputs = "args_0,args_1", outputs = "rets_0,rets_1"}} {
    %0:2 = tf_executor.graph {
      %1:3 = tf_executor.island wraps "tf.IdentityN"(%arg0, %arg1) {T = ["tfdtype$DT_FLOAT", "tfdtype$DT_INT32"], device = "", name = "identity"} : (tensor<*xf32>, tensor<2x4x6x8xi32>) -> (tensor<*xf32>, tensor<2x4x6x8xi32>)
      tf_executor.fetch %1#0, %1#1 : tensor<*xf32>, tensor<2x4x6x8xi32>
    }

  attributes  {tf.entry_function = {inputs = "args_0,args_1", outputs = "rets_0,rets_1"}} {
    %0:2 = tf_executor.graph {
      %1:3 = tf_executor.island wraps "tf.IdentityN"(%arg0, %arg1) {T = ["tfdtype$DT_FLOAT", "tfdtype$DT_INT32"], device = "", name = "identity"} : (tensor<*xf32>, tensor<2x4x6x8xi32>) -> (tensor<*xf32>, tensor<2x4x6x8xi32>)
      tf_executor.fetch %1#0, %1#1 : tensor<*xf32>, tensor<2x4x6x8xi32>
    }

func.func @empty_island_multiple_data_results(%arg0: tensor<*xf32>, %arg1: tensor<*xi32>) {
  tf_executor.graph {
    %0:3 = tf_executor.island {
      // CHECK: %[[IDENTITY_N:.*]]:2 = "tf.IdentityN"
      // CHECK-SAME: (%[[ARG_0]], %[[ARG_1]])
      // CHECK: tf_executor.yield %[[IDENTITY_N]]#0, %[[IDENTITY_N]]#1
      tf_executor.yield %arg0, %arg1 : tensor<*xf32>, tensor<*xi32>
    }
    tf_executor.fetch
  }
  func.return
}

Aditya Manthramurthy <******@****.***> 1714599073 -0700

Laura Kassovic <******@****.***> 1701107622 -0800

  %val1 = "tf.Identity"(%val0) {_class = ["loc:@v"]} : (tensor<96x!tf_type.f32ref>) -> tensor<96xf32>
  func.return
}

// -----

// Test case: Two ReadVariable ops.

func.func @f() {
  // CHECK: "tf.VarHandleOp"

  // During lowering to resource variables, this pass will preserve the
  // locations of the ReadVariableOps as Identity ops to keep the original graph
  // composition and order.

//
// input   min cst       max cst          input   min cst       max cst
//  \       |             |                \       |             |
//   \  (tf.Identity) (tf.Identity)   =>    \  (tf.Identity) (tf.Identity)
//    \     |             |                  \     |             |
//       tf.FakeQuantWithMinMaxVars       tf.FakeQuantWithMinMaxVars
//                   |                                 |