%2 = "tf.AddN"(%arg0, %1, %1) : (tensor<2xi32>, tensor<2xi32>, tensor<2xi32>) -> tensor<2xi32>
  %3 = "tf.AddN"(%1, %arg0, %1) : (tensor<2xi32>, tensor<2xi32> , tensor<2xi32>) -> tensor<2xi32>
  %4 = "tf.AddN"(%1, %1) : (tensor<2xi32>, tensor<2xi32>) -> tensor<2xi32>
  %5 = "tf.AddN"(%arg0, %1, %0) : (tensor<2xi32>, tensor<2xi32>, tensor<2xi32>) -> tensor<2xi32>
  func.return %2, %3, %4, %5: tensor<2xi32>, tensor<2xi32>, tensor<2xi32>, tensor<2xi32>
}

  %group_assignment = "tf.Const"() { value = dense<[[0, 1]]> : tensor<1x2xi32> } : () -> tensor<1x2xi32>
  %instance_key = "tf.Const"() { value = dense<3> : tensor<i32> } : () -> tensor<i32>
  %group_size, %group_key = "tf.CollectiveAssignGroupV2"(%group_assignment, %rank, %key_base) {} : (tensor<1x2xi32>, tensor<i32>, tensor<i32>) -> (tensor<i32>, tensor<i32>)
  // CHECK-NOT: "tf.CollectiveAssignGroupV2"

        }) : () -> tensor<f32>
        "tf_device.cluster"() ({
          %group = "tf.Const"() <{value = dense<[[0, 1]]> : tensor<1x2xi32>}> : () -> tensor<1x2xi32>
          %arg1_reduced = "tf.XlaAllReduce"(%arg1_id, %group) <{mode = "CrossReplica", reduce_op = "Add"}> : (tensor<f32>, tensor<1x2xi32>) -> tensor<f32>
          "tf.OpA"(%arg1_reduced) : (tensor<f32>) -> ()
          tf_device.return
        }) : () -> ()

}

func.func @einsum_reducesum(%arg0: tensor<2x5x7xf32>, %arg1: tensor<5x2xf32>) -> tensor<5x7xf32> {
  %0 = "tf.Einsum"(%arg0, %arg1) {T = "tfdtype$DT_FLOAT", equation = "lbh,bl->bh"}: (tensor<2x5x7xf32>, tensor<5x2xf32>) -> tensor<5x7xf32>
  func.return %0 : tensor<5x7xf32>
  // CHECK-LABEL: einsum_reducesum
  // CHECK-DAG: %[[cst:.*]] = arith.constant dense<[1, 2, 0]> : tensor<3xi32>

    "vhlo.return_v1"(%lhs) : (tensor<f32>) -> ()
  }): (tensor<200x100x300xf32>, tensor<10x2xi32>, tensor<10x300xf32>) ->
      tensor<200x100x300xf32>
  func.return %0 : tensor<200x100x300xf32>
}

// CHECK-LABEL: func.func private @scatter(%arg0: tensor<200x100x300xf32>, %arg1: tensor<10x2xi32>, %arg2: tensor<10x300xf32>) -> tensor<200x100x300xf32> {

// CHECK: [[UNPACK:%.*]]:3 = "tf.Unpack"(%arg0) <{axis = 0 : i64}> : (tensor<3x2x2xf32>) -> (tensor<2x2xf32>, tensor<2x2xf32>, tensor<2x2xf32>)
// CHECK: [[SCALAR_ZERO:%.*]] = arith.constant dense<0> : tensor<i32>
// CHECK: [[CONCAT:%.*]] = "tf.Concat"([[SCALAR_ZERO]], [[UNPACK]]#0, [[UNPACK]]#1, [[UNPACK]]#2) : (tensor<i32>, tensor<2x2xf32>, tensor<2x2xf32>, tensor<2x2xf32>) -> tensor<?x2xf32>

func.func @reshape_vector_shape(tensor<4x4x4xf32>) -> tensor<16x4xf32> {
^bb0(%arg0: tensor<4x4x4xf32>) :
  %shape0 = arith.constant dense<[[16, 4]]> : tensor<1x2xi32>
  // expected-error @+1 {{'tfl.reshape' op requires 'shape' to be rank 1, but got 2}}
  %1 = "tfl.reshape"(%arg0, %shape0) : (tensor<4x4x4xf32>, tensor<1x2xi32>) -> tensor<16x4xf32>
  func.return %1 : tensor<16x4xf32>
}

// -----

// Note: Argument quantization sequence omitted.
// CHECK: stablehlo.dot_general %{{.+}}, %{{.+}}, contracting_dims = [1] x [0] : (tensor<1x1024xi8>, tensor<1024x3xi8>) -> tensor<1x3xi32>

// Note: Result dequantization sequence omitted.
// CHECK: return %{{.+}} : tensor<1x3xf32>

// -----

// Tests that a simple dot_general (lifted as a function) with CustomAggregators

    return %3: tensor<512xi32>
  }
func.func private @_func(%arg0: tensor<?xi32, #mhlo.type_extensions<bounds = [512]>> {mhlo.sharding = "\08\01\1A\01\01\22\01\00"}) -> (tensor<512xi32>) {
    %0 = "tf.A"(%arg0) {} : (tensor<?xi32, #mhlo.type_extensions<bounds = [512]>>) -> tensor<512xi32>
    return %0 : tensor<512xi32>
  }
}

// -----

    // CHECK: %[[rhs:.*]] = "tf.Cast"(%arg1) <{Truncate = false}> : (tensor<2xi32>) -> tensor<2x!tf_type.qint32>
    // CHECK: %[[res_qint:.*]] = "tf.UniformQuantizedAdd"(%[[lhs]], %[[rhs]]
    // CHECK: %[[res_int:.*]] = "tf.Cast"(%[[res_qint]]) <{Truncate = false}> : (tensor<2x!tf_type.qint32>) -> tensor<2xi32>
    // CHECK: return %[[res_int]] : tensor<2xi32>