// CHECK: return %[[v2]] : tensor<2x1x1x2xf32>
}

func.func @einsum_with_runtime_outputshape1(%arg0 : tensor<?x36x32xf32>, %arg1 : tensor<?x36x?x32xf32>) -> tensor<?x36x?xf32> {
  %0 = "tf.Einsum"(%arg0, %arg1) {device = "", equation = "bij,binj->bin"} : (tensor<?x36x32xf32>, tensor<?x36x?x32xf32>) -> tensor<?x36x?xf32>
  func.return %0 : tensor<?x36x?xf32>
// CHECK-LABEL: einsum_with_runtime_outputshape1

// CHECK:           [[VAL_50:%.*]] = tensor.cast [[VAL_51:%.*]] : tensor<1x3xf32> to tensor<1x?xf32>
// CHECK:           return [[VAL_50]] : tensor<1x?xf32>

  func.return %0 : tensor<1x112x112x32xf32>
}

func.func @fully_connected(%arg0: tensor<1x37xf32>, %arg1: tensor<40x37xf32>, %arg2: tensor<40xf32>) -> tensor<1x40xf32> {
  // CHECK: _arithmetic_count = 3000 : i64

// =============================================================================

func.func @sigmoidGrad(%arg0: tensor<?x32xf32>, %arg1: tensor<?x32xf32>) -> tensor<?x32xf32> {
  %0 = "tf.SigmoidGrad"(%arg0, %arg1) : (tensor<?x32xf32>, tensor<?x32xf32>) -> tensor<?x32xf32>
  func.return %0 : tensor<?x32xf32>
// CHECK-LABEL: sigmoidGrad
// CHECK-NEXT: [[ONE:%.+]] = arith.constant dense<1.000000e+00> : tensor<f32>

// 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>
// CHECK: [[LENGTHS:%.*]] = arith.constant dense<0> : tensor<0xi64>

  // CHECK-SAME: %[[ARG3:[^:[:space:]]+]]
  func.func private @main(%arg0: tensor<1x4xf32>, %arg1: tensor<4x3xf32>) -> tensor<1x3xf32> attributes {_from_xla_call_module} {
    %0 = stablehlo.dot_general %arg0, %arg1, contracting_dims = [1] x [0] : (tensor<1x4xf32>, tensor<4x3xf32>) -> tensor<1x3xf32>
    return %0 : tensor<1x3xf32>
    // CHECK: %[[DOT:.+]] = stablehlo.dot_general %[[ARG2]], %[[ARG3]]
    // CHECK: return %[[DOT]]
  }
}

  %1 = "tf.FakeQuantWithMinMaxVarsPerChannel"(%arg1, %cst_1, %cst_2) {
    device = "", narrow_range = true, num_bits = 8 : i64
  } : (tensor<4x3xf32>, tensor<3xf32>, tensor<3xf32>) -> tensor<4x3xf32>
  func.return %0, %1 : tensor<3xf32>, tensor<4x3xf32>
}

// CHECK: %[[q1:.*]] = "quantfork.qcast"(%arg0)
// CHECK-SAME: tensor<3x!quant.uniform<i8:f32, 0.0076314610593459188:-3>>

func.func @main(tensor<40x37xf32>, tensor<40x37xf32>) -> tensor<40x40xf32> {
^bb0(%arg0: tensor<40x37xf32>, %arg1: tensor<40x37xf32>):
  %cst = arith.constant dense<1.0> : tensor<40xf32>
  %0:2 = "tfl.fully_connected"(%arg0, %arg1, %cst) {fused_activation_function = "NONE", keep_num_dims = false, weights_format = "DEFAULT"} : (tensor<40x37xf32>, tensor<40x37xf32>, tensor<40xf32>) -> (tensor<40x40xf32>, tensor<40x40xf32>)

func.func @main(tensor<40x37xf32>, tensor<40x37xf32>) -> tensor<40x40xf32> {
^bb0(%arg0: tensor<40x37xf32>, %arg1: tensor<40x37xf32>):
  %cst = arith.constant dense<1.0> : tensor<40xf32>
  %0:2 = "tfl.fully_connected"(%arg0, %arg1, %cst) {fused_activation_function = "NONE", keep_num_dims = false, weights_format = "DEFAULT"} : (tensor<40x37xf32>, tensor<40x37xf32>, tensor<40xf32>) -> (tensor<40x40xf32>, tensor<40x40xf32>)

  func.return %0 : tensor<3x4x4xf32>
}

func.func @batchmatmulv2_rhs_batch(%arg0: tensor<4x2xf32>, %arg1: tensor<3x2x4xf32>) -> tensor<3x4x4xf32> {
// CHECK-LABEL:   func @batchmatmulv2_rhs_batch
// CHECK:           "mhlo.dynamic_broadcast_in_dim"({{.*}}, {{.*}}) <{broadcast_dimensions = dense<[1, 2]> : tensor<2xi64>}>