func.func @testSoftmax(%arg0: tensor<10x10xf32>) -> tensor<10x10xf32> {
  // CHECK: _arithmetic_count = 6400 : i64
  %0 = "tfl.softmax"(%arg0) {beta = 1.000000e+00 : f32} : (tensor<10x10xf32>) -> tensor<10x10xf32>
  func.return %0 : tensor<10x10xf32>
}

func.func @testTanh(%arg0: tensor<10x10xf32>) -> tensor<10x10xf32> {
  // CHECK: _arithmetic_count = 6400 : i64
  %0 = "tfl.tanh"(%arg0) : (tensor<10x10xf32>) -> tensor<10x10xf32>

// CHECK-LABEL: testFoldEnsureShapeOp
func.func @testFoldEnsureShapeOp(%arg0: tensor<10x20xf32>) -> (tensor<10x20xf32>, tensor<10x20xf32>, tensor<20x10xf32>) {
  %0 = "tf.EnsureShape"(%arg0) {shape = #tf_type.shape<10x20>} : (tensor<10x20xf32>) -> tensor<10x20xf32>
  %1 = "tf.EnsureShape"(%arg0) {shape = #tf_type.shape<?x20>} : (tensor<10x20xf32>) -> tensor<10x20xf32>
  // Failing case which should not be folded.

  %0 = "tfl.unidirectional_sequence_lstm"(%arg0, %arg1, %arg2, %arg3, %arg4, %arg5, %arg6, %arg7, %arg8, %arg9, %arg10, %arg11, %arg12, %arg13, %arg14, %arg15, %arg16, %arg17, %arg18, %arg19,...

}

// -----

func.func @testBatchMatMulV2(%lhs: tensor<f32>, %rhs: tensor<10x10xf32>) {
  // expected-error @+1 {{requires lhs operand to have rank at least two}}
  %0 = "tf.BatchMatMulV2"(%lhs, %rhs) : (tensor<f32>, tensor<10x10xf32>) -> tensor<10x10xf32>
}

// -----

func.func @testBatchMatMulV2(%lhs: tensor<10x10xf32>, %rhs: tensor<f32>) {
  // expected-error @+1 {{requires rhs operand to have rank at least two}}

func.func @RemoveFcZeroBias(%arg0: tensor<1x37xf32>, %arg1: tensor<40x37xf32>) -> tensor<1x40xf32> {
  %0 = "tfl.pseudo_const"() {value = dense<0.0> : tensor<40xf32>} : () -> tensor<40xf32>
  %1 = "tfl.fully_connected"(%arg0, %arg1, %0) {fused_activation_function = "NONE", keep_num_dims = false, weights_format = "DEFAULT"} : (tensor<1x37xf32>, tensor<40x37xf32>, tensor<40xf32>) -> tensor<1x40xf32>
// CHECK: "tfl.fully_connected"

    %add_out, %add_control = tf_executor.island wraps "tf.AddV2"(%const_out, %const_out) {_xla_compile_device_type = "TPU", _replication_info = "cluster"} : (tensor<4x4xf32>, tensor<4x4xf32>) -> tensor<4x4xf32>
    %replicated_out, %replicated_control = tf_executor.island wraps "tf.TPUReplicatedOutput"(%add_out) : (tensor<4x4xf32>) -> (tensor<4x4xf32>)

# CHECK:"tf.MlirPassthroughOp"
# CHECK: mlir_module = "\0Afunc @main(%arg0 : tensor<10xf32>, %arg1 : tensor<10xf32>) -> tensor<10x10xf32> {\0A %add = \22tf.Add\22(%arg0, %arg1) : (tensor<10xf32>, tensor<10xf32>) -> tensor<10xf32>\0A %ret = \22magic.op\22(%add, %add) : (tensor<10xf32>, tensor<10xf32>) -> tensor<10x10xf32>\0A return %ret : tensor<10x10xf32>\0A}\0A"}> {device = ""} : (tensor<10xf32>, tensor<10xf32>) -> tensor<*xf32>

node {
  name: "x"
  op: "Placeholder"

  %2 = "tf.Const"() {value = dense<0.0> : tensor<2x2xf32>} : () -> tensor<2x2xf32>
  %3 = "tf.Const"() {value = dense<0.0> : tensor<4x4xf32>} : () -> tensor<4x4xf32>
  "tf.SendTPUEmbeddingGradients"(%2, %3) {_tpu_embedding_layer = "call1", config = "\0A\0B\0C\0D", operandSegmentSizes = array<i32: 2, 0>} : (tensor<2x2xf32>, tensor<4x4xf32>) -> ()

    : ( tensor<1x28x28xf32>,
        tensor<20x20xf32>, tensor<20x20xf32>, tensor<20x20xf32>, tensor<20x20xf32>,
        tensor<20x20xf32>, tensor<20x20xf32>, tensor<20x20xf32>, tensor<20x20xf32>,
        tensor<20xf32>, tensor<20xf32>, tensor<20xf32>,
        tensor<20xf32>, tensor<20xf32>, tensor<20xf32>, tensor<20xf32>,
        tensor<20x20xf32>, none,
        tensor<1x20xf32>, tensor<1x20xf32>,

func.func @float_matmul(
  %arg0: tensor<1x10xf32>, %arg1: tensor<10x10xf32>) -> (tensor<*xf32>, tensor<*xf32>, tensor<*xf32>) {
  %cst = "tf.Const"() {value = dense<0.000000e+00> : tensor<10xf32>} : () -> tensor<10xf32>
  %0 = "tf.MatMul"(%arg0, %arg1) {
    transpose_a = false, transpose_b = false
  } : (tensor<1x10xf32>, tensor<10x10xf32>) -> tensor<*xf32>