func.func @Int64SliceBeginSize(%arg0: tensor<4x128x32xf32>) -> tensor<1x128x32xf32> {
  %0 = "tfl.pseudo_const"() {value = dense<0> : tensor<3xi64>} : () -> tensor<3xi64>
  %1 = "tfl.pseudo_const"() {value = dense<[1, 128, 32]> : tensor<3xi64>} : () -> tensor<3xi64>
  %2 = "tfl.slice"(%arg0, %0, %1) : (tensor<4x128x32xf32>, tensor<3xi64>, tensor<3xi64>) -> tensor<1x128x32xf32>
  func.return %2 : tensor<1x128x32xf32>

// CHECK:      %[[R1:.*]] = "mhlo.concatenate"(%arg1, %arg3, %arg5, %arg7) <{dimension = 0 : i64}> : (tensor<1x128x72xf32>, tensor<1x128x72xf32>, tensor<1x128x72xf32>, tensor<1x128x72xf32>) -> tensor<4x128x72xf32>
// CHECK:      %[[R2:.*]] = "mhlo.dot_general"(%[[R0]], %[[R1]]) <{
// CHECK-SAME:   dot_dimension_numbers = #mhlo.dot<
// CHECK-SAME:     lhs_batching_dimensions = [0],

func.func private @func_20_GPU_FLOAT(%arg0: tensor<128x128xf32>, %arg1: tensor<3xi32>) -> tensor<1x128x128xf32> attributes {tac.device = "GPU", tac.inference_type = "FLOAT", tac.interface_name = "func_20"} {
  %0 = "tfl.reshape"(%arg0, %arg1) {tac.device = "GPU", tac.inference_type = "FLOAT"} : (tensor<128x128xf32>, tensor<3xi32>) -> tensor<1x128x128xf32>
  func.return %0 : tensor<1x128x128xf32>
}