// CHECK: return %[[v5]] : tensor<?x?x8xf32>
}

func.func @einsum_no_reshape(%arg0 : tensor<?x?x8x128xf32>, %arg1 : tensor<1x?x8x128xf32>) -> tensor<?x?x8x?xf32> {
  %0 = "tf.Einsum"(%arg0, %arg1) {device = "", equation = "BQNH,BTNH->BQNT"} : (tensor<?x?x8x128xf32>, tensor<1x?x8x128xf32>) -> tensor<?x?x8x?xf32>
  func.return %0 : tensor<?x?x8x?xf32>
// CHECK-LABEL: einsum_no_reshape

// CHECK:           %[[VAL_15:.*]] = "tfl.reshape"(%[[VAL_14]], %[[VAL_5]]) {tac.device = "GPU", tac.inference_type = "FLOAT"} : (tensor<384x128xf32>, tensor<3xi32>) -> tensor<1x384x128xf32>

  %2 = "tf.Conv2D"(%1, %arg1) {padding = "VALID", strides = [1, 1, 1, 1]} : (tensor<2x68x1x3xf32>, tensor<5x1x3x8xf32>) -> tensor<2x64x1x8xf32>
  %3 = "tf.Squeeze"(%2) {squeeze_dims = [-2]} : (tensor<2x64x1x8xf32>) -> tensor<2x64x8xf32>
  %4 = "tf.BatchToSpaceND"(%3, %cst_1, %cst) : (tensor<2x64x8xf32>, tensor<1xi32>, tensor<1x2xi32>) -> tensor<1x128x8xf32>
  func.return %4 : tensor<1x128x8xf32>

      // CHECK-SAME:     scatter_dimension = 0
      //
      %1 = "tf.XlaReduceScatter"(%arg0, %cst_0, %cst) {reduce_op = "Add"} : (tensor<128x128xf32>, tensor<4x2xi32>, tensor<i32>) -> tensor<64x128xf32>
      func.return %1 : tensor<64x128xf32>
  }

  // CHECK-LABEL: func @tf_mod
  func.func @tf_mod(%arg1: tensor<2x2xf32>) -> tensor<2x2xf32> {
    %cst = "tf.Const"() {value = dense<7.000000e+00> : tensor<f32>} : () -> tensor<f32>