%0 = "tfl.pseudo_qconst"() { qtype = tensor<3x3x!quant.uniform<i32:f32:1, {1.0, 0.5:1, 0.25:1}>>, value = dense<1> : tensor<3x3xi32>} : () -> tensor<3x3x!quant.uniform<i32:f32:1, {1.0, 0.5:1, 0.25:1}>>

// CHECK-NEXT:  %[[argmax:.*]] = "tfl.arg_max"(%[[softmax]], %[[cst]]) : (tensor<128x16xf32>, tensor<i32>) -> tensor<128xi32>
  %3 = "tfl.arg_max"(%2, %cst) : (tensor<128x16xf32>, tensor<i32>) -> tensor<128xi32>
// CHECK-NEXT:  return %[[softmax]], %[[argmax]] : tensor<128x16xf32>, tensor<128xi32>
  func.return %2, %3 : tensor<128x16xf32>, tensor<128xi32>
}

// CHECK-LABEL: PruneUnusedLstm

  // CHECK:  %5 = "tfl.arg_max"(%arg0, %cst) : (tensor<1x32x1xf32>, tensor<1xi32>) -> tensor<1x1xi32>
  // CHECK:  return %4, %5 : tensor<1x1xf32>, tensor<1x1xi32>
}

// CHECK-LABEL:   func @convert_pytorch_argmax
func.func @convert_pytorch_argmax(%arg0: tensor<1x9xi32>) -> tensor<1xi32> {
  %0 = mhlo.constant dense<0> : tensor<i32>
  %1 = mhlo.constant dense<-2147483648> : tensor<i32>

  } : (tensor<2x7xi64>) -> (tensor<2x7xi32>)
  %1 = "tf.Identity"(%0) : (tensor<2x7xi32>) -> tensor<2x7xi32>
  func.return %1 : tensor<2x7xi32>
}

// CHECK-LABEL: func @func
// CHECK-SAME: %arg0: tensor<2x7xi64> {mhlo.sharding = "\08\03\1A\03\01\01\02\22\02\00\01B\01\00"
func.func @func(%arg0: tensor<2x7xi64>) -> (tensor<2x7xi32>) {

// CHECK-LABEL: reduce_window_max_activation_transpose
// CHECK-SAME: (%[[ARG:.+]]: tensor<16x8xf32>) -> tensor<4x8xf32>
func.func @reduce_window_max_activation_transpose_rank2(%arg0: tensor<16x8xf32>) -> tensor<4x8xf32> {
  %0 = stablehlo.constant dense<0xFF800000> : tensor<f32>  // -inf
  %1 = stablehlo.transpose %arg0, dims = [1, 0] : (tensor<16x8xf32>) -> tensor<8x16xf32>
  %2 = "stablehlo.reduce_window"(%1, %0) ({

  %3 = "tf.AddV2"(%rhs_dilation, %rhs_dilation) {device = ""} : (tensor<3xi32>, tensor<3xi32>) -> tensor<3xi32>
  %4 = "tf.AddV2"(%padding, %padding) {device = ""} : (tensor<3x2xi32>, tensor<3x2xi32>) -> tensor<3x2xi32>
  %5 = "tf.AddV2"(%strides, %strides) {device = ""} : (tensor<3xi32>, tensor<3xi32>) -> tensor<3xi32>

  return %0 : tensor<8x4x14x14x16xf32>
}

  %group_assignment = "tf.Const"() { value = dense<[[0],[1]]> : tensor<2x1xi32> } : () -> tensor<2x1xi32>
  // CHECK: "mhlo.all_reduce"
  // CHECK{LITERAL}: replica_groups = dense<[[0], [1]]> : tensor<2x1xi64>
  // CHECK-NOT: channel_handle
  // CHECK: mhlo.add
  %0 = "tf.XlaAllReduce"(%input, %group_assignment) {reduce_op = "Add", mode = "CrossReplica"} : (tensor<f32>, tensor<2x1xi32>) -> tensor<f32>
  func.return %0 : tensor<f32>
}

// -----

  %1 = "tfl.pseudo_qconst"() {qtype = tensor<128x!quant.uniform<i32:f32, 0.7>>, value = dense<0> : tensor<128xi32>} : () -> tensor<128x!quant.uniform<i32:f32, 0.7>>

func.func @add_dense_dense_int_mixing_1_n() -> tensor<2x2xi32> {
  %cst_0 = arith.constant dense<[[1, 2]]> : tensor<1x2xi32>
  %cst_1 = arith.constant dense<[[3], [4]]> : tensor<2x1xi32>

  %0 = "tfl.add"(%cst_0, %cst_1) {fused_activation_function = "NONE"} : (tensor<1x2xi32>, tensor<2x1xi32>) -> tensor<2x2xi32>

  func.return %0 : tensor<2x2xi32>

        }) : () -> 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
        }) : () -> ()