// CHECK-LABEL:   func @batchmatmulv2_lhs_batch
// CHECK:           "mhlo.dynamic_broadcast_in_dim"({{.*}}, {{.*}}) <{broadcast_dimensions = dense<[1, 2]> : tensor<2xi64>}>
// CHECK:           "mhlo.dot_general"({{.*}}, {{.*}}) <{
// CHECK-SAME:        lhs_batching_dimensions = [0]
// CHECK-SAME:        rhs_batching_dimensions = [0]
// CHECK-SAME:        lhs_contracting_dimensions = [2]

// CHECK: %[[BITCAST2:.+]] = "tfl.bitcast"(%[[ARG2]]) : (tensor<i64>) -> tensor<1xi64>
// CHECK: %[[MIN2:.+]] = "tfl.minimum"(%[[BITCAST2]], %[[MAX2]]) : (tensor<1xi64>, tensor<1xi64>) -> tensor<1xi64>
// CHECK: %[[CONCAT:.+]] = "tfl.concatenation"(%[[MIN1]], %[[MIN2]]) <{axis = 0 : i32, fused_activation_function = "NONE"}> : (tensor<1xi64>, tensor<1xi64>) -> tensor<2xi64>

  %cst_0 = arith.constant dense<[2, -1]> : tensor<2xi64>
  %0 = "tf.FakeQuantWithMinMaxVars"(%arg0, %arg1, %arg2) {num_bits = 5, narrow_range = false} : (tensor<1x2xf32>, tensor<f32>, tensor<f32>) -> tensor<1x2xf32>
  %1 = "tf.Reshape"(%0, %cst_0) : (tensor<1x2xf32>, tensor<2xi64>) -> tensor<2x1xf32>
  %2 = "tf.Reshape"(%0, %cst_0) : (tensor<1x2xf32>, tensor<2xi64>) -> tensor<2x1xf32>
  func.return %1, %2 : tensor<2x1xf32>, tensor<2x1xf32>

  %1 = "arith.constant"() {value = dense<0> : tensor<16xi64>} : () -> tensor<16xi64>
  %2 = "tfl.conv_2d"(%arg0, %0, %1) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_h = 1 : i32, stride_w = 1 : i32} : (tensor<1x32x32x8x!quant.uniform<i16:f32, 1.0>>, tensor<16x1x1x8x!quant.uniform<i8:f32, 1.0>>, tensor<16xi64>) -> tensor<1x32x32x16x!quant.uniform<i16:f32, 1.0>>

// CHECK: %[[CUSTOM_AGGREGATOR_1:.*]], %[[MIN_1:.*]], %[[MAX_1:.*]], %[[HISTOGRAM_1:.*]] = "tf.CustomAggregator"
// CKECK-SAME: <{calibration_method = 5 : i32, id = "1", num_bins = 32 : i32, max_percentile = 0.000000e+00 : f32, min_percentile = 0.000000e+00 : f32}> : (tensor<1x3x4x3xf32>) -> (tensor<1x3x4x3xf32>, tensor<f32>, tensor<f32>, tensor<512xi64>)

  %cst_0 = arith.constant dense<[2, -1]> : tensor<2xi64>
  %0 = "tf.FakeQuantWithMinMaxVars"(%arg0, %arg1, %arg2) {num_bits = 3, narrow_range = false} : (tensor<1x2xf32>, tensor<f32>, tensor<f32>) -> tensor<1x2xf32>
  %1 = "tf.Reshape"(%0, %cst_0) : (tensor<1x2xf32>, tensor<2xi64>) -> tensor<2x1xf32>
  %2 = "tf.Reshape"(%0, %cst_0) : (tensor<1x2xf32>, tensor<2xi64>) -> tensor<2x1xf32>
  func.return %1, %2 : tensor<2x1xf32>, tensor<2x1xf32>

    %range_nume = "tf.Sub"(%r_max_f64, %r_min_f64) : (tensor<1xf64>, tensor<1xf64>) -> tensor<1xf64>
    %range_deno = "tf.Sub"(%i8_max_f64, %i8_min_f64) : (tensor<f64>, tensor<f64>) -> tensor<f64>

    %scale_f64 = "tf.Div"(%range_nume, %range_deno) : (tensor<1xf64>, tensor<f64>) -> tensor<1xf64>
    %scale = "tf.Cast"(%scale_f64) : (tensor<1xf64>) -> tensor<1xf32>

    // Add comparison with minimum if needed

        %1 = "tf.OpB"() {_XlaSharding = "\08\03\1A\02\02\01\22\02\00\02"} : () -> tensor<2x2xi64>
        // expected-error @+1 {{should have the same sharding}}
        %2 = "tf.OpC"(%0, %1) {_xla_map_outside_compilation = "0", _xla_outside_compilation = "from_launch"} : (tensor<2x2xi64>, tensor<2x2xi64>) -> tensor<2x2xi64>
        "tf.OpD"(%2) : (tensor<2x2xi64>) -> ()
        tf_device.return

  %1 = "tfl.shape"(%conv_s) : (tensor<?x26x26x26x16xf32>) -> tensor<5xi64>
  %2 = "tfl.broadcast_args"(%1, %cst) : (tensor<5xi64>, tensor<1xi64>) -> tensor<5xi64>
  %broad1 = "tfl.broadcast_to"(%conv_s, %2) : (tensor<?x26x26x26x16xf32>, tensor<5xi64>) -> tensor<?x26x26x26x16xf32>
  %broad2 = "tfl.broadcast_to"(%b, %2) : (tensor<16xf32>, tensor<5xi64>) -> tensor<?x26x26x26x16xf32>

    %24 = "tf.Softmax"(%18) : (tensor<2x1001xf32>) -> tensor<2x1001xf32>
    %25 = "tf.ArgMax"(%24, %2) : (tensor<2x1001xf32>, tensor<i32>) -> tensor<2xi64>
    %26 = "tf.Cast"(%25) {Truncate = false} : (tensor<2xi64>) -> tensor<2xf32>
    %27 = "tf.Equal"(%14, %26) {incompatible_shape_error = true} : (tensor<2xf32>, tensor<2xf32>) -> tensor<2xi1>