// CHECK: %[[v0:.*]] = "tf.Transpose"(%arg1, %[[cst]]) : (tensor<2x5x3x7xf32>, tensor<4xi32>) -> tensor<2x5x7x3xf32>
  // CHECK: %[[v1:.*]] = "tf.Reshape"(%arg0, %[[cst_1]]) : (tensor<2x5x7xf32>, tensor<4xi64>) -> tensor<2x5x1x7xf32>
  // CHECK: %[[v2:.*]] = "tf.BatchMatMulV2"(%[[v1]], %[[v0]]) <{adj_x = false, adj_y = false}> : (tensor<2x5x1x7xf32>, tensor<2x5x7x3xf32>) -> tensor<2x5x1x3xf32>

      tf_executor.fetch %outputs#1, %outputs#0 : tensor<2x2x1x2xi32>, tensor<2x2x1x2xf32>
    }
    func.return %0#0, %0#1 : tensor<2x2x1x2xi32>, tensor<2x2x1x2xf32>
  }
}

// CHECK:        name: "serving_default_x"
// CHECK-NEXT:   op: "_Arg"

    %12 = "quantfork.qcast"(%11) {volatile} : (tensor<2x3x1x3xf32>) -> tensor<2x3x1x3x!quant.uniform<i8:f32, 3.000000e-01:1>>
    %13 = "quantfork.dcast"(%12) : (tensor<2x3x1x3x!quant.uniform<i8:f32, 3.000000e-01:1>>) -> tensor<2x3x1x3xf32>
    return %13 : tensor<2x3x1x3xf32>
  }

  // CHECK: quantized_dot_general_fn_1

      stablehlo.return %2 : tensor<f32>
  }) {padding = dense<[[0, 0], [1, 1], [1, 1], [0, 0]]> : tensor<4x2xi64>, window_dimensions = array<i64: 1, 3, 3, 1>} : (tensor<2x3x1x3xf32>, tensor<f32>) -> tensor<2x3x1x3xf32>
  return %1 : tensor<2x3x1x3xf32>
}

// -----

// CHECK-LABEL: @bitcast_convert_i32_f32(%arg0: tensor<1x256128xi32>) -> tensor<1x256128xbf16>

  // CHECK-NEXT: return [[RESULT]] : tensor<1x128x128x8xf32>
}

func.func @testDilatedConvWithExpandSqueeze1(%arg0: tensor<1x128x128xf32>, %arg1: tensor<5x5x1x1xf32>, %arg2: tensor<128xf32>) -> tensor<1x128x128xf32> {
  %cst = arith.constant dense<[2, 2]> : tensor<2xi32>
  %cst_0 = "tf.Const"() { value = dense<3> : tensor<i32> } : () -> tensor<i32>

# CHECK:           %[[VAL_12:.*]] = "tfl.pack"(%[[VAL_10]], %[[VAL_11]]) <{axis = 0 : i32, values_count = 2 : i32}> : (tensor<5x7xf32>, tensor<5x7xf32>) -> tensor<2x5x7xf32>
# CHECK:           return %[[VAL_12]] : tensor<2x5x7xf32>

// PerChannel-NOT: tensor<2x3x3x2xf32>
// PerChannel-SAME: tensor<2x3x1x6xf32>
// PerChannel: %[[PARTITIONEDCALL_0:.*]] = "tf.PartitionedCall"(%arg0, %[[CONST_1:.*]]) <{config = "", config_proto = "", executor_type = "", f = @composite_depthwise_conv2d_fn_0}> {_tfl_quant_trait = "fully_quantizable"} : (tensor<1x3x4x3xf32>, tensor<2x3x1x6xf32>) -> tensor<*xf32>

// CHECK: %[[CONST_1:.*]] = arith.constant dense
// CHECK-NOT: tensor<2x3x3x2xf32>
// CHECK-SAME: tensor<2x3x1x6xf32>
// CHECK: %[[PARTITIONEDCALL_0:.*]] = "tf.PartitionedCall"(%arg0, %[[CONST_1:.*]]) <{config = "", config_proto = "", executor_type = "", f = @composite_depthwise_conv2d_fn_0}> {_tfl_quant_trait = "fully_quantizable"} : (tensor<1x3x4x3xf32>, tensor<2x3x1x6xf32>) -> tensor<*xf32>

  return %2 : tensor<?x2x2x1xi8>
}

// -----

// CHECK-LABEL: func @convolution_add_add_f32
func.func @convolution_add_add_f32(
    %lhs: tensor<?x3x2x1xf32>, %rhs: tensor<2x1x1x1xf32>,
    %zp_offset: tensor<?x2x2x1xf32>, %bias: tensor<1xf32>
  ) -> tensor<?x2x2x1xf32> {
  // CHECK-DAG: %[[conv:.*]] = mhlo.convolution
  // CHECK-DAG: %[[combined:.*]] = chlo.broadcast_add %[[conv:.*]], %[[zp_offset:.*]]

// CHECK:           })
// CHECK-SAME:        -> tensor<2x4x6x7xf32>
// CHECK:           %[[RESULT:.*]] = mhlo.divide %[[DIVIDEND]], %[[DIVISOR]] : tensor<2x4x6x7xf32>
// CHECK:           return %[[RESULT]] : tensor<2x4x6x7xf32>
// CHECK:         }
func.func @avgpool_same_padding(%arg0: tensor<2x12x21x7xf32>) -> tensor<2x4x6x7xf32> {