} : (tensor<1x6x6x3xf32>, tensor<2x3x3x2x!tf_type.qint8>, tensor<f32>, tensor<i32>) -> tensor<1x4x1x2xf32>
  func.return %0 : tensor<1x4x1x2xf32>
}

// -----

// CHECK-LABEL: func @uniform_quantized_convolution_hybrid_same
func.func @uniform_quantized_convolution_hybrid_same(%input: tensor<1x2x2x3xf32>) -> tensor<1x2x1x2xf32> {

    %cst_1 = "tf.Const"() {device = "", value = dense<[-0.0291469581, 0.0106381178]> : tensor<2xf32>} : () -> tensor<2xf32>

// CHECK:           %[[VAL_7:.*]] = "tfl.concatenation"(%[[VAL_5]], %[[VAL_6]]) <{axis = 3 : i32, fused_activation_function = "NONE"}> : (tensor<1x1x1x1xf32>, tensor<1x1x1x1xf32>) -> tensor<1x1x1x2xf32>
// CHECK:           %[[VAL_8:.*]] = "tfl.reshape"(%[[VAL_7]], %[[VAL_3]]) : (tensor<1x1x1x2xf32>, tensor<1xi32>) -> tensor<2xf32>
// CHECK:           %[[VAL_9:.*]] = "tfl.reshape"(%[[VAL_8]], %[[VAL_4]]) : (tensor<2xf32>, tensor<2xi32>) -> tensor<2x1xf32>

  return %0 : tensor<1x1x1x2xf32>
}
func.func private @XlaCallModule_aten.avg_pool2d.default.impl_5(%arg0: tensor<1x1x1x7xf32>) -> tensor<1x1x1x2xf32>

// CHECK-LABEL: avg_pool2d_6
// CHECK: %cst = arith.constant dense<[0, 2, 3, 1]> : tensor<4xi32>

  // CHECK: return %[[RESULT]]

  %indices = "tf.Const"() {value = dense<[1, 0]> : tensor<2xi32>} : () -> tensor<2xi32>
  %0 = "tf.DynamicStitch"(%indices, %arg0) : (tensor<2xi32>, tensor<2x2x2xf32>) -> tensor<2x2x2xf32>
  func.return %0 : tensor<2x2x2xf32>
}

// CHECK-LABEL: func @DynamicStitch_dynamic

  %2 = "tf.Const"() {value = dense<[2, 4, 4, 1]> : tensor<4xi32>} : () -> tensor<4xi32>
  %3 = "tf.Sub"(%0, %arg1) {device = ""} : (tensor<1x4x4x2xf32>, tensor<2x4x4x2xf32>) -> tensor<2x4x4x2xf32>
  %4 = "tf.Reshape"(%3, %1) {device = ""} : (tensor<2x4x4x2xf32>, tensor<3xi32>) -> tensor<2x16x2xf32>

// CHECK:           %[[VAL_8:.*]] = "tfl.reshape"(%[[VAL_7]], %[[VAL_3]]) {tac.device = "GPU", tac.inference_type = "FLOAT"} : (tensor<1x1x1x2xf32>, tensor<1xi32>) -> tensor<2xf32>

//
// ```
// %0 = stablehlo.constant dense<3>
// %1 = stablehlo.constant dense<4>
// %2 = stablehlo.constant dense<2>
// %3 = stablehlo.convolution(%%arg0, %%arg1) :
//          (tensor<?x3x4x3xf32>, tensor<2x3x3x2xf32>) -> tensor<?x3x4x2xf32>
// %4 = stablehlo.get_dimension_size %3, dim = 0 :
//          (tensor<?x3x4x2xf32>) -> tensor<i32>
// %5 = stablehlo.reshape %4 :
//          (tensor<i32>) -> tensor<1xi32>

}

func.func @squeezeSingleAxis(%arg0: tensor<2x1x2xf32>) -> tensor<2x2xf32> {
  %0 = "tf.Squeeze"(%arg0) {squeeze_dims = [1]} : (tensor<2x1x2xf32>) -> tensor<2x2xf32>
  func.return %0 : tensor<2x2xf32>

// CHECK-LABEL:squeezeSingleAxis
// CHECK:  "tfl.squeeze"(%arg0) <{squeeze_dims = [1]}> : (tensor<2x1x2xf32>) -> tensor<2x2xf32>
}

func.func @squeezeTwoAxes(%arg0: tensor<1x2x1x2xf32>) -> tensor<2x2xf32> {

    slice_sizes = array<i64: 1, 1, 2, 2>,
    indices_are_sorted = false
  } : (tensor<3x4x2x2xf32>, tensor<2x3x2xi64>) -> tensor<2x3x2x2xf32>
  return %0 : tensor<2x3x2x2xf32>
}

// CHECK-LABEL: func @gather_float
// CHECK: stablehlo.gather
// CHECK-NOT: tfl.gather_nd
// CHECK-NOT: tfl.gather

// -----

// Tests that a quantized `stablehlo.dynamic_slice` is converted to `tfl.slice`.