'has_bias': [True, False],
          'has_batch_norm': [True, False],
          'target_opset': [quant_opts_pb2.XLA],
      }])
  )
  @test_util.run_in_graph_and_eager_modes
  def test_qat_conv_model(
      self,
      activation_fn: Optional[ops.Operation],
      has_bias: bool,
      has_batch_norm: bool,
      target_opset: quant_opts_pb2.OpSet,
  ):

        filter_shape,
        self._input_saved_model_path,
        bias_fn,
        activation_fn,
        has_batch_norm,
        strides,
        dilations,
        'SAME',
        has_func_alias,
    )
    # TODO: b/331809306 - Investigate why these test fail then re-enable.
    if has_batch_norm and (bias_fn or not input_shape_dynamic):
      return

      input_shape: Sequence[int],
      filter_shape: Sequence[int],
      saved_model_path: str,
      bias_fn: Optional[ops.Operation] = None,
      activation_fn: Optional[ops.Operation] = None,
      has_batch_norm: bool = False,
      strides: Sequence[int] = (1, 1, 1, 1),
      dilations: Sequence[int] = (1, 1, 1, 1),
      padding: str = 'SAME',
      has_func_alias: bool = False,
  ) -> module.Module:

    return GatherModel(use_variable)

  def _create_depthwise_conv2d_model(
      self,
      input_shape: Sequence[int],
      filter_shape: Sequence[int],
      has_bias: bool = False,
      has_batch_norm: bool = False,
      activation_fn: Optional[ops.Operation] = None,
      strides: Sequence[int] = (1, 2, 2, 1),
      dilations: Sequence[int] = (1, 1, 1, 1),
      padding: str = 'SAME',
  ):