%rst = "tf.Conv2D"(%arg, %fq) {T = "tfdtype$DT_FLOAT", data_format = "NHWC", dilations = [1, 2, 3, 1], padding = "SAME", strides = [1, 4, 5, 1]} : (tensor<256x32x32x3xf32>, tensor<3x3x3x16xf32>) -> tensor<256x8x7x16xf32>
  func.return %rst : tensor<256x8x7x16xf32>

        offset = [0.5] * self.out_channel_size
        mean, variance = scale, offset
        out = nn_ops.conv2d(
            input_tensor,
            self.filters,
            strides=strides,
            dilations=dilations,
            padding=padding,
            data_format='NHWC',
            name='sample/conv',
        )
        if bias_fn is not None:
          out = nn_ops.bias_add(out, self.bias)

  %0 = "tf.Conv2D"(%output, %cst) <{data_format = "NHWC", dilations = [1, 1, 1, 1], explicit_paddings = [], padding = "SAME", strides = [1, 2, 2, 1], use_cudnn_on_gpu = true}> {attr_map = "0:strides,1:use_cudnn_on_gpu,2:padding,3:explicit_paddings,4:dilations", device = ""} : (tensor<1x3x4x3xf32>, tensor<2x3x3x2xf32>) -> tensor<1x2x2x2xf32>

        %w = "tf.Mul"(%cst, %cst_2) : (tensor<2x3x3x1xf32>, tensor<f32>) -> tensor<2x3x3x1xf32>
        %0 = "tf.DepthwiseConv2dNative"(%arg0, %w) {data_format = "NHWC", device = "", dilations = [1, 1, 1, 1], explicit_paddings = [], padding = "SAME", strides = [1, 1, 1, 1]} : (tensor<*xf32>, tensor<2x3x3x1xf32>) -> tensor<?x?x?x3xf32>

      const int64_t dim = GetTensorSpatialDimIndex(num_dims, data_format, i);
      spatial_dim_indices.push_back(dim);

      const int64_t dilation = get_int(dilations[dim]);
      rhs_dilations.push_back(dilation);
      const int64_t stride = get_int(strides[dim]);
      window_strides.push_back(stride);

      Value pad_low, pad_high;

        mean, variance = scale, offset
        out = nn_ops.conv2d(
            input_tensor,
            self.filters,
            strides=strides,
            dilations=dilations,
            padding=padding,
            data_format='NHWC',
        )
        if has_bias:
          out = nn_ops.bias_add(out, self.bias, data_format='NHWC')
        if has_batch_norm:

    has_batch_norm = True
    dilations = None
    input_shape = (1, 3, 4, 3)
    filter_shape = (2, 3, 3, 2)
    strides = (1, 1, 1, 1)
    model = self._create_conv2d_model(
        input_shape,
        filter_shape,
        self._input_saved_model_path,
        bias_fn,
        activation_fn,
        has_batch_norm,
        strides,
        dilations,
    )

    state.stride_height = height;
    state.stride_width = width;

    if (TFIntListIs1XY1(op, "dilations", &height, &width)) {
      state.dilation_height_factor = height;
      state.dilation_width_factor = width;
    } else {
      // If the 'dilations' attribute is missing, we use the default value (1)
      // for both dilation height and width factor.
      state.dilation_height_factor = intAttrOne;

  int64_t dilations_size = dilations.size();
  if (dilations_size != num_dims)
    return emitOptionalError(
        location, "requires dilations attribute length to be ", num_dims);
  if (llvm::any_of(dilations, is_not_positive))
    return emitOptionalError(location, "requires positive dilations");

  return success();
}

// Verifies that,

    %5 = "tf.Conv2D"(%1, %3) {
      padding = "VALID", strides = [1, 1, 1, 1],
      attr_map = "strides:0,use_cudnn_on_gpu:1,padding:2,explicit_paddings:3,dilations:4"
    } : (tensor<*xi32>, tensor<*xi32>) -> tensor<*xi32>
    func.return %5 : tensor<*xi32>
  }

  // DepthwiseConv2D with (simulated) int32 accumulation.
  func.func private @internal_depthwise_conv2d_fn(