}

// Returns true if the given TensorFlow op does not have a `data_format`
// attribute (then default to "NDHWC"), or its `data_format` attribute is
// "NDHWC".
inline bool TFDataFormatIsNDHWC(Operation *op) {
  auto attr = op->getAttrOfType<StringAttr>("data_format");
  return !attr || attr.getValue() == "NDHWC";
}

// Returns true if the given `op`
//   * has an attribute with the given `name`,

                                        input, num, axis);
  return success();
}

LogicalResult ConvertTFConv3DOp::matchAndRewrite(
    Operation* op, PatternRewriter& rewriter) const {
  if (!TFDataFormatIsNDHWC(op)) return failure();

  auto tf_op = cast<TF::Conv3DOp>(op);

  IntegerAttr stride_depth, stride_height, stride_width;
  if (!TFIntListIs1XYZ1(op, "strides", &stride_depth, &stride_height,

      if (!exponential_avg_factor)
        exponential_avg_factor =
            rewriter.getFloatAttr(rewriter.getF32Type(), 1.0f);
    }
    if (!TFDataFormatIsNHWC(fused_batch_norm_op) &&
        !TFDataFormatIsNDHWC(fused_batch_norm_op))
      return failure();

    if (!(((*root.getODSResults(1).begin()).use_empty()))) {
      return rewriter.notifyMatchFailure(
          fused_batch_norm_op, [&](::mlir::Diagnostic &diag) {