std::vector<PartialTensorShape> shapes;
  shapes.reserve(num_shapes);
  for (int i = 0; i < num_shapes; ++i) {
    if (num_dims[i] < 0) {
      shapes.emplace_back();
    } else {
      shapes.emplace_back(absl::Span<const int64_t>(
          reinterpret_cast<const int64_t*>(dims[i]), num_dims[i]));
    }
  }
  desc->node_builder.Attr(attr_name, shapes);
}

*   Tracing/timeline support for distributed runtime (no GPU profiler yet).
*   C API gives access to inferred shapes with `TF_GraphGetTensorNumDims` and
    `TF_GraphGetTensorShape`.
*   Shape functions for core ops have moved to C++ via
    `REGISTER_OP(...).SetShapeFn(...)`. Python shape inference RegisterShape
    calls use the C++ shape functions with `common_shapes.call_cpp_shape_fn`. A
    future release will remove `RegisterShape` from python.

TEST_F(CApiAttributesTest, ShapeList) {
  const int64_t shape_1[] = {1, 3};
  const int64_t shape_2[] = {2, 4, 6};
  const int64_t* list[] = {&shape_1[0], &shape_2[0]};
  const size_t list_size = TF_ARRAYSIZE(list);
  const int ndims[] = {TF_ARRAYSIZE(shape_1), TF_ARRAYSIZE(shape_2)};
  const int total_ndims = 5;  // ndims[0] + ndims[1]

  auto desc = init("list(shape)");
  TF_SetAttrShapeList(desc, "v", list, ndims, list_size);

/// tip | Dica

A coisa importante e "mágica" aqui é que `get_current_user` terá diferentes listas de `scopes` para validar para cada *operação de rota*.

## Проверка `scopes` { #verify-the-scopes }

Теперь проверяем, что все требуемые scopes — этой зависимостью и всеми зависящими (включая операции пути) — присутствуют среди scopes, предоставленных в полученном токене, иначе вызываем `HTTPException`.

Для этого используем `security_scopes.scopes`, содержащий `list` со всеми этими scopes как `str`.

## Verificar los `scopes` { #verify-the-scopes }

Ahora verificamos que todos los scopes requeridos, por esta dependencia y todos los dependientes (incluyendo *path operations*), estén incluidos en los scopes proporcionados en el token recibido, de lo contrario, lanzamos una `HTTPException`.

Para esto, usamos `security_scopes.scopes`, que contiene una `list` con todos estos scopes como `str`.

Sebastián Ramírez <******@****.***> 1766840096 -0800

    ShapeHandle shape_handle = c.output(i);
    TF_ShapeAndType& shape = output_shapes_result->items[i];
    shape.num_dims = c.Rank(shape_handle);
    if (shape.num_dims == InferenceContext::kUnknownRank) {
      shape.dims = nullptr;
      continue;
    }
    shape.dims = new int64_t[shape.num_dims];
    for (size_t j = 0; j < shape.num_dims; ++j) {
      shape.dims[j] = c.Value(c.Dim(shape_handle, j));
    }
  }

        "A parameter-less dependency must have a callable dependency"
    )
    own_oauth_scopes: list[str] = []
    if isinstance(depends, params.Security) and depends.scopes:
        own_oauth_scopes.extend(depends.scopes)
    return get_dependant(
        path=path,
        call=depends.dependency,
        scope=depends.scope,
        own_oauth_scopes=own_oauth_scopes,
    )

  if (reader_ != nullptr) {
    status = reader_->GetTensor(name, out_tensor);
  } else {
    tensorflow::DataType dtype;
    tensorflow::TensorShape shape;
    status = v2_reader_->LookupDtypeAndShape(name, &dtype, &shape);
    if (status.ok()) {
      out_tensor->reset(new Tensor(dtype, shape));
      status = v2_reader_->Lookup(name, out_tensor->get());
      if (!status.ok()) out_tensor->reset();
    }
  }
  if (!status.ok()) {