updateHighlightInfo(highlightInfo, width);
                final HttpSession session = req.getSession(false);
                if (session != null) {
                    session.setAttribute(SCREEN_WIDTH, width);
                }
            } else {
                final HttpSession session = req.getSession(false);
                if (session != null) {

 */
@Experimental
@Immutable
public interface DependencyResolverRequest {

    enum RequestType {
        COLLECT,
        FLATTEN,
        RESOLVE
    }

    @Nonnull
    Session getSession();

    @Nonnull
    RequestType getRequestType();

    @Nonnull
    Optional<Project> getProject();

    @Nonnull
    Optional<Artifact> getRootArtifact();

    @Nonnull

            try {
                String interpolated =
                        ssi.interpolate(interpolatedModel.getParent().getVersion());
                interpolatedModel.getParent().setVersion(interpolated);
            } catch (Exception e) {
                ModelProblemCollectorRequest mpcr = new ModelProblemCollectorRequest(
                                Severity.ERROR, ModelProblem.Version.BASE)

            final EntityCreator<T, SearchResponse, SearchHit> creator) {
        return search(index, searchRequestBuilder -> {
            searchRequestBuilder.setVersion(true);
            return condition.build(searchRequestBuilder);
        }, (queryBuilder, execTime, searchResponse) -> searchResponse.map(response -> {
            final SearchHit[] hits = response.getHits().getHits();

    *   symbolic tensors outside the scope of the model are used in custom loss
        functions. The flag can be disabled for these cases and ideally the
        usage pattern will need to be fixed.
    *   Mark Keras `set_session` as `compat.v1` only.
    *   `tf.keras.estimator.model_to_estimator` now supports exporting to
        `tf.train.Checkpoint format`, which allows the saved checkpoints to be
        compatible with `model.load_weights`.