Statistics of Evaluating Pretrained Transformer-based Models on the Task of Fine-Grained Named Entity Recognition