I used longformer for classification purpose where the document is longer than 4096 (>10k). It runs fine and gives okay result. However I wonder is there is something that I am missing? or did the transformers library automatically chunk my long document?