chunked/nltk_chunking/model_doc_detr/chunk_31.txt

Next, standard cross-entropy (for
the classes) and a linear combination of the L1 and generalized IoU loss (for the
bounding boxes) are used to optimize the parameters of the model.