Reproduce paper results

Dear Mayer,

Running your code I don't get the same results as those reported in the paper. I used the run_train.sh script as-is. I show the two results in the following table. Also, I attach the outputs (predictions and evaluation results) for the three test sets.

Please, I need to get accurate results because I want to compare a model of mine on your dataset and also run your model on a dataset of mine.

Source	Embedding	Neoplasm				Glaucoma				Mixed
		f1	F1	C-F1	E-F1	f1	F1	C-F1	E-F1	f1	F1	C-F1	E-F1
Paper	fine-tuning SciBERT	0.90	0.87	0.88	0.92	0.91	0.89	0.93	0.91	0.91	0.88	0.90	0.93
My run	fine-tuning SciBERT	0.90	0.83	0.75	0.91	0.92	0.84	0.83	0.92	0.90	0.83	0.77	0.91

sequence_tagging_predictions_neoplasm.conll

sequence_tagging_predictions_mixed.conll

sequence_tagging_predictions_glaucoma.conll

eval_results_neoplasm.txt

eval_results_mixed.txt

eval_results_glaucoma.txt