icdar23-entrydetector_plaintext_breaks_indents_left_diff_right_ref

This model is a fine-tuned version of HueyNemud/das22-10-camembert_pretrained on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0078
• Ebegin: {'precision': 0.9920303605313093, 'recall': 0.9830763444904099, 'f1': 0.9875330562901399, 'number': 2659}
• Eend: {'precision': 0.9958443520967133, 'recall': 0.9850523168908819, 'f1': 0.9904189366898367, 'number': 2676}
• Overall Precision: 0.9939
• Overall Recall: 0.9841
• Overall F1: 0.9890
• Overall Accuracy: 0.9982

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 7500

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	0.07	300	0.0314	0.9572	0.9870	0.9719	0.9956
0.1574	0.14	600	0.0145	0.9897	0.9834	0.9866	0.9979
0.1574	0.21	900	0.0098	0.9896	0.9917	0.9907	0.9985
0.0161	0.29	1200	0.0079	0.9919	0.9921	0.9920	0.9987
0.0107	0.36	1500	0.0072	0.9895	0.9928	0.9911	0.9986
0.0107	0.43	1800	0.0116	0.9900	0.9877	0.9888	0.9981
0.0114	0.5	2100	0.0069	0.9965	0.9898	0.9931	0.9988
0.0114	0.57	2400	0.0055	0.9955	0.9907	0.9931	0.9989
0.0082	0.64	2700	0.0051	0.9870	0.9956	0.9913	0.9985
0.0062	0.72	3000	0.0046	0.9903	0.9957	0.9930	0.9988
0.0062	0.79	3300	0.0038	0.9957	0.9929	0.9943	0.9990
0.0051	0.86	3600	0.0038	0.9956	0.9943	0.9949	0.9992
0.0051	0.93	3900	0.0047	0.9902	0.9942	0.9921	0.9987
0.0041	1.0	4200	0.0035	0.9979	0.9917	0.9948	0.9991
0.0029	1.07	4500	0.0036	0.9973	0.9926	0.9949	0.9992
0.0029	1.14	4800	0.0038	0.9969	0.9916	0.9942	0.9990
0.0034	1.22	5100	0.0036	0.9953	0.9935	0.9944	0.9991

Framework versions

• Transformers 4.26.1
• Pytorch 1.13.1+cu116
• Datasets 2.9.0
• Tokenizers 0.13.2