hmBERT-CoNLL-cp3

This model is a fine-tuned version of dbmdz/bert-base-historic-multilingual-cased on the conll2003 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0572
• Precision: 0.9121
• Recall: 0.9243
• F1: 0.9182
• Accuracy: 0.9862

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: 5e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	0.06	25	0.4115	0.3643	0.3728	0.3685	0.9007
No log	0.11	50	0.2243	0.6393	0.6908	0.6641	0.9460
No log	0.17	75	0.1617	0.7319	0.7637	0.7475	0.9580
No log	0.23	100	0.1544	0.7282	0.7637	0.7455	0.9585
No log	0.28	125	0.1341	0.7595	0.8117	0.7847	0.9644
No log	0.34	150	0.1221	0.7980	0.8251	0.8114	0.9693
No log	0.4	175	0.1013	0.7968	0.8344	0.8152	0.9719
No log	0.46	200	0.1076	0.8265	0.8403	0.8333	0.9732
No log	0.51	225	0.0883	0.8453	0.8635	0.8543	0.9763
No log	0.57	250	0.0973	0.8439	0.8633	0.8535	0.9763
No log	0.63	275	0.0883	0.8497	0.8655	0.8575	0.9765
No log	0.68	300	0.0879	0.8462	0.8642	0.8551	0.9766
No log	0.74	325	0.0781	0.8592	0.8834	0.8711	0.9787
No log	0.8	350	0.0725	0.8697	0.8928	0.8811	0.9803
No log	0.85	375	0.0755	0.8687	0.8943	0.8813	0.9807
No log	0.91	400	0.0666	0.8781	0.9004	0.8891	0.9822
No log	0.97	425	0.0658	0.8877	0.8995	0.8936	0.9823
No log	1.03	450	0.0645	0.8951	0.9036	0.8993	0.9837
No log	1.08	475	0.0697	0.8864	0.9039	0.8951	0.9831
0.1392	1.14	500	0.0688	0.8824	0.8994	0.8908	0.9824
0.1392	1.2	525	0.0681	0.8950	0.9049	0.8999	0.9827
0.1392	1.25	550	0.0676	0.8855	0.8977	0.8915	0.9823
0.1392	1.31	575	0.0618	0.8940	0.9088	0.9014	0.9842
0.1392	1.37	600	0.0644	0.8945	0.9076	0.9010	0.9840
0.1392	1.42	625	0.0641	0.8936	0.9086	0.9010	0.9837
0.1392	1.48	650	0.0619	0.8969	0.9120	0.9044	0.9846
0.1392	1.54	675	0.0608	0.9045	0.9105	0.9075	0.9848
0.1392	1.59	700	0.0624	0.9038	0.9143	0.9091	0.9851
0.1392	1.65	725	0.0596	0.9062	0.9170	0.9116	0.9852
0.1392	1.71	750	0.0580	0.8995	0.9143	0.9069	0.9848
0.1392	1.77	775	0.0582	0.9082	0.9172	0.9127	0.9858
0.1392	1.82	800	0.0588	0.9024	0.9179	0.9101	0.9852
0.1392	1.88	825	0.0592	0.9020	0.9219	0.9119	0.9856
0.1392	1.94	850	0.0600	0.9054	0.9182	0.9118	0.9852
0.1392	1.99	875	0.0568	0.9068	0.9202	0.9135	0.9861
0.1392	2.05	900	0.0571	0.9131	0.9212	0.9171	0.9861
0.1392	2.11	925	0.0577	0.9110	0.9204	0.9157	0.9858
0.1392	2.16	950	0.0605	0.9127	0.9243	0.9185	0.9860
0.1392	2.22	975	0.0575	0.9109	0.9224	0.9166	0.9867
0.0392	2.28	1000	0.0572	0.9121	0.9243	0.9182	0.9862
0.0392	2.33	1025	0.0567	0.9171	0.9253	0.9212	0.9870
0.0392	2.39	1050	0.0570	0.9193	0.9295	0.9244	0.9871
0.0392	2.45	1075	0.0584	0.9155	0.9276	0.9215	0.9867
0.0392	2.51	1100	0.0591	0.9168	0.9286	0.9227	0.9867
0.0392	2.56	1125	0.0577	0.9182	0.9312	0.9246	0.9874
0.0392	2.62	1150	0.0570	0.9184	0.9283	0.9233	0.9870
0.0392	2.68	1175	0.0563	0.9191	0.9298	0.9245	0.9872
0.0392	2.73	1200	0.0565	0.9180	0.9313	0.9246	0.9872
0.0392	2.79	1225	0.0559	0.9190	0.9298	0.9244	0.9873
0.0392	2.85	1250	0.0562	0.9185	0.9293	0.9239	0.9873
0.0392	2.9	1275	0.0564	0.9175	0.9285	0.9230	0.9872
0.0392	2.96	1300	0.0563	0.9181	0.9295	0.9237	0.9873

Framework versions

• Transformers 4.20.1
• Pytorch 1.12.0
• Datasets 2.4.0
• Tokenizers 0.12.1