twitter-roberta-base-CoNLL

This model is a fine-tuned version of cardiffnlp/twitter-roberta-base on the conll2003 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0423
• Precision: 0.9531
• Recall: 0.9613
• F1: 0.9572
• Accuracy: 0.9926

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: 6e-05
• train_batch_size: 64
• eval_batch_size: 1024
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	0.11	25	0.2063	0.6517	0.6659	0.6587	0.9386
No log	0.23	50	0.0810	0.8373	0.8766	0.8565	0.9771
No log	0.34	75	0.0651	0.8937	0.9058	0.8997	0.9827
No log	0.45	100	0.0537	0.9014	0.9135	0.9074	0.9849
No log	0.57	125	0.0464	0.9097	0.9244	0.9170	0.9867
No log	0.68	150	0.0423	0.9243	0.9350	0.9296	0.9885
No log	0.8	175	0.0381	0.9250	0.9438	0.9343	0.9900
No log	0.91	200	0.0388	0.9264	0.9446	0.9354	0.9896
No log	1.02	225	0.0394	0.9328	0.9441	0.9384	0.9898
No log	1.14	250	0.0423	0.9348	0.9458	0.9403	0.9896
No log	1.25	275	0.0432	0.9304	0.9406	0.9355	0.9892
No log	1.36	300	0.0382	0.9393	0.9473	0.9433	0.9901
No log	1.48	325	0.0381	0.9326	0.9504	0.9414	0.9901
No log	1.59	350	0.0387	0.9337	0.9524	0.9429	0.9902
No log	1.7	375	0.0365	0.9404	0.9475	0.9439	0.9901
No log	1.82	400	0.0382	0.9431	0.9517	0.9474	0.9905
No log	1.93	425	0.0373	0.9399	0.9524	0.9461	0.9903
No log	2.05	450	0.0367	0.9440	0.9556	0.9497	0.9910
No log	2.16	475	0.0396	0.9400	0.9551	0.9475	0.9907
0.0771	2.27	500	0.0353	0.9442	0.9574	0.9508	0.9912
0.0771	2.39	525	0.0394	0.9401	0.9507	0.9454	0.9906
0.0771	2.5	550	0.0370	0.9447	0.9522	0.9485	0.9910
0.0771	2.61	575	0.0352	0.9404	0.9541	0.9472	0.9908
0.0771	2.73	600	0.0386	0.9345	0.9554	0.9448	0.9908
0.0771	2.84	625	0.0366	0.9428	0.9576	0.9502	0.9916
0.0771	2.95	650	0.0353	0.9427	0.9546	0.9486	0.9913
0.0771	3.07	675	0.0359	0.9412	0.9544	0.9478	0.9911
0.0771	3.18	700	0.0356	0.9476	0.9593	0.9534	0.9920
0.0771	3.3	725	0.0345	0.9484	0.9586	0.9535	0.9918
0.0771	3.41	750	0.0345	0.9427	0.9557	0.9492	0.9916
0.0771	3.52	775	0.0364	0.9389	0.9569	0.9478	0.9914
0.0771	3.64	800	0.0360	0.9430	0.9584	0.9507	0.9915
0.0771	3.75	825	0.0387	0.9458	0.9552	0.9505	0.9915
0.0771	3.86	850	0.0347	0.9468	0.9576	0.9521	0.9917
0.0771	3.98	875	0.0357	0.9445	0.9574	0.9509	0.9915
0.0771	4.09	900	0.0382	0.9464	0.9578	0.9521	0.9918
0.0771	4.2	925	0.0391	0.9475	0.9562	0.9518	0.9918
0.0771	4.32	950	0.0428	0.9466	0.9547	0.9506	0.9912
0.0771	4.43	975	0.0404	0.9459	0.9554	0.9506	0.9913
0.0118	4.55	1000	0.0403	0.9375	0.9549	0.9461	0.9909
0.0118	4.66	1025	0.0369	0.9482	0.9586	0.9534	0.9919
0.0118	4.77	1050	0.0374	0.9457	0.9584	0.9520	0.9918
0.0118	4.89	1075	0.0359	0.9507	0.9571	0.9539	0.9923
0.0118	5.0	1100	0.0373	0.9453	0.9594	0.9523	0.9919
0.0118	5.11	1125	0.0370	0.9499	0.9594	0.9546	0.9924
0.0118	5.23	1150	0.0388	0.9510	0.9601	0.9555	0.9922
0.0118	5.34	1175	0.0395	0.9486	0.9559	0.9522	0.9920
0.0118	5.45	1200	0.0391	0.9495	0.9591	0.9543	0.9924
0.0118	5.57	1225	0.0378	0.9517	0.9588	0.9552	0.9923
0.0118	5.68	1250	0.0388	0.9515	0.9615	0.9565	0.9924
0.0118	5.8	1275	0.0384	0.9512	0.9610	0.9560	0.9924
0.0118	5.91	1300	0.0395	0.9530	0.9613	0.9571	0.9924
0.0118	6.02	1325	0.0408	0.9499	0.9569	0.9534	0.9919
0.0118	6.14	1350	0.0412	0.9481	0.9616	0.9548	0.9922
0.0118	6.25	1375	0.0413	0.9521	0.9591	0.9556	0.9924
0.0118	6.36	1400	0.0412	0.9466	0.9584	0.9525	0.9917
0.0118	6.48	1425	0.0405	0.9504	0.9608	0.9556	0.9921
0.0118	6.59	1450	0.0400	0.9517	0.9615	0.9566	0.9925
0.0118	6.7	1475	0.0398	0.9510	0.9594	0.9552	0.9923
0.0049	6.82	1500	0.0395	0.9523	0.9615	0.9569	0.9925
0.0049	6.93	1525	0.0392	0.9520	0.9623	0.9571	0.9927
0.0049	7.05	1550	0.0390	0.9511	0.9593	0.9552	0.9923
0.0049	7.16	1575	0.0393	0.9520	0.9611	0.9565	0.9925
0.0049	7.27	1600	0.0389	0.9512	0.9613	0.9562	0.9925
0.0049	7.39	1625	0.0405	0.9518	0.9613	0.9565	0.9924
0.0049	7.5	1650	0.0410	0.9512	0.9606	0.9559	0.9925
0.0049	7.61	1675	0.0408	0.9526	0.9613	0.9569	0.9925
0.0049	7.73	1700	0.0436	0.9482	0.9610	0.9545	0.9922
0.0049	7.84	1725	0.0419	0.9495	0.9625	0.9560	0.9924
0.0049	7.95	1750	0.0429	0.9525	0.9618	0.9571	0.9926
0.0049	8.07	1775	0.0419	0.9509	0.9615	0.9562	0.9924
0.0049	8.18	1800	0.0422	0.9510	0.9601	0.9555	0.9923
0.0049	8.3	1825	0.0417	0.9521	0.9603	0.9562	0.9924
0.0049	8.41	1850	0.0415	0.9529	0.9611	0.9570	0.9925
0.0049	8.52	1875	0.0416	0.9523	0.9611	0.9567	0.9924
0.0049	8.64	1900	0.0419	0.9504	0.9608	0.9556	0.9922
0.0049	8.75	1925	0.0417	0.9520	0.9610	0.9564	0.9924
0.0049	8.86	1950	0.0419	0.9535	0.9621	0.9578	0.9926
0.0049	8.98	1975	0.0422	0.9531	0.9620	0.9575	0.9927
0.0022	9.09	2000	0.0423	0.9531	0.9613	0.9572	0.9926
0.0022	9.2	2025	0.0426	0.9520	0.9615	0.9567	0.9925
0.0022	9.32	2050	0.0425	0.9515	0.9606	0.9560	0.9925
0.0022	9.43	2075	0.0422	0.9517	0.9613	0.9565	0.9925
0.0022	9.55	2100	0.0423	0.9513	0.9606	0.9560	0.9925
0.0022	9.66	2125	0.0424	0.9513	0.9605	0.9559	0.9925
0.0022	9.77	2150	0.0423	0.9522	0.9611	0.9566	0.9925
0.0022	9.89	2175	0.0423	0.9522	0.9613	0.9567	0.9925
0.0022	10.0	2200	0.0422	0.9525	0.9616	0.9570	0.9925

Framework versions

• Transformers 4.20.1
• Pytorch 1.12.0
• Datasets 2.3.2
• Tokenizers 0.12.1