turkish-hs-group-prediction

This model is a fine-tuned version of dbmdz/bert-base-turkish-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1554
• Accuracy: 0.9734
• Macro F1: 0.9634

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-06
• train_batch_size: 16
• eval_batch_size: 20
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Macro F1
2.0921	0.1391	100	1.7605	0.4421	0.1828
1.5316	0.2782	200	1.1793	0.7269	0.5084
0.9739	0.4172	300	0.6451	0.8912	0.7465
0.6016	0.5563	400	0.3865	0.9147	0.8015
0.3992	0.6954	500	0.2805	0.9421	0.8851
0.2769	0.8345	600	0.2051	0.9538	0.9163
0.1817	0.9736	700	0.1844	0.9562	0.9230
0.1648	1.1127	800	0.1579	0.9593	0.9278
0.1282	1.2517	900	0.1479	0.9624	0.9426
0.1299	1.3908	1000	0.1426	0.9656	0.9467
0.1115	1.5299	1100	0.1288	0.9679	0.9556
0.1144	1.6690	1200	0.1434	0.9671	0.9542
0.0742	1.8081	1300	0.1393	0.9687	0.9459
0.1048	1.9471	1400	0.1339	0.9703	0.9571
0.0775	2.0862	1500	0.1416	0.9695	0.9555
0.0783	2.2253	1600	0.1150	0.9703	0.9580
0.0738	2.3644	1700	0.1133	0.9742	0.9628
0.0717	2.5035	1800	0.1355	0.9718	0.9593
0.0685	2.6426	1900	0.1195	0.9718	0.9589
0.0717	2.7816	2000	0.1249	0.9750	0.9643
0.0718	2.9207	2100	0.1308	0.9734	0.9638
0.0429	3.0598	2200	0.1366	0.9734	0.9639
0.0486	3.1989	2300	0.1256	0.9734	0.9636
0.0577	3.3380	2400	0.1238	0.9718	0.9627
0.0528	3.4771	2500	0.1251	0.9734	0.9639
0.0423	3.6161	2600	0.1327	0.9734	0.9639
0.0682	3.7552	2700	0.1434	0.9742	0.9649
0.0438	3.8943	2800	0.1357	0.9734	0.9641
0.046	4.0334	2900	0.1469	0.9718	0.9624
0.0481	4.1725	3000	0.1378	0.9726	0.9632
0.0304	4.3115	3100	0.1403	0.9734	0.9639
0.0406	4.4506	3200	0.1359	0.9750	0.9655
0.0429	4.5897	3300	0.1315	0.9734	0.9641
0.047	4.7288	3400	0.1365	0.9726	0.9633
0.0345	4.8679	3500	0.1497	0.9734	0.9643
0.0246	5.0070	3600	0.1507	0.9726	0.9630
0.0275	5.1460	3700	0.1443	0.9742	0.9646
0.0229	5.2851	3800	0.1554	0.9734	0.9641
0.0226	5.4242	3900	0.1543	0.9742	0.9649
0.045	5.5633	4000	0.1588	0.9726	0.9629
0.0321	5.7024	4100	0.1333	0.9742	0.9640
0.0312	5.8414	4200	0.1379	0.9742	0.9643
0.0191	5.9805	4300	0.1542	0.9734	0.9641
0.0236	6.1196	4400	0.1600	0.9742	0.9642
0.0225	6.2587	4500	0.1567	0.9742	0.9645
0.0255	6.3978	4600	0.1469	0.9726	0.9628
0.0193	6.5369	4700	0.1582	0.9726	0.9625
0.0228	6.6759	4800	0.1609	0.9742	0.9645
0.0302	6.8150	4900	0.1489	0.9742	0.9645
0.0332	6.9541	5000	0.1480	0.9734	0.9632
0.0237	7.0932	5100	0.1495	0.9757	0.9662
0.0224	7.2323	5200	0.1410	0.9750	0.9654
0.0085	7.3713	5300	0.1443	0.9750	0.9654
0.0106	7.5104	5400	0.1476	0.9750	0.9652
0.0215	7.6495	5500	0.1435	0.9734	0.9634
0.0198	7.7886	5600	0.1474	0.9750	0.9655
0.0367	7.9277	5700	0.1517	0.9734	0.9635
0.0171	8.0668	5800	0.1464	0.9750	0.9655
0.0133	8.2058	5900	0.1514	0.9750	0.9654
0.02	8.3449	6000	0.1482	0.9734	0.9634
0.0186	8.4840	6100	0.1480	0.9726	0.9622
0.0199	8.6231	6200	0.1472	0.9750	0.9654
0.0185	8.7622	6300	0.1513	0.9750	0.9654
0.0117	8.9013	6400	0.1624	0.9734	0.9634
0.0211	9.0403	6500	0.1620	0.9742	0.9645
0.0095	9.1794	6600	0.1604	0.9742	0.9645
0.0222	9.3185	6700	0.1597	0.9734	0.9634
0.0182	9.4576	6800	0.1591	0.9734	0.9634
0.0158	9.5967	6900	0.1567	0.9734	0.9634
0.0146	9.7357	7000	0.1554	0.9734	0.9634
0.0112	9.8748	7100	0.1554	0.9734	0.9634

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 2.14.4
• Tokenizers 0.21.1