turkish-hs-2class-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: 0.5811
• Accuracy: 0.8879
• Macro F1: 0.8829

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
0.5915	0.1460	100	0.4860	0.7703	0.7399
0.4414	0.2920	200	0.3925	0.8122	0.8044
0.3889	0.4380	300	0.3714	0.8323	0.8180
0.3713	0.5839	400	0.3467	0.8487	0.8376
0.3442	0.7299	500	0.3240	0.8569	0.8505
0.3427	0.8759	600	0.3084	0.8678	0.8612
0.3295	1.0219	700	0.3067	0.8651	0.8598
0.282	1.1679	800	0.3136	0.8678	0.8631
0.29	1.3139	900	0.3050	0.8696	0.8604
0.2993	1.4599	1000	0.3068	0.8660	0.8614
0.2835	1.6058	1100	0.2951	0.8769	0.8697
0.2751	1.7518	1200	0.3079	0.8742	0.8694
0.2717	1.8978	1300	0.2912	0.8806	0.8732
0.2615	2.0438	1400	0.2887	0.8833	0.8779
0.2413	2.1898	1500	0.3005	0.8824	0.8779
0.2446	2.3358	1600	0.3081	0.8788	0.8742
0.2022	2.4818	1700	0.3216	0.8815	0.8754
0.245	2.6277	1800	0.3109	0.8861	0.8812
0.2341	2.7737	1900	0.3369	0.8742	0.8712
0.239	2.9197	2000	0.2988	0.8879	0.8828
0.2228	3.0657	2100	0.3174	0.8824	0.8751
0.1819	3.2117	2200	0.3371	0.8815	0.8747
0.1841	3.3577	2300	0.3331	0.8861	0.8810
0.2208	3.5036	2400	0.3237	0.8906	0.8864
0.1934	3.6496	2500	0.3337	0.8897	0.8852
0.1947	3.7956	2600	0.3446	0.8861	0.8798
0.2043	3.9416	2700	0.3472	0.8888	0.8836
0.154	4.0876	2800	0.3549	0.8888	0.8842
0.1657	4.2336	2900	0.3664	0.8833	0.8764
0.1456	4.3796	3000	0.3905	0.8879	0.8829
0.1774	4.5255	3100	0.3801	0.8861	0.8807
0.1971	4.6715	3200	0.3943	0.8842	0.8796
0.182	4.8175	3300	0.3607	0.8906	0.8855
0.1751	4.9635	3400	0.3829	0.8879	0.8828
0.123	5.1095	3500	0.4142	0.8833	0.8776
0.1678	5.2555	3600	0.4128	0.8888	0.8833
0.1339	5.4015	3700	0.4287	0.8879	0.8820
0.164	5.5474	3800	0.4406	0.8842	0.8798
0.1521	5.6934	3900	0.4191	0.8906	0.8855
0.1476	5.8394	4000	0.4343	0.8806	0.8750
0.171	5.9854	4100	0.4305	0.8861	0.8802
0.1171	6.1314	4200	0.4552	0.8824	0.8780
0.1237	6.2774	4300	0.4561	0.8870	0.8808
0.1261	6.4234	4400	0.4696	0.8833	0.8778
0.1345	6.5693	4500	0.4848	0.8888	0.8830
0.1172	6.7153	4600	0.5006	0.8842	0.8797
0.1574	6.8613	4700	0.4770	0.8833	0.8782
0.1144	7.0073	4800	0.4923	0.8851	0.8803
0.1129	7.1533	4900	0.5090	0.8861	0.8811
0.1086	7.2993	5000	0.5437	0.8879	0.8833
0.1155	7.4453	5100	0.5360	0.8861	0.8814
0.121	7.5912	5200	0.5260	0.8851	0.8790
0.0984	7.7372	5300	0.5583	0.8806	0.8756
0.1267	7.8832	5400	0.5423	0.8861	0.8809
0.1087	8.0292	5500	0.5492	0.8879	0.8832
0.1117	8.1752	5600	0.5604	0.8851	0.8808
0.112	8.3212	5700	0.5614	0.8833	0.8784
0.1037	8.4672	5800	0.5707	0.8815	0.8773
0.0813	8.6131	5900	0.5736	0.8806	0.8761
0.1034	8.7591	6000	0.5672	0.8842	0.8793
0.0985	8.9051	6100	0.5761	0.8824	0.8778
0.1043	9.0511	6200	0.5739	0.8833	0.8785
0.082	9.1971	6300	0.5752	0.8851	0.8802
0.1019	9.3431	6400	0.5816	0.8815	0.8767
0.0743	9.4891	6500	0.5814	0.8842	0.8793
0.1045	9.6350	6600	0.5843	0.8815	0.8767
0.0743	9.7810	6700	0.5831	0.8833	0.8784
0.1138	9.9270	6800	0.5811	0.8879	0.8829

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0