CNEC_2_0_Supertypes_slavicbert

This model is a fine-tuned version of DeepPavlov/bert-base-bg-cs-pl-ru-cased on the cnec dataset. It achieves the following results on the evaluation set:

• Loss: 0.2859
• Precision: 0.8603
• Recall: 0.8905
• F1: 0.8752
• Accuracy: 0.9654

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: 2e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.1989	1.0	1799	0.1639	0.8057	0.8410	0.8230	0.9544
0.1512	2.0	3598	0.1679	0.8105	0.8550	0.8322	0.9550
0.1085	3.0	5397	0.1516	0.8253	0.8662	0.8452	0.9582
0.0823	4.0	7196	0.1586	0.8374	0.8765	0.8565	0.9608
0.0529	5.0	8995	0.1802	0.8346	0.8670	0.8505	0.9602
0.0507	6.0	10794	0.2033	0.8249	0.8699	0.8468	0.9603
0.0441	7.0	12593	0.2032	0.8401	0.8724	0.8559	0.9614
0.0271	8.0	14392	0.2247	0.8450	0.8740	0.8593	0.9604
0.0289	9.0	16191	0.2319	0.8385	0.8794	0.8585	0.9613
0.0214	10.0	17990	0.2623	0.8462	0.8703	0.8581	0.9609
0.0173	11.0	19789	0.2553	0.8432	0.8748	0.8587	0.9614
0.0149	12.0	21588	0.2760	0.8582	0.8827	0.8703	0.9631
0.0143	13.0	23387	0.2748	0.8530	0.8843	0.8684	0.9630
0.0095	14.0	25186	0.2796	0.8543	0.8864	0.8701	0.9632
0.0049	15.0	26985	0.2944	0.8512	0.8810	0.8658	0.9627
0.0047	16.0	28784	0.2836	0.8524	0.8848	0.8683	0.9644
0.0047	17.0	30583	0.2902	0.8490	0.8827	0.8655	0.9646
0.0039	18.0	32382	0.2888	0.8603	0.8881	0.8740	0.9650
0.0026	19.0	34181	0.2917	0.8585	0.8897	0.8738	0.9644
0.0047	20.0	35980	0.2859	0.8603	0.8905	0.8752	0.9654

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0