CNEC_2_0_ext_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: 0.2252
• Precision: 0.8578
• Recall: 0.8864
• F1: 0.8719
• Accuracy: 0.9697

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: 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: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.1347	4.46	1000	0.1375	0.8279	0.8620	0.8446	0.9656
0.0681	8.93	2000	0.1519	0.8345	0.8710	0.8524	0.9668
0.0406	13.39	3000	0.1663	0.8519	0.8789	0.8652	0.9679
0.0276	17.86	4000	0.1719	0.8623	0.8888	0.8754	0.9690
0.02	22.32	5000	0.1920	0.8505	0.8809	0.8654	0.9686
0.015	26.79	6000	0.1984	0.8570	0.8893	0.8729	0.9693
0.0108	31.25	7000	0.2048	0.8587	0.8864	0.8723	0.9692
0.0092	35.71	8000	0.2179	0.8606	0.8888	0.8745	0.9696
0.0076	40.18	9000	0.2252	0.8564	0.8878	0.8718	0.9696
0.0057	44.64	10000	0.2262	0.8571	0.8873	0.8720	0.9698
0.0054	49.11	11000	0.2252	0.8578	0.8864	0.8719	0.9697

Framework versions

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