CNEC_1_1_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: 0.2572
• Precision: 0.8607
• Recall: 0.8915
• F1: 0.8758
• Accuracy: 0.9627

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.3946	1.72	500	0.1925	0.7835	0.8471	0.8141	0.9467
0.1653	3.44	1000	0.1627	0.8340	0.8675	0.8504	0.9572
0.1183	5.15	1500	0.1700	0.8378	0.8808	0.8588	0.9595
0.0869	6.87	2000	0.1901	0.8554	0.8728	0.8640	0.9589
0.0661	8.59	2500	0.2037	0.8482	0.8867	0.8670	0.9595
0.053	10.31	3000	0.2011	0.8460	0.8867	0.8659	0.9609
0.043	12.03	3500	0.2216	0.8555	0.8888	0.8718	0.9593
0.0358	13.75	4000	0.2245	0.8492	0.8878	0.8680	0.9603
0.0296	15.46	4500	0.2401	0.8513	0.8872	0.8689	0.9603
0.0264	17.18	5000	0.2415	0.8564	0.8862	0.8710	0.9610
0.0212	18.9	5500	0.2570	0.8557	0.8872	0.8712	0.9622
0.0205	20.62	6000	0.2540	0.8567	0.8883	0.8722	0.9616
0.0167	22.34	6500	0.2573	0.8568	0.8894	0.8728	0.9614
0.0161	24.05	7000	0.2572	0.8607	0.8915	0.8758	0.9627

Framework versions

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