mBERT-2

This model is a fine-tuned version of bert-base-multilingual-cased on the CERED(Czech Relationship Extraction dataset)-version-2 dataset. (https://lindat.mff.cuni.cz/repository/xmlui/handle/11234/1-3266?locale-attribute=cs) It achieves the following results on the test set:

• Loss: 0.9469
• Accuracy: 0.8845
• Micro Precision: 0.8845
• Micro Recall: 0.8845
• Micro F1: 0.8845
• Macro Precision: 0.8512
• Macro Recall: 0.8349
• Macro F1: 0.8398

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-05
• train_batch_size: 12
• eval_batch_size: 12
• seed: 42
• optimizer: Use 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: 8

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Micro Precision	Micro Recall	Micro F1	Macro Precision	Macro Recall	Macro F1
0.6019	1.0	30148	0.5612	0.8398	0.8398	0.8398	0.8398	0.8271	0.7687	0.7761
0.4904	2.0	60296	0.5216	0.8564	0.8564	0.8564	0.8564	0.8172	0.8079	0.8032
0.3624	3.0	90444	0.5196	0.8656	0.8656	0.8656	0.8656	0.8374	0.8006	0.8094
0.3242	4.0	120592	0.5662	0.8720	0.8720	0.8720	0.8720	0.8468	0.8222	0.8275
0.2522	5.0	150740	0.6456	0.8717	0.8717	0.8717	0.8717	0.8329	0.8374	0.8298
0.1707	6.0	180888	0.7233	0.8764	0.8764	0.8764	0.8764	0.8442	0.8356	0.8343
0.103	7.0	211036	0.8474	0.8794	0.8794	0.8794	0.8794	0.8440	0.8403	0.8385
0.0646	8.0	241184	0.9117	0.8836	0.8836	0.8836	0.8836	0.8465	0.8455	0.8423

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3