BERTInvoiceCzechV03

This model is a fine-tuned version of TomasFAV/BERTInvoiceCzechV0 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0683
• Precision: 0.8635
• Recall: 0.8866
• F1: 0.8749
• Accuracy: 0.9833

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 2
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	20	0.1657	0.6566	0.7456	0.6983	0.9547
No log	2.0	40	0.1222	0.7308	0.7821	0.7556	0.9653
No log	3.0	60	0.1044	0.7512	0.8233	0.7856	0.9707
No log	4.0	80	0.0872	0.7799	0.8532	0.8149	0.9747
No log	5.0	100	0.0897	0.7791	0.8808	0.8268	0.9748
No log	6.0	120	0.0850	0.7725	0.8730	0.8197	0.9742
No log	7.0	140	0.0704	0.8433	0.88	0.8613	0.9812
No log	8.0	160	0.0749	0.8291	0.8649	0.8466	0.9799
No log	9.0	180	0.0752	0.8187	0.8753	0.8461	0.9794
No log	10.0	200	0.0687	0.8440	0.8761	0.8598	0.9815
No log	11.0	220	0.0671	0.8436	0.8816	0.8621	0.9820
No log	12.0	240	0.0711	0.8376	0.8913	0.8636	0.9809
No log	13.0	260	0.0683	0.8638	0.8870	0.8753	0.9833
No log	14.0	280	0.0686	0.8488	0.8870	0.8675	0.9818
No log	15.0	300	0.0690	0.8439	0.8816	0.8623	0.9816
No log	16.0	320	0.0669	0.8469	0.8827	0.8644	0.9819
No log	17.0	340	0.0699	0.8404	0.8897	0.8644	0.9814
No log	18.0	360	0.0684	0.8532	0.8870	0.8698	0.9825
No log	19.0	380	0.0701	0.8408	0.8920	0.8656	0.9815
No log	20.0	400	0.0685	0.8560	0.8909	0.8731	0.9827

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2