2026-09-15-crf-classweights-clean
This model is a fine-tuned version of FacebookAI/xlm-roberta-large on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 9.0167
- Precision: 0.7802
- Recall: 0.8170
- F1: 0.7982
- Accuracy: 0.9722
- Academicdiscipline F1: 0.3810
- Ambiguouslydefinedconcept F1: 0.7126
- Discoursephenomenon F1: 0.6553
- Graphemicphenomenon F1: 0.0
- Languagerelatedterm F1: 0.8463
- Languageresourceinformation F1: 0.7573
- Lexicalphenomenon F1: 0.7078
- Morphologicalphenomenon F1: 0.8116
- Morphosyntacticphenomenon F1: 0.8484
- New Tag F1: 0.8132
- Otherlinguisticterm F1: 0.7301
- Phonologicalphenomenon F1: 0.8689
- Semanticphenomenon F1: 0.7162
- Syntacticphenomenon F1: 0.7993
- Topnode Dummy F1: 0.6951
- Unclassifiedlinguisticconcept F1: 0.8957
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: 8
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- 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
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 8
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy | Academicdiscipline F1 | Ambiguouslydefinedconcept F1 | Discoursephenomenon F1 | Graphemicphenomenon F1 | Languagerelatedterm F1 | Languageresourceinformation F1 | Lexicalphenomenon F1 | Morphologicalphenomenon F1 | Morphosyntacticphenomenon F1 | New Tag F1 | Otherlinguisticterm F1 | Phonologicalphenomenon F1 | Semanticphenomenon F1 | Syntacticphenomenon F1 | Topnode Dummy F1 | Unclassifiedlinguisticconcept F1 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 312.1506 | 1.0 | 667 | 14.2308 | 0.5736 | 0.5128 | 0.5415 | 0.9475 | 0.0 | 0.0 | 0.0 | 0.0 | 0.7340 | 0.4569 | 0.1235 | 0.5443 | 0.6434 | 0.4791 | 0.4146 | 0.7183 | 0.0 | 0.5204 | 0.2474 | 0.0 |
| 59.568 | 2.0 | 1334 | 8.5685 | 0.7454 | 0.7158 | 0.7303 | 0.9653 | 0.5333 | 0.5676 | 0.5118 | 0.0 | 0.8092 | 0.7196 | 0.5543 | 0.7578 | 0.8027 | 0.7481 | 0.6569 | 0.8371 | 0.4966 | 0.7043 | 0.5542 | 0.7876 |
| 28.8857 | 3.0 | 2001 | 7.8781 | 0.7259 | 0.7960 | 0.7593 | 0.9668 | 0.4706 | 0.6102 | 0.6054 | 0.0 | 0.8286 | 0.7156 | 0.5800 | 0.7703 | 0.8256 | 0.7661 | 0.6836 | 0.8516 | 0.6523 | 0.7490 | 0.6463 | 0.7903 |
| 19.8109 | 4.0 | 2668 | 7.2640 | 0.7695 | 0.7953 | 0.7822 | 0.9704 | 0.4211 | 0.7018 | 0.6 | 0.0 | 0.8345 | 0.7672 | 0.6977 | 0.7912 | 0.8411 | 0.8046 | 0.7097 | 0.8510 | 0.6949 | 0.7633 | 0.6663 | 0.8739 |
| 16.3548 | 5.0 | 3335 | 7.7692 | 0.7591 | 0.8204 | 0.7886 | 0.9706 | 0.3333 | 0.7556 | 0.6587 | 0.0 | 0.8359 | 0.7729 | 0.7140 | 0.8051 | 0.8424 | 0.8192 | 0.7234 | 0.8649 | 0.6895 | 0.7720 | 0.6674 | 0.8831 |
| 9.8307 | 6.0 | 4002 | 8.1815 | 0.7628 | 0.8245 | 0.7924 | 0.9709 | 0.4000 | 0.7381 | 0.6950 | 0.0 | 0.8397 | 0.7579 | 0.6833 | 0.8064 | 0.8406 | 0.8081 | 0.7238 | 0.8666 | 0.6983 | 0.7925 | 0.6825 | 0.8957 |
| 7.181 | 7.0 | 4669 | 8.5412 | 0.7888 | 0.8076 | 0.7981 | 0.9725 | 0.4444 | 0.7273 | 0.6545 | 0.0 | 0.8475 | 0.7577 | 0.6837 | 0.7996 | 0.8472 | 0.8155 | 0.7323 | 0.8670 | 0.7013 | 0.7972 | 0.6982 | 0.9052 |
| 5.9979 | 8.0 | 5336 | 9.0167 | 0.7802 | 0.8170 | 0.7982 | 0.9722 | 0.3810 | 0.7126 | 0.6553 | 0.0 | 0.8463 | 0.7573 | 0.7078 | 0.8116 | 0.8484 | 0.8132 | 0.7301 | 0.8689 | 0.7162 | 0.7993 | 0.6951 | 0.8957 |
Framework versions
- Transformers 4.57.6
- Pytorch 2.5.1+cu121
- Datasets 3.2.0
- Tokenizers 0.22.2