<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
rubert-tiny2-srl
This model is a fine-tuned version of cointegrated/rubert-tiny2 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.2041
- Addressee Precision: 0.7273
- Addressee Recall: 0.8
- Addressee F1: 0.7619
- Addressee Number: 10
- Benefactive Precision: 0.0
- Benefactive Recall: 0.0
- Benefactive F1: 0.0
- Benefactive Number: 1
- Causator Precision: 0.8824
- Causator Recall: 0.8333
- Causator F1: 0.8571
- Causator Number: 18
- Cause Precision: 0.6667
- Cause Recall: 0.1538
- Cause F1: 0.25
- Cause Number: 13
- Contrsubject Precision: 0.6667
- Contrsubject Recall: 0.3333
- Contrsubject F1: 0.4444
- Contrsubject Number: 6
- Deliberative Precision: 1.0
- Deliberative Recall: 0.4
- Deliberative F1: 0.5714
- Deliberative Number: 5
- Experiencer Precision: 0.7660
- Experiencer Recall: 0.8
- Experiencer F1: 0.7826
- Experiencer Number: 90
- Object Precision: 0.7576
- Object Recall: 0.6868
- Object F1: 0.7205
- Object Number: 182
- Predicate Precision: 0.9713
- Predicate Recall: 0.9967
- Predicate F1: 0.9839
- Predicate Number: 306
- Overall Precision: 0.8719
- Overall Recall: 0.8415
- Overall F1: 0.8565
- Overall Accuracy: 0.9429
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: 0.00018632464179881193
- train_batch_size: 4
- eval_batch_size: 1
- seed: 755657
- gradient_accumulation_steps: 8
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.02
- num_epochs: 2
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Addressee Precision | Addressee Recall | Addressee F1 | Addressee Number | Benefactive Precision | Benefactive Recall | Benefactive F1 | Benefactive Number | Causator Precision | Causator Recall | Causator F1 | Causator Number | Cause Precision | Cause Recall | Cause F1 | Cause Number | Contrsubject Precision | Contrsubject Recall | Contrsubject F1 | Contrsubject Number | Deliberative Precision | Deliberative Recall | Deliberative F1 | Deliberative Number | Experiencer Precision | Experiencer Recall | Experiencer F1 | Experiencer Number | Object Precision | Object Recall | Object F1 | Object Number | Predicate Precision | Predicate Recall | Predicate F1 | Predicate Number | Overall Precision | Overall Recall | Overall F1 | Overall Accuracy |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.2845 | 1.0 | 181 | 0.2356 | 0.8 | 0.8 | 0.8000 | 10 | 0.0 | 0.0 | 0.0 | 1 | 0.7895 | 0.8333 | 0.8108 | 18 | 0.0 | 0.0 | 0.0 | 13 | 0.0 | 0.0 | 0.0 | 6 | 0.0 | 0.0 | 0.0 | 5 | 0.7320 | 0.7889 | 0.7594 | 90 | 0.7740 | 0.6209 | 0.6890 | 182 | 0.9744 | 0.9935 | 0.9838 | 306 | 0.875 | 0.8098 | 0.8412 | 0.9376 |
| 0.1875 | 1.99 | 362 | 0.2041 | 0.7273 | 0.8 | 0.7619 | 10 | 0.0 | 0.0 | 0.0 | 1 | 0.8824 | 0.8333 | 0.8571 | 18 | 0.6667 | 0.1538 | 0.25 | 13 | 0.6667 | 0.3333 | 0.4444 | 6 | 1.0 | 0.4 | 0.5714 | 5 | 0.7660 | 0.8 | 0.7826 | 90 | 0.7576 | 0.6868 | 0.7205 | 182 | 0.9713 | 0.9967 | 0.9839 | 306 | 0.8719 | 0.8415 | 0.8565 | 0.9429 |
Framework versions
- Transformers 4.28.1
- Pytorch 2.0.0+cu117
- Datasets 2.11.0
- Tokenizers 0.13.3
