<!-- 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. -->
mental_bert_classifier
This model is a fine-tuned version of Zamoranesis/mental_bert on hackathon-somos-nlp-2023/DiagTrast. It achieves the following results on the evaluation set:
- Loss: 0.2426
- F1 Class 0: 0.8852
- F1 Class 1: 0.9512
- F1 Class 2: 0.8421
- F1 Class 3: 0.8539
- F1 Class 4: 0.9412
- F1: 0.8947
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: 0.005
- train_batch_size: 64
- eval_batch_size: 64
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- lr_scheduler_warmup_steps: 100
- training_steps: 1000
Training results
| Training Loss | Epoch | Step | Validation Loss | F1 Class 0 | F1 Class 1 | F1 Class 2 | F1 Class 3 | F1 Class 4 | F1 |
|---|---|---|---|---|---|---|---|---|---|
| 0.7979 | 6.25 | 100 | 0.3557 | 0.8421 | 0.9512 | 0.8214 | 0.8889 | 0.9091 | 0.8825 |
| 0.2559 | 12.5 | 200 | 0.2823 | 0.9333 | 0.9412 | 0.8364 | 0.8602 | 0.9362 | 0.9015 |
| 0.1963 | 18.75 | 300 | 0.2610 | 0.9180 | 0.9756 | 0.7778 | 0.8352 | 0.9231 | 0.8859 |
| 0.1717 | 25.0 | 400 | 0.2534 | 0.9180 | 0.9630 | 0.8 | 0.8261 | 0.9412 | 0.8897 |
| 0.1511 | 31.25 | 500 | 0.2476 | 0.8667 | 0.9512 | 0.8148 | 0.8298 | 0.96 | 0.8845 |
| 0.1501 | 37.5 | 600 | 0.2513 | 0.9 | 0.9630 | 0.8 | 0.8261 | 0.9231 | 0.8824 |
| 0.1427 | 43.75 | 700 | 0.2581 | 0.9180 | 0.9756 | 0.8475 | 0.8810 | 0.8889 | 0.9022 |
| 0.1457 | 50.0 | 800 | 0.2428 | 0.8852 | 0.9512 | 0.8 | 0.8261 | 0.92 | 0.8765 |
| 0.1311 | 56.25 | 900 | 0.2462 | 0.9 | 0.9512 | 0.8 | 0.8352 | 0.9231 | 0.8819 |
| 0.1346 | 62.5 | 1000 | 0.2426 | 0.8852 | 0.9512 | 0.8421 | 0.8539 | 0.9412 | 0.8947 |
Framework versions
- Transformers 4.33.3
- Pytorch 2.0.1+cu118
- Datasets 2.14.5
- Tokenizers 0.13.3
