fb-data2vec-finetuned-finance-classification
This model is a fine-tuned version of facebook/data2vec-text-base on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.8993
- Accuracy: 0.8557
- F1: 0.8563
- Precision: 0.8576
- Recall: 0.8557
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: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 15
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | Precision | Recall |
|---|---|---|---|---|---|---|---|
| No log | 1.0 | 285 | 0.6704 | 0.6680 | 0.6262 | 0.7919 | 0.6680 |
| 0.6626 | 2.0 | 570 | 0.4731 | 0.8360 | 0.8350 | 0.8346 | 0.8360 |
| 0.6626 | 3.0 | 855 | 0.4598 | 0.8458 | 0.8454 | 0.8452 | 0.8458 |
| 0.3666 | 4.0 | 1140 | 0.4758 | 0.8360 | 0.8352 | 0.8353 | 0.8360 |
| 0.3666 | 5.0 | 1425 | 0.5683 | 0.8340 | 0.8342 | 0.8353 | 0.8340 |
| 0.2316 | 6.0 | 1710 | 0.6234 | 0.8419 | 0.8421 | 0.8447 | 0.8419 |
| 0.2316 | 7.0 | 1995 | 0.7186 | 0.8379 | 0.8385 | 0.8395 | 0.8379 |
| 0.1523 | 8.0 | 2280 | 0.7268 | 0.8439 | 0.8442 | 0.8455 | 0.8439 |
| 0.0928 | 9.0 | 2565 | 0.7364 | 0.8439 | 0.8452 | 0.8494 | 0.8439 |
| 0.0928 | 10.0 | 2850 | 0.7975 | 0.8478 | 0.8476 | 0.8476 | 0.8478 |
| 0.054 | 11.0 | 3135 | 0.9019 | 0.8498 | 0.8509 | 0.8554 | 0.8498 |
| 0.054 | 12.0 | 3420 | 0.8779 | 0.8538 | 0.8548 | 0.8578 | 0.8538 |
| 0.036 | 13.0 | 3705 | 0.8914 | 0.8617 | 0.8626 | 0.8652 | 0.8617 |
| 0.036 | 14.0 | 3990 | 0.8976 | 0.8538 | 0.8547 | 0.8572 | 0.8538 |
| 0.0232 | 15.0 | 4275 | 0.8993 | 0.8557 | 0.8563 | 0.8576 | 0.8557 |
Framework versions
- Transformers 4.18.0
- Pytorch 1.11.0+cu113
- Datasets 2.1.0
- Tokenizers 0.12.1