DistilRoberta-financial-sentiment
This model is a fine-tuned version of distilroberta-base on the financial_phrasebank dataset. It achieves the following results on the evaluation set:
- Loss: 0.1116
- Accuracy: 0.9823
Base Model description
This model is a distilled version of the RoBERTa-base model. It follows the same training procedure as DistilBERT. The code for the distillation process can be found here. This model is case-sensitive: it makes a difference between English and English.
The model has 6 layers, 768 dimension and 12 heads, totalizing 82M parameters (compared to 125M parameters for RoBERTa-base). On average DistilRoBERTa is twice as fast as Roberta-base.
Training Data
Polar sentiment dataset of sentences from financial news. The dataset consists of 4840 sentences from English language financial news categorised by sentiment. The dataset is divided by agreement rate of 5-8 annotators.
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- trainbatchsize: 8
- evalbatchsize: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lrschedulertype: linear
- num_epochs: 5
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| No log | 1.0 | 255 | 0.1670 | 0.9646 |
| 0.209 | 2.0 | 510 | 0.2290 | 0.9558 |
| 0.209 | 3.0 | 765 | 0.2044 | 0.9558 |
| 0.0326 | 4.0 | 1020 | 0.1116 | 0.9823 |
| 0.0326 | 5.0 | 1275 | 0.1127 | 0.9779 |
Framework versions
- Transformers 4.10.2
- Pytorch 1.9.0+cu102
- Datasets 1.12.1
- Tokenizers 0.10.3
评论