metadata
license: apache-2.0
tags:
- generated_from_trainer
- financial
- stocks
- sentiment
widget:
- text: Operating profit totaled EUR 9.4 mn , down from EUR 11.7 mn in 2004 .
datasets:
- financial_phrasebank
metrics:
- accuracy
model-index:
- name: distilRoberta-financial-sentiment
results:
- task:
name: Text Classification
type: text-classification
dataset:
name: financial_phrasebank
type: financial_phrasebank
args: sentences_allagree
metrics:
- name: Accuracy
type: accuracy
value: 0.9823008849557522
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
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: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: 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