|
--- |
|
license: apache-2.0 |
|
tags: |
|
- generated_from_trainer |
|
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 |
|
--- |
|
|
|
<!-- 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. --> |
|
|
|
# distilRoberta-financial-sentiment |
|
|
|
This model is a fine-tuned version of [distilroberta-base](https://huggingface.co/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 |
|
|
