metadata

license: apache-2.0
thumbnail: >-
  https://huggingface.co/mrm8488/distilroberta-finetuned-financial-news-sentiment-analysis/resolve/main/logo_fin.png
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