|
--- |
|
license: mit |
|
base_model: xlm-roberta-base |
|
tags: |
|
- generated_from_trainer |
|
metrics: |
|
- accuracy |
|
model-index: |
|
- name: xlm-yoruba-tweets-classifications |
|
results: [] |
|
datasets: |
|
- shmuhammad/AfriSenti-twitter-sentiment |
|
language: |
|
- yo |
|
pipeline_tag: text-classification |
|
widget: |
|
- test: Àti àwọn tí wọ́n ń fi wá ṣe yẹ̀yẹ́. Ẹ fiwọ́n sílẹ̀, ara ló ń ta wọ́n |
|
--- |
|
|
|
<!-- 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. --> |
|
|
|
# xlm-yoruba-tweets-classifications |
|
|
|
This model is a fine-tuned version of [xlm-roberta-base](https://huggingface.co/xlm-roberta-base) on an [shmuhammad/AfriSenti-twitter-sentiment](https://huggingface.co/datasets/shmuhammad/AfriSenti-twitter-sentiment) |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.7641 |
|
- Accuracy: 0.6871 |
|
|
|
## Model description |
|
|
|
This model is a fine-tuned version of the [xlm-roberta-base](https://huggingface.co/xlm-roberta-base) pre-trained model, specifically trained on the [shmuhammad/AfriSenti-twitter-sentiment](https://huggingface.co/datasets/shmuhammad/AfriSenti-twitter-sentiment) dataset focusing on Yoruba tweets. It aims to perform sentiment classification on Yoruba tweets. |
|
## Key details: |
|
- Type: Fine-tuned language model |
|
- Base model: xlm-roberta-base |
|
- Task: Yoruba tweet sentiment classification |
|
- Dataset: shmuhammad/AfriSenti-twitter-sentiment (Yoruba subset) |
|
|
|
|
|
## Intended uses: |
|
- Classifying sentiment (positive, negative, neutral) on Yoruba tweets. |
|
- Can be used as a starting point for further fine-tuning on specific Yoruba tweet classification tasks. |
|
|
|
## Limitations: |
|
- Trained on a limited dataset, potentially impacting performance on unseen data. |
|
- Fine-tuned only for sentiment classification, not suitable for other tasks. |
|
- Accuracy might not be optimal for all applications. |
|
|
|
|
|
## Training and evaluation data |
|
|
|
- train: Dataset({ |
|
features: ['tweet', 'label'], |
|
num_rows: 8522 |
|
}) |
|
- validation: Dataset({ |
|
features: ['tweet', 'label'], |
|
num_rows: 2090 |
|
}) |
|
|
|
|
|
### 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: 3 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Accuracy | |
|
|:-------------:|:-----:|:----:|:---------------:|:--------:| |
|
| 0.9621 | 1.0 | 1066 | 0.9099 | 0.6120 | |
|
| 0.8269 | 2.0 | 2132 | 0.7536 | 0.6627 | |
|
| 0.7239 | 3.0 | 3198 | 0.7641 | 0.6871 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.35.2 |
|
- Pytorch 2.1.0+cu121 |
|
- Datasets 2.16.1 |
|
- Tokenizers 0.15.1 |