|
--- |
|
tags: |
|
- generated_from_trainer |
|
model-index: |
|
- name: fine-tuned-viquad-hgf |
|
results: [] |
|
--- |
|
|
|
<!-- 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. --> |
|
|
|
# fine-tuned-viquad-hgf |
|
|
|
This model is a fine-tuned version of [bhavikardeshna/xlm-roberta-base-vietnamese](https://huggingface.co/bhavikardeshna/xlm-roberta-base-vietnamese) on the [UIT-ViQuAD](https://github.com/windhashira06/Demo-QA-Extraction-system/blob/main/Dataset/UIT-ViQuAD.json) dataset. |
|
|
|
## Model description |
|
|
|
The model is described in [Cascading Adaptors to Leverage English Data to Improve Performance of |
|
Question Answering for Low-Resource Languages](https://arxiv.org/pdf/2112.09866v1.pdf) paper |
|
|
|
## Training and evaluation data |
|
|
|
A new dataset for the low-resource language as Vietnamese to evaluate MRC models. This dataset comprises over 23,000 human-generated question-answer pairs based on 5,109 passages of 174 Vietnamese articles from Wikipedia. However in processing, I eliminated more than 3000 questions with no answers. |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 2e-05 |
|
- train_batch_size: 32 |
|
- eval_batch_size: 32 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- num_epochs: 3 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Training results |
|
|
|
- **EM**: 52.38 |
|
- **F1-SCORE**: 77.67 |
|
|
|
### Framework versions |
|
|
|
- Transformers 4.26.1 |
|
- Pytorch 1.13.1+cu116 |
|
- Datasets 2.10.1 |
|
- Tokenizers 0.13.2 |
|
|
