metadata
license: apache-2.0
base_model: Twitter/twhin-bert-base
tags:
- text-classification
- generated_from_trainer
- lao-extractive-summarization
metrics:
- accuracy
- precision
- recall
- f1
model-index:
- name: fine-tuned-bert-extractive-summarization
results: []
language:
- lo
fine-tuned-bert-extractive-summarization
This model is a fine-tuned version of Twitter/twhin-bert-base on the LaoNews dataset for Lao text extractive summarization. It achieves the following results on the evaluation set:
- Loss: 0.5566
- Accuracy: 0.6995
- Precision: 0.6947
- Recall: 0.6995
- F1: 0.6961
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: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 3
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 |
|---|---|---|---|---|---|---|---|
| 0.5748 | 1.0 | 7107 | 0.5609 | 0.6916 | 0.6858 | 0.6916 | 0.6873 |
| 0.5552 | 2.0 | 14215 | 0.5659 | 0.6839 | 0.6931 | 0.6839 | 0.6870 |
| 0.5364 | 3.0 | 21321 | 0.5566 | 0.6995 | 0.6947 | 0.6995 | 0.6961 |
Framework versions
- Transformers 4.39.3
- Pytorch 2.3.0+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2