<!-- 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. -->
ABSA-MaskedLM-Vietnamese
This model is a fine-tuned version of vinai/phobert-base on the vietnamese_students_feedback dataset. It achieves the following results on the evaluation set:
- Loss: 0.0351
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: 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: 3.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.077 | 1.0 | 220 | 0.0398 |
| 0.0415 | 2.0 | 440 | 0.0370 |
| 0.0381 | 3.0 | 660 | 0.0351 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2
