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