bertweet-base

This model is a fine-tuned version of vinai/bertweet-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6796
• F1 Macro: 0.8476
• F1: 0.8811
• F1 Neg: 0.8141
• Acc: 0.855
• Prec: 0.9267
• Recall: 0.8398
• Mcc: 0.7020

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: 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: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Macro	F1	F1 Neg	Acc	Prec	Recall	Mcc
0.6221	1.0	1161	0.5233	0.7216	0.8315	0.6116	0.765	0.7682	0.9062	0.4689
0.4332	2.0	2322	0.4843	0.7862	0.8680	0.7045	0.8175	0.8081	0.9375	0.5946
0.3714	3.0	3483	0.5872	0.8405	0.8963	0.7846	0.86	0.8521	0.9453	0.6914
0.2856	4.0	4644	0.5511	0.8589	0.8984	0.8194	0.87	0.8984	0.8984	0.7179
0.2199	5.0	5805	0.6796	0.8476	0.8811	0.8141	0.855	0.9267	0.8398	0.7020

Framework versions

• Transformers 4.40.1
• Pytorch 2.3.0+cu121
• Datasets 2.18.0
• Tokenizers 0.19.1