BERT-tiny-emotion-intent

This model is a fine-tuned version of google/bert_uncased_L-2_H-128_A-2 on the emotion dataset. It achieves the following results on the evaluation set:

• Loss: 0.3620
• Accuracy: 0.91

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 33
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.2603	1.0	1000	0.7766	0.7815
0.5919	2.0	2000	0.4117	0.884
0.367	3.0	3000	0.3188	0.8995
0.2848	4.0	4000	0.2928	0.8985
0.2395	5.0	5000	0.2906	0.898
0.2094	6.0	6000	0.2887	0.907
0.1884	7.0	7000	0.2831	0.9065
0.1603	8.0	8000	0.3044	0.9065
0.1519	9.0	9000	0.3124	0.9095
0.1291	10.0	10000	0.3256	0.9065
0.1179	11.0	11000	0.3651	0.9035
0.1091	12.0	12000	0.3620	0.91
0.0977	13.0	13000	0.3992	0.907
0.0914	14.0	14000	0.4285	0.908
0.0876	15.0	15000	0.4268	0.9055

Framework versions

• Transformers 4.22.1
• Pytorch 1.12.1+cu113
• Datasets 2.5.1
• Tokenizers 0.12.1