hbertv1-emotion-logit_KD_new

This model is a fine-tuned version of gokuls/bert_12_layer_model_v1_complete_training_new_48 on the emotion dataset. It achieves the following results on the evaluation set:

• Loss: 0.6758
• Accuracy: 0.884

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: 64
• eval_batch_size: 64
• seed: 33
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.3717	1.0	250	1.7665	0.6635
1.3315	2.0	500	1.1797	0.7825
0.9074	3.0	750	0.7837	0.864
0.636	4.0	1000	0.7591	0.861
0.5276	5.0	1250	0.7692	0.85
0.4694	6.0	1500	0.7129	0.871
0.4317	7.0	1750	0.6381	0.8765
0.3791	8.0	2000	0.6646	0.879
0.3551	9.0	2250	0.6414	0.882
0.3236	10.0	2500	0.6436	0.8795
0.301	11.0	2750	0.6758	0.884
0.2854	12.0	3000	0.6706	0.8695
0.2698	13.0	3250	0.6216	0.8825
0.2554	14.0	3500	0.6592	0.879
0.2311	15.0	3750	0.7509	0.877
0.2275	16.0	4000	0.7177	0.8775

Framework versions

• Transformers 4.35.2
• Pytorch 1.14.0a0+410ce96
• Datasets 2.15.0
• Tokenizers 0.15.0