mobilebert_sa_GLUE_Experiment_logit_kd_qqp_128

This model is a fine-tuned version of google/mobilebert-uncased on the GLUE QQP dataset. It achieves the following results on the evaluation set:

• Loss: 0.6884
• Accuracy: 0.7872
• F1: 0.7062
• Combined Score: 0.7467

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: 128
• eval_batch_size: 128
• seed: 10
• 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	F1	Combined Score
0.9518	1.0	2843	0.8352	0.7536	0.6530	0.7033
0.8249	2.0	5686	0.7766	0.7607	0.6219	0.6913
0.7847	3.0	8529	0.7625	0.7648	0.6402	0.7025
0.7498	4.0	11372	0.7551	0.7638	0.6197	0.6917
0.7137	5.0	14215	0.7387	0.7691	0.6545	0.7118
0.6762	6.0	17058	0.7165	0.7753	0.6720	0.7237
0.6373	7.0	19901	0.7042	0.7783	0.6765	0.7274
0.6045	8.0	22744	0.7075	0.7799	0.6902	0.7350
0.5729	9.0	25587	0.7233	0.7792	0.6639	0.7215
0.545	10.0	28430	0.7088	0.7805	0.7180	0.7493
0.5183	11.0	31273	0.6884	0.7872	0.7062	0.7467
0.4948	12.0	34116	0.7064	0.7869	0.7076	0.7472
0.4724	13.0	36959	0.7053	0.7884	0.7120	0.7502
0.4514	14.0	39802	0.7314	0.7903	0.7024	0.7464
0.4321	15.0	42645	0.7112	0.7891	0.7228	0.7560
0.4152	16.0	45488	0.7410	0.7909	0.7211	0.7560

Framework versions

• Transformers 4.26.0
• Pytorch 1.14.0a0+410ce96
• Datasets 2.9.0
• Tokenizers 0.13.2