mobilebert_sa_GLUE_Experiment_logit_kd_qqp_256

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: 0.6619
• Accuracy: 0.7949
• F1: 0.7224
• Combined Score: 0.7586

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
0.9454	1.0	2843	0.8257	0.7556	0.6563	0.7059
0.8165	2.0	5686	0.7682	0.7590	0.6049	0.6820
0.7741	3.0	8529	0.7514	0.7638	0.6203	0.6920
0.7325	4.0	11372	0.7354	0.7675	0.6288	0.6981
0.6849	5.0	14215	0.7063	0.7785	0.6818	0.7302
0.6399	6.0	17058	0.6906	0.7828	0.6876	0.7352
0.6005	7.0	19901	0.6771	0.7868	0.6993	0.7430
0.5666	8.0	22744	0.6809	0.7897	0.7138	0.7517
0.5365	9.0	25587	0.6807	0.7886	0.6921	0.7403
0.5097	10.0	28430	0.6827	0.7873	0.7260	0.7566
0.4856	11.0	31273	0.6619	0.7949	0.7224	0.7586
0.4653	12.0	34116	0.7002	0.7948	0.7197	0.7572
0.4438	13.0	36959	0.6900	0.7965	0.7203	0.7584
0.4255	14.0	39802	0.6847	0.7981	0.7284	0.7632
0.4072	15.0	42645	0.6893	0.7917	0.7336	0.7627
0.391	16.0	45488	0.7136	0.7957	0.7300	0.7629

Framework versions

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