mobilebert_sa_GLUE_Experiment_logit_kd_pretrain_mrpc

This model is a fine-tuned version of gokuls/mobilebert_sa_pre-training-complete on the GLUE MRPC dataset. It achieves the following results on the evaluation set:

• Loss: 0.2291
• Accuracy: 0.8578
• F1: 0.8993
• Combined Score: 0.8786

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
0.536	1.0	29	0.4134	0.7279	0.8284	0.7782
0.3419	2.0	58	0.3005	0.8284	0.8801	0.8543
0.2413	3.0	87	0.2707	0.8235	0.8780	0.8507
0.1852	4.0	116	0.3247	0.8284	0.8837	0.8561
0.1524	5.0	145	0.2856	0.8431	0.8900	0.8666
0.1297	6.0	174	0.2999	0.8456	0.8948	0.8702
0.1219	7.0	203	0.2797	0.8529	0.8986	0.8758
0.1141	8.0	232	0.2462	0.8603	0.9005	0.8804
0.1127	9.0	261	0.2557	0.8578	0.8982	0.8780
0.1091	10.0	290	0.2853	0.8480	0.8967	0.8724
0.1007	11.0	319	0.2472	0.8554	0.8981	0.8767
0.0979	12.0	348	0.2431	0.8505	0.8950	0.8727
0.0954	13.0	377	0.2456	0.8578	0.9007	0.8793
0.0946	14.0	406	0.2526	0.8578	0.9017	0.8798
0.0946	15.0	435	0.2291	0.8578	0.8993	0.8786
0.0938	16.0	464	0.2452	0.8603	0.9029	0.8816
0.0919	17.0	493	0.2365	0.8652	0.9050	0.8851
0.0916	18.0	522	0.2363	0.8652	0.9060	0.8856
0.0915	19.0	551	0.2432	0.8652	0.9063	0.8857
0.0905	20.0	580	0.2297	0.8652	0.9057	0.8854

Framework versions

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