mobilebert_sa_GLUE_Experiment_logit_kd_pretrain_stsb

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

• Loss: 0.2919
• Pearson: 0.8665
• Spearmanr: 0.8642
• Combined Score: 0.8654

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	Pearson	Spearmanr	Combined Score
1.1501	1.0	45	0.4726	0.7774	0.7922	0.7848
0.364	2.0	90	0.3480	0.8457	0.8455	0.8456
0.259	3.0	135	0.3156	0.8582	0.8590	0.8586
0.2054	4.0	180	0.4231	0.8551	0.8549	0.8550
0.1629	5.0	225	0.3245	0.8668	0.8654	0.8661
0.1263	6.0	270	0.3192	0.8649	0.8625	0.8637
0.1021	7.0	315	0.3337	0.8655	0.8629	0.8642
0.0841	8.0	360	0.3061	0.8601	0.8577	0.8589
0.0713	9.0	405	0.3600	0.8576	0.8555	0.8566
0.0587	10.0	450	0.3135	0.8620	0.8600	0.8610
0.0488	11.0	495	0.3006	0.8641	0.8620	0.8631
0.0441	12.0	540	0.3308	0.8645	0.8621	0.8633
0.0385	13.0	585	0.3468	0.8620	0.8601	0.8610
0.0346	14.0	630	0.3175	0.8658	0.8634	0.8646
0.0298	15.0	675	0.2919	0.8665	0.8642	0.8654
0.0299	16.0	720	0.3103	0.8649	0.8628	0.8639
0.0263	17.0	765	0.3325	0.8620	0.8599	0.8609
0.0237	18.0	810	0.3092	0.8636	0.8611	0.8623
0.0213	19.0	855	0.3169	0.8653	0.8631	0.8642
0.0196	20.0	900	0.2985	0.8647	0.8624	0.8636

Framework versions

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