bert-base-uncased-finetuned-cola_HW2_sepehr_bakhshi_dropout_05

This model is a fine-tuned version of bert-base-uncased on the glue dataset. It achieves the following results on the evaluation set:

• Loss: 1.4260
• Matthews Correlation: 0.5780

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: 7.530341440816975e-06
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Matthews Correlation
0.5137	1.0	535	0.4936	0.4808
0.362	2.0	1070	0.4270	0.5781
0.2679	3.0	1605	0.6409	0.5148
0.2046	4.0	2140	0.5658	0.5892
0.1736	5.0	2675	0.7711	0.5624
0.1378	6.0	3210	0.8053	0.5956
0.1137	7.0	3745	0.9714	0.5523
0.0903	8.0	4280	0.9119	0.5735
0.0839	9.0	4815	1.0448	0.5839
0.0629	10.0	5350	1.2056	0.5521
0.0577	11.0	5885	1.1880	0.5889
0.0505	12.0	6420	1.1722	0.5836
0.0519	13.0	6955	1.2863	0.5884
0.0369	14.0	7490	1.2971	0.5608
0.032	15.0	8025	1.3024	0.5785
0.0244	16.0	8560	1.3904	0.5737
0.0166	17.0	9095	1.4044	0.5778
0.0185	18.0	9630	1.4234	0.5650
0.0168	19.0	10165	1.4384	0.5727
0.0224	20.0	10700	1.4260	0.5780

Framework versions

• Transformers 4.28.1
• Pytorch 2.0.0+cu118
• Datasets 2.12.0
• Tokenizers 0.13.3