distilbert-base-uncased-finetuned-mrpc

This model is a fine-tuned version of distilbert-base-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5682
• Accuracy: 0.7164
• F1: 0.2022

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: 2e-05
• 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	Accuracy	F1
0.5789	1.0	635	0.5764	0.7055	0.0
0.543	2.0	1270	0.5682	0.7164	0.2022
0.4661	3.0	1905	0.6358	0.7164	0.2812
0.2338	4.0	2540	0.9347	0.6844	0.4670
0.1538	5.0	3175	1.3556	0.6758	0.4740
0.1067	6.0	3810	1.6163	0.7016	0.3322
0.0928	7.0	4445	2.0786	0.6984	0.3609
0.0438	8.0	5080	2.1976	0.6945	0.4309
0.0312	9.0	5715	2.1931	0.6969	0.4209
0.0311	10.0	6350	2.4030	0.6883	0.4158
0.0281	11.0	6985	2.3715	0.7148	0.3739
0.0166	12.0	7620	2.6843	0.6984	0.3390
0.0167	13.0	8255	2.7291	0.6922	0.3604
0.0181	14.0	8890	2.7929	0.6906	0.3851
0.0147	15.0	9525	2.8976	0.7117	0.3303
0.0103	16.0	10160	3.0229	0.6859	0.3964
0.0047	17.0	10795	3.0616	0.6836	0.3817
0.0136	18.0	11430	3.0513	0.6875	0.3730
0.005	19.0	12065	3.0634	0.6930	0.3732
0.0042	20.0	12700	3.0611	0.7	0.3642

Framework versions

• Transformers 4.28.1
• Pytorch 2.0.1+cu117
• Datasets 1.18.4
• Tokenizers 0.12.1