superglue_rte-gpt2

This model is a fine-tuned version of gpt2 on the super_glue dataset. It achieves the following results on the evaluation set:

• Loss: 4.4821
• Accuracy: 0.5435

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: 4
• eval_batch_size: 32
• 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
0.778	1.0	623	0.6845	0.5797
0.7042	2.0	1246	0.6909	0.5797
0.7022	3.0	1869	0.6608	0.5507
0.7145	4.0	2492	0.7206	0.5797
0.6183	5.0	3115	0.8510	0.5435
0.5855	6.0	3738	1.7010	0.5362
0.5468	7.0	4361	2.3186	0.5362
0.4411	8.0	4984	2.6790	0.5435
0.3226	9.0	5607	2.6486	0.5507
0.2479	10.0	6230	3.2958	0.5362
0.1632	11.0	6853	3.3893	0.5290
0.1526	12.0	7476	3.2382	0.5942
0.1127	13.0	8099	4.0889	0.4855
0.0902	14.0	8722	3.7049	0.5580
0.0997	15.0	9345	3.6377	0.5290
0.083	16.0	9968	3.6723	0.6087
0.0612	17.0	10591	4.2905	0.5870
0.0357	18.0	11214	4.4611	0.5145
0.0643	19.0	11837	4.4033	0.5217
0.0348	20.0	12460	4.4821	0.5435

Framework versions

• Transformers 4.32.1
• Pytorch 1.13.0+cu117
• Datasets 2.15.0
• Tokenizers 0.13.3