N_roberta_agnews_padding50model

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

• Loss: 0.5524
• Accuracy: 0.9486

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
0.1998	1.0	7500	0.2132	0.9382
0.1682	2.0	15000	0.2009	0.9475
0.1506	3.0	22500	0.2273	0.9446
0.1294	4.0	30000	0.2495	0.9482
0.1028	5.0	37500	0.2612	0.9459
0.0797	6.0	45000	0.2966	0.9457
0.0646	7.0	52500	0.3040	0.9458
0.0531	8.0	60000	0.3825	0.9446
0.0443	9.0	67500	0.3838	0.9425
0.0345	10.0	75000	0.3968	0.9475
0.0395	11.0	82500	0.4132	0.9474
0.019	12.0	90000	0.4612	0.9453
0.0219	13.0	97500	0.4559	0.9458
0.0067	14.0	105000	0.4692	0.9467
0.0065	15.0	112500	0.5118	0.9461
0.0045	16.0	120000	0.5115	0.9470
0.004	17.0	127500	0.5326	0.9472
0.0079	18.0	135000	0.5088	0.9483
0.0039	19.0	142500	0.5359	0.9504
0.0024	20.0	150000	0.5524	0.9486

Framework versions

• Transformers 4.33.2
• Pytorch 2.0.1+cu117
• Datasets 2.14.5
• Tokenizers 0.13.3