deberta-v3-large__sst2__train-16-5

This model is a fine-tuned version of microsoft/deberta-v3-large on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5433
• Accuracy: 0.7924

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: 4
• eval_batch_size: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.6774	1.0	7	0.7450	0.2857
0.7017	2.0	14	0.7552	0.2857
0.6438	3.0	21	0.7140	0.4286
0.3525	4.0	28	0.5570	0.7143
0.2061	5.0	35	0.5303	0.8571
0.0205	6.0	42	0.6706	0.8571
0.0068	7.0	49	0.8284	0.8571
0.0029	8.0	56	0.9281	0.8571
0.0015	9.0	63	0.9871	0.8571
0.0013	10.0	70	1.0208	0.8571
0.0008	11.0	77	1.0329	0.8571
0.0005	12.0	84	1.0348	0.8571
0.0004	13.0	91	1.0437	0.8571
0.0005	14.0	98	1.0512	0.8571
0.0004	15.0	105	1.0639	0.8571

Framework versions

• Transformers 4.15.0
• Pytorch 1.10.2+cu102
• Datasets 1.18.2
• Tokenizers 0.10.3