deberta-v3-large__sst2__train-16-4

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: 0.6329
• Accuracy: 0.6392

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
0.6945	1.0	7	0.7381	0.2857
0.7072	2.0	14	0.7465	0.2857
0.6548	3.0	21	0.7277	0.4286
0.5695	4.0	28	0.6738	0.5714
0.4615	5.0	35	0.8559	0.5714
0.0823	6.0	42	1.0983	0.5714
0.0274	7.0	49	1.9937	0.5714
0.0106	8.0	56	2.2209	0.5714
0.0039	9.0	63	2.2114	0.5714
0.0031	10.0	70	2.2808	0.5714
0.0013	11.0	77	2.3707	0.5714
0.0008	12.0	84	2.4902	0.5714
0.0005	13.0	91	2.5208	0.5714
0.0007	14.0	98	2.5683	0.5714

Framework versions

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