deberta-v3-large__sst2__train-16-0

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: 0.9917
• Accuracy: 0.7705

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
0.7001	1.0	7	0.7327	0.2857
0.6326	2.0	14	0.6479	0.5714
0.5232	3.0	21	0.5714	0.5714
0.3313	4.0	28	0.6340	0.7143
0.3161	5.0	35	0.6304	0.7143
0.0943	6.0	42	0.4719	0.8571
0.0593	7.0	49	0.5000	0.7143
0.0402	8.0	56	0.3530	0.8571
0.0307	9.0	63	0.3499	0.8571
0.0033	10.0	70	0.3258	0.8571
0.0021	11.0	77	0.3362	0.8571
0.0012	12.0	84	0.4591	0.8571
0.0036	13.0	91	0.4661	0.8571
0.001	14.0	98	0.5084	0.8571
0.0017	15.0	105	0.5844	0.8571
0.0005	16.0	112	0.6645	0.8571
0.002	17.0	119	0.7422	0.8571
0.0006	18.0	126	0.7354	0.8571
0.0005	19.0	133	0.7265	0.8571
0.0005	20.0	140	0.7207	0.8571

Framework versions

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