deberta-v3-large__sst2__train-8-9

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: 0.6013
• Accuracy: 0.7210

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
0.6757	1.0	3	0.7810	0.25
0.6506	2.0	6	0.8102	0.25
0.6463	3.0	9	0.8313	0.25
0.5813	4.0	12	0.8858	0.25
0.4635	5.0	15	0.8220	0.25
0.3992	6.0	18	0.7226	0.5
0.3281	7.0	21	0.6707	0.75
0.2276	8.0	24	0.7515	0.75
0.1674	9.0	27	0.6971	0.75
0.0873	10.0	30	0.5419	0.75
0.0525	11.0	33	0.5025	0.75
0.0286	12.0	36	0.5229	0.75
0.0149	13.0	39	0.5660	0.75
0.0082	14.0	42	0.6954	0.75
0.006	15.0	45	0.8649	0.75
0.0043	16.0	48	1.0011	0.75
0.0035	17.0	51	1.0909	0.75
0.0021	18.0	54	1.1615	0.75
0.0017	19.0	57	1.2147	0.75
0.0013	20.0	60	1.2585	0.75
0.0016	21.0	63	1.2917	0.75

Framework versions

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