distilbert-base-uncased__sst2__train-16-5

This model is a fine-tuned version of distilbert-base-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6537
• Accuracy: 0.6332

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
0.6925	1.0	7	0.6966	0.2857
0.6703	2.0	14	0.7045	0.2857
0.6404	3.0	21	0.7205	0.2857
0.555	4.0	28	0.7548	0.2857
0.5179	5.0	35	0.6745	0.5714
0.3038	6.0	42	0.7260	0.5714
0.2089	7.0	49	0.8016	0.5714
0.1303	8.0	56	0.8202	0.5714
0.0899	9.0	63	0.9966	0.5714
0.0552	10.0	70	1.1887	0.5714
0.0333	11.0	77	1.2163	0.5714
0.0169	12.0	84	1.2874	0.5714
0.0136	13.0	91	1.3598	0.5714
0.0103	14.0	98	1.4237	0.5714
0.0089	15.0	105	1.4758	0.5714

Framework versions

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