distilbert-base-uncased__sst2__train-32-2

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: 0.4805
• Accuracy: 0.7699

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
0.7124	1.0	13	0.6882	0.5385
0.6502	2.0	26	0.6715	0.5385
0.6001	3.0	39	0.6342	0.6154
0.455	4.0	52	0.5713	0.7692
0.2605	5.0	65	0.5562	0.7692
0.1258	6.0	78	0.6799	0.7692
0.0444	7.0	91	0.8096	0.7692
0.0175	8.0	104	0.9281	0.6923
0.0106	9.0	117	0.9826	0.6923
0.0077	10.0	130	1.0254	0.7692
0.0056	11.0	143	1.0667	0.7692
0.0042	12.0	156	1.1003	0.7692
0.0036	13.0	169	1.1299	0.7692
0.0034	14.0	182	1.1623	0.6923
0.003	15.0	195	1.1938	0.6923

Framework versions

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