distilbert-base-uncased__sst2__train-16-6

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: 0.8356
• Accuracy: 0.6480

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
0.6978	1.0	7	0.6807	0.4286
0.6482	2.0	14	0.6775	0.4286
0.6051	3.0	21	0.6623	0.5714
0.486	4.0	28	0.6710	0.5714
0.4612	5.0	35	0.5325	0.7143
0.2233	6.0	42	0.4992	0.7143
0.1328	7.0	49	0.4753	0.7143
0.0905	8.0	56	0.2416	1.0
0.0413	9.0	63	0.2079	1.0
0.0356	10.0	70	0.2234	0.8571
0.0217	11.0	77	0.2639	0.8571
0.0121	12.0	84	0.2977	0.8571
0.0105	13.0	91	0.3468	0.8571
0.0085	14.0	98	0.3912	0.8571
0.0077	15.0	105	0.4000	0.8571
0.0071	16.0	112	0.4015	0.8571
0.0078	17.0	119	0.3865	0.8571
0.0059	18.0	126	0.3603	0.8571
0.0051	19.0	133	0.3231	0.8571

Framework versions

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