distilbert-base-uncased__sst2__train-32-4

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: 0.5001
• Accuracy: 0.7650

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
0.7175	1.0	13	0.6822	0.5385
0.6559	2.0	26	0.6533	0.6154
0.6052	3.0	39	0.5762	0.7692
0.4587	4.0	52	0.4477	0.8462
0.2459	5.0	65	0.4288	0.7692
0.1001	6.0	78	0.5219	0.7692
0.0308	7.0	91	0.8540	0.7692
0.014	8.0	104	0.7789	0.7692
0.0083	9.0	117	0.7996	0.7692
0.0064	10.0	130	0.8342	0.7692
0.0049	11.0	143	0.8612	0.7692
0.0036	12.0	156	0.8834	0.7692
0.0032	13.0	169	0.9067	0.7692
0.003	14.0	182	0.9332	0.7692
0.0028	15.0	195	0.9511	0.7692

Framework versions

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