distilbert-base-uncased__sst2__train-32-3

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: 0.5694
• Accuracy: 0.7073

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
0.7118	1.0	13	0.6844	0.5385
0.6587	2.0	26	0.6707	0.6154
0.6067	3.0	39	0.6295	0.5385
0.4714	4.0	52	0.5811	0.6923
0.2444	5.0	65	0.5932	0.7692
0.1007	6.0	78	0.7386	0.6923
0.0332	7.0	91	0.6962	0.6154
0.0147	8.0	104	0.8200	0.7692
0.0083	9.0	117	0.9250	0.7692
0.0066	10.0	130	0.9345	0.7692
0.005	11.0	143	0.9313	0.7692
0.0036	12.0	156	0.9356	0.7692
0.0031	13.0	169	0.9395	0.7692
0.0029	14.0	182	0.9504	0.7692

Framework versions

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