distilbert-base-uncased__hate_speech_offensive__train-16-7

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: 0.9011
• Accuracy: 0.578

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
1.0968	1.0	10	1.1309	0.0
1.0709	2.0	20	1.1237	0.1
0.9929	3.0	30	1.1254	0.1
0.878	4.0	40	1.1206	0.5
0.7409	5.0	50	1.0831	0.1
0.5663	6.0	60	0.9830	0.6
0.4105	7.0	70	0.9919	0.5
0.2912	8.0	80	1.0472	0.6
0.1013	9.0	90	1.1617	0.4
0.0611	10.0	100	1.2789	0.6
0.039	11.0	110	1.4091	0.4
0.0272	12.0	120	1.4974	0.4
0.0189	13.0	130	1.4845	0.5
0.018	14.0	140	1.4924	0.5
0.0131	15.0	150	1.5206	0.6
0.0116	16.0	160	1.5858	0.5

Framework versions

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