Edit model card

distilbert-base-uncased__hate_speech_offensive__train-32-6

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: 1.0523
  • Accuracy: 0.663

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
1.0957 1.0 19 1.0696 0.6
1.0107 2.0 38 1.0047 0.55
0.8257 3.0 57 0.8358 0.8
0.6006 4.0 76 0.7641 0.6
0.4172 5.0 95 0.5931 0.8
0.2639 6.0 114 0.5570 0.7
0.1314 7.0 133 0.5017 0.65
0.0503 8.0 152 0.3115 0.75
0.023 9.0 171 0.4353 0.85
0.0128 10.0 190 0.5461 0.75
0.0092 11.0 209 0.5045 0.8
0.007 12.0 228 0.5014 0.8
0.0064 13.0 247 0.5070 0.8
0.0049 14.0 266 0.4681 0.8
0.0044 15.0 285 0.4701 0.8
0.0039 16.0 304 0.4862 0.8
0.0036 17.0 323 0.4742 0.8
0.0035 18.0 342 0.4652 0.8

Framework versions

  • Transformers 4.15.0
  • Pytorch 1.10.2+cu102
  • Datasets 1.18.2
  • Tokenizers 0.10.3
Downloads last month
21
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.