roberta-large-finetuned-ours-DS

This model is a fine-tuned version of roberta-large on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 2.3369
  • Accuracy: 0.75
  • Precision: 0.7054
  • Recall: 0.6949
  • F1: 0.6974

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: 1e-05
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 43
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 20

Training results

Training Loss Epoch Step Validation Loss Accuracy Precision Recall F1
1.0561 0.99 99 0.8773 0.615 0.4054 0.5584 0.4591
0.762 1.98 198 0.6514 0.715 0.6735 0.6672 0.6588
0.5661 2.97 297 0.6806 0.71 0.6764 0.6608 0.6435
0.3699 3.96 396 0.8358 0.71 0.6611 0.6691 0.6570
0.2184 4.95 495 1.1627 0.7 0.6597 0.6337 0.6414
0.1743 5.94 594 1.0544 0.725 0.6831 0.6949 0.6831
0.098 6.93 693 1.4757 0.73 0.6885 0.6902 0.6892
0.0813 7.92 792 1.8146 0.73 0.6840 0.6772 0.6800
0.0435 8.91 891 1.6697 0.755 0.7141 0.7127 0.7132
0.0209 9.9 990 1.8931 0.755 0.7102 0.7070 0.7082
0.0201 10.89 1089 2.1934 0.74 0.6971 0.6866 0.6907
0.0095 11.88 1188 2.1389 0.75 0.7014 0.6915 0.6932
0.0141 12.87 1287 2.1902 0.74 0.6942 0.6943 0.6936
0.0112 13.86 1386 2.5021 0.73 0.6889 0.6669 0.6741
0.0054 14.85 1485 2.3840 0.73 0.6819 0.6715 0.6746
0.0088 15.84 1584 2.3224 0.74 0.6909 0.6825 0.6787
0.003 16.83 1683 2.2641 0.75 0.7054 0.6949 0.6974
0.0017 17.82 1782 2.3361 0.75 0.7077 0.6968 0.7012
0.0014 18.81 1881 2.3041 0.755 0.7131 0.7009 0.7051
0.0083 19.8 1980 2.3369 0.75 0.7054 0.6949 0.6974

Framework versions

  • Transformers 4.20.1
  • Pytorch 1.10.1+cu111
  • Datasets 2.3.2
  • Tokenizers 0.12.1
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.