Edit model card

result_xlmr_siqa

This model is a fine-tuned version of xlm-roberta-large on the super_glue dataset. It trained first on SIQA dataset. It achieves the following results on the evaluation set:

  • Loss: 1.4143
  • Accuracy: 0.79

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: 8
  • eval_batch_size: 8
  • seed: 44
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 10

Training results

Training Loss Epoch Step Validation Loss Accuracy
0.0152 0.2 10 1.0207 0.77
0.001 0.4 20 0.7651 0.82
0.0013 0.6 30 0.7756 0.79
0.0012 0.8 40 1.2054 0.8
0.0005 1.0 50 1.3034 0.79
0.0008 1.2 60 1.1920 0.76
0.0138 1.4 70 0.9139 0.76
0.0003 1.6 80 0.9160 0.78
0.0001 1.8 90 1.1525 0.8
0.0085 2.0 100 0.8657 0.79
0.0033 2.2 110 0.8925 0.79
0.0055 2.4 120 1.2264 0.78
0.0014 2.6 130 1.4958 0.8
0.0031 2.8 140 1.4250 0.79
0.0138 3.0 150 1.4240 0.81
0.0304 3.2 160 1.4179 0.8
0.0 3.4 170 1.4685 0.8
0.0 3.6 180 1.4897 0.8
0.0015 3.8 190 1.2689 0.8
0.0001 4.0 200 1.0355 0.78
0.0007 4.2 210 1.1339 0.77
0.0002 4.4 220 1.1915 0.79
0.0001 4.6 230 1.1300 0.8
0.001 4.8 240 1.1464 0.79
0.0001 5.0 250 1.2227 0.78
0.0 5.2 260 1.3048 0.81
0.0 5.4 270 1.3418 0.79
0.0093 5.6 280 1.3442 0.78
0.0004 5.8 290 1.2721 0.8
0.0035 6.0 300 1.1852 0.77
0.0016 6.2 310 1.1745 0.77
0.0003 6.4 320 1.1138 0.8
0.0002 6.6 330 1.2342 0.79
0.0055 6.8 340 1.3594 0.79
0.0 7.0 350 1.4109 0.79
0.0 7.2 360 1.4677 0.78
0.0 7.4 370 1.4951 0.77
0.0 7.6 380 1.4987 0.77
0.0004 7.8 390 1.4517 0.77
0.0 8.0 400 1.4632 0.77
0.0 8.2 410 1.4825 0.78
0.0008 8.4 420 1.4486 0.79
0.0 8.6 430 1.4426 0.79
0.0 8.8 440 1.4216 0.79
0.0 9.0 450 1.4166 0.79
0.0 9.2 460 1.4161 0.79
0.0 9.4 470 1.4172 0.79
0.0003 9.6 480 1.4179 0.79
0.0286 9.8 490 1.4155 0.79
0.0 10.0 500 1.4143 0.79

Framework versions

  • Transformers 4.34.0
  • Pytorch 2.1.0
  • Datasets 2.14.5
  • Tokenizers 0.14.0
Downloads last month
37
Inference API (serverless) does not yet support transformers models for this pipeline type.

Finetuned from

Dataset used to train Erland/result_xlmr_siqa