Edit model card

bert-base-cased-qnli

This model is a fine-tuned version of bert-base-cased on the GLUE QNLI dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2569
  • Accuracy: 0.9087

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: 16
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.06
  • num_epochs: 10.0

Training results

Training Loss Epoch Step Validation Loss Accuracy
0.6529 0.08 500 0.4945 0.7734
0.4731 0.15 1000 0.3888 0.8406
0.4113 0.23 1500 0.3605 0.8523
0.4006 0.31 2000 0.3175 0.8649
0.3615 0.38 2500 0.3220 0.8678
0.3662 0.46 3000 0.3198 0.8667
0.3578 0.53 3500 0.2970 0.8779
0.3349 0.61 4000 0.2864 0.8834
0.3548 0.69 4500 0.2664 0.8946
0.3254 0.76 5000 0.2651 0.8929
0.3212 0.84 5500 0.2943 0.8817
0.3195 0.92 6000 0.2630 0.8960
0.3044 0.99 6500 0.2766 0.8882
0.245 1.07 7000 0.3268 0.9013
0.2415 1.15 7500 0.2987 0.8991
0.248 1.22 8000 0.3373 0.8812
0.2328 1.3 8500 0.3474 0.8808
0.2427 1.37 9000 0.2986 0.8913
0.243 1.45 9500 0.2835 0.8982
0.2375 1.53 10000 0.2893 0.8909
0.2253 1.6 10500 0.2942 0.8960
0.2429 1.68 11000 0.2852 0.9028
0.238 1.76 11500 0.2672 0.9037
0.2344 1.83 12000 0.2672 0.9021
0.2368 1.91 12500 0.2754 0.9019
0.2424 1.99 13000 0.2569 0.9087
0.1542 2.06 13500 0.3807 0.9026
0.1416 2.14 14000 0.3915 0.8980
0.1445 2.21 14500 0.4292 0.8984
0.1631 2.29 15000 0.4097 0.8971
0.1512 2.37 15500 0.3880 0.9012
0.1624 2.44 16000 0.4083 0.8955
0.1616 2.52 16500 0.3950 0.9039
0.1587 2.6 17000 0.3579 0.9103
0.1615 2.67 17500 0.3931 0.9012
0.1623 2.75 18000 0.3697 0.9059
0.1687 2.83 18500 0.3473 0.9037
0.1627 2.9 19000 0.3851 0.8982
0.1593 2.98 19500 0.4039 0.9019
0.1136 3.05 20000 0.4835 0.9024
0.0965 3.13 20500 0.5061 0.9012
0.0931 3.21 21000 0.5279 0.8991
0.0993 3.28 21500 0.4856 0.9019
0.1187 3.36 22000 0.4883 0.9032
0.1008 3.44 22500 0.5054 0.9015
0.1013 3.51 23000 0.5025 0.9023
0.1092 3.59 23500 0.4485 0.8986
0.1135 3.67 24000 0.5123 0.8977
0.1042 3.74 24500 0.4884 0.9010
0.1178 3.82 25000 0.4130 0.9006
0.1041 3.89 25500 0.4847 0.8951
0.1025 3.97 26000 0.4706 0.8986
0.0833 4.05 26500 0.5018 0.9054
0.0632 4.12 27000 0.5426 0.9055
0.0613 4.2 27500 0.5629 0.9021
0.0703 4.28 28000 0.5763 0.9033
0.0687 4.35 28500 0.5274 0.9039
0.0707 4.43 29000 0.5682 0.9019
0.0717 4.51 29500 0.5382 0.9004
0.0692 4.58 30000 0.5901 0.9010
0.0701 4.66 30500 0.5817 0.8990
0.0708 4.73 31000 0.5580 0.9019
0.07 4.81 31500 0.5640 0.8990
0.0725 4.89 32000 0.5768 0.8993
0.0701 4.96 32500 0.5289 0.9035
0.0441 5.04 33000 0.6401 0.9010
0.0388 5.12 33500 0.6446 0.8991
0.0417 5.19 34000 0.6327 0.9039
0.039 5.27 34500 0.6385 0.9048
0.0407 5.35 35000 0.6510 0.9030
0.0446 5.42 35500 0.5788 0.9030
0.0422 5.5 36000 0.6723 0.9024
0.04 5.58 36500 0.6602 0.9037
0.0514 5.65 37000 0.6407 0.9024
0.0462 5.73 37500 0.6145 0.9048
0.0479 5.8 38000 0.5881 0.9008
0.0503 5.88 38500 0.6001 0.9008
0.0385 5.96 39000 0.6464 0.9052
0.0436 6.03 39500 0.6683 0.9039
0.0296 6.11 40000 0.6965 0.9037
0.0275 6.19 40500 0.7193 0.9039
0.0258 6.26 41000 0.7229 0.9039
0.0223 6.34 41500 0.6974 0.9074
0.0268 6.42 42000 0.7045 0.9028
0.0297 6.49 42500 0.7289 0.9055
0.0295 6.57 43000 0.6810 0.9050
0.0265 6.64 43500 0.6833 0.9065
0.0268 6.72 44000 0.7155 0.9035
0.0293 6.8 44500 0.7632 0.9015
0.0289 6.87 45000 0.7229 0.9033
0.0286 6.95 45500 0.6671 0.9054
0.024 7.03 46000 0.7659 0.9033
0.0141 7.1 46500 0.7981 0.9023
0.0212 7.18 47000 0.7021 0.9088
0.0183 7.26 47500 0.7122 0.9096
0.0221 7.33 48000 0.7080 0.9065
0.0146 7.41 48500 0.7344 0.9074
0.0181 7.48 49000 0.7273 0.9105
0.0161 7.56 49500 0.7332 0.9083
0.0193 7.64 50000 0.7117 0.9094
0.0165 7.71 50500 0.7797 0.9070
0.0173 7.79 51000 0.7128 0.9107
0.0158 7.87 51500 0.7420 0.9110
0.0157 7.94 52000 0.7198 0.9129
0.0142 8.02 52500 0.7033 0.9114
0.0107 8.1 53000 0.7761 0.9109
0.0122 8.17 53500 0.7777 0.9110
0.013 8.25 54000 0.7880 0.9076
0.0092 8.32 54500 0.8070 0.9074
0.0102 8.4 55000 0.8113 0.9055
0.0099 8.48 55500 0.8153 0.9079
0.0087 8.55 56000 0.8045 0.9112
0.0133 8.63 56500 0.8173 0.9081
0.017 8.71 57000 0.7646 0.9101
0.0093 8.78 57500 0.7681 0.9120
0.0085 8.86 58000 0.8067 0.9083
0.0137 8.94 58500 0.7645 0.9107
0.0099 9.01 59000 0.7742 0.9123
0.0074 9.09 59500 0.8052 0.9059
0.0057 9.16 60000 0.7906 0.9132
0.0046 9.24 60500 0.7915 0.9136
0.0036 9.32 61000 0.8114 0.9101
0.0036 9.39 61500 0.8382 0.9087
0.0079 9.47 62000 0.8013 0.9110
0.0079 9.55 62500 0.8116 0.9099
0.0054 9.62 63000 0.8143 0.9110
0.0036 9.7 63500 0.8138 0.9114
0.0053 9.78 64000 0.8158 0.9107
0.0035 9.85 64500 0.8146 0.9112
0.0051 9.93 65000 0.8168 0.9107

Framework versions

  • Transformers 4.21.3
  • Pytorch 1.7.1
  • Datasets 1.18.3
  • Tokenizers 0.11.6
Downloads last month
6
Inference API
This model can be loaded on Inference API (serverless).

Dataset used to train WillHeld/bert-base-cased-qnli

Evaluation results