emrecan's picture
Update README.md
c71182e
metadata
language:
  - tr
tags:
  - zero-shot-classification
  - nli
  - pytorch
pipeline_tag: zero-shot-classification
license: mit
datasets:
  - nli_tr
metrics:
  - accuracy
widget:
  - text: Dolar yükselmeye devam ediyor.
    candidate_labels: ekonomi, siyaset, spor
  - text: Senaryo çok saçmaydı, beğendim diyemem.
    candidate_labels: olumlu, olumsuz

bert-base-turkish-cased_allnli_tr

This model is a fine-tuned version of dbmdz/bert-base-turkish-cased on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.5771
  • Accuracy: 0.7978

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: 32
  • eval_batch_size: 32
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 3

Training results

Training Loss Epoch Step Validation Loss Accuracy
0.8559 0.03 1000 0.7577 0.6798
0.6612 0.07 2000 0.7263 0.6958
0.6115 0.1 3000 0.6431 0.7364
0.5916 0.14 4000 0.6347 0.7407
0.5719 0.17 5000 0.6317 0.7483
0.5575 0.2 6000 0.6034 0.7544
0.5521 0.24 7000 0.6148 0.7568
0.5393 0.27 8000 0.5931 0.7610
0.5382 0.31 9000 0.5866 0.7665
0.5306 0.34 10000 0.5881 0.7594
0.5295 0.37 11000 0.6120 0.7632
0.5225 0.41 12000 0.5620 0.7759
0.5112 0.44 13000 0.5641 0.7769
0.5133 0.48 14000 0.5571 0.7798
0.5023 0.51 15000 0.5719 0.7722
0.5017 0.54 16000 0.5482 0.7844
0.5111 0.58 17000 0.5503 0.7800
0.4929 0.61 18000 0.5502 0.7836
0.4923 0.65 19000 0.5424 0.7843
0.4894 0.68 20000 0.5417 0.7851
0.4877 0.71 21000 0.5514 0.7841
0.4818 0.75 22000 0.5494 0.7848
0.4898 0.78 23000 0.5450 0.7859
0.4823 0.82 24000 0.5417 0.7878
0.4806 0.85 25000 0.5354 0.7875
0.4779 0.88 26000 0.5338 0.7848
0.4744 0.92 27000 0.5277 0.7934
0.4678 0.95 28000 0.5507 0.7871
0.4727 0.99 29000 0.5603 0.7789
0.4243 1.02 30000 0.5626 0.7894
0.3955 1.05 31000 0.5324 0.7939
0.4022 1.09 32000 0.5322 0.7925
0.3976 1.12 33000 0.5450 0.7920
0.3913 1.15 34000 0.5464 0.7948
0.406 1.19 35000 0.5406 0.7958
0.3875 1.22 36000 0.5489 0.7878
0.4024 1.26 37000 0.5427 0.7925
0.3988 1.29 38000 0.5335 0.7904
0.393 1.32 39000 0.5415 0.7923
0.3988 1.36 40000 0.5385 0.7962
0.3912 1.39 41000 0.5383 0.7950
0.3949 1.43 42000 0.5415 0.7931
0.3902 1.46 43000 0.5438 0.7893
0.3948 1.49 44000 0.5348 0.7906
0.3921 1.53 45000 0.5361 0.7890
0.3944 1.56 46000 0.5419 0.7953
0.3959 1.6 47000 0.5402 0.7967
0.3926 1.63 48000 0.5429 0.7925
0.3854 1.66 49000 0.5346 0.7959
0.3864 1.7 50000 0.5241 0.7979
0.385 1.73 51000 0.5149 0.8002
0.3871 1.77 52000 0.5325 0.8002
0.3819 1.8 53000 0.5332 0.8022
0.384 1.83 54000 0.5419 0.7873
0.3899 1.87 55000 0.5225 0.7974
0.3894 1.9 56000 0.5358 0.7977
0.3838 1.94 57000 0.5264 0.7988
0.3881 1.97 58000 0.5280 0.7956
0.3756 2.0 59000 0.5601 0.7969
0.3156 2.04 60000 0.5936 0.7925
0.3125 2.07 61000 0.5898 0.7938
0.3179 2.11 62000 0.5591 0.7981
0.315 2.14 63000 0.5853 0.7970
0.3122 2.17 64000 0.5802 0.7979
0.3105 2.21 65000 0.5758 0.7979
0.3076 2.24 66000 0.5685 0.7980
0.3117 2.28 67000 0.5799 0.7944
0.3108 2.31 68000 0.5742 0.7988
0.3047 2.34 69000 0.5907 0.7921
0.3114 2.38 70000 0.5723 0.7937
0.3035 2.41 71000 0.5944 0.7955
0.3129 2.45 72000 0.5838 0.7928
0.3071 2.48 73000 0.5929 0.7949
0.3061 2.51 74000 0.5794 0.7967
0.3068 2.55 75000 0.5892 0.7954
0.3053 2.58 76000 0.5796 0.7962
0.3117 2.62 77000 0.5763 0.7981
0.3062 2.65 78000 0.5852 0.7964
0.3004 2.68 79000 0.5793 0.7966
0.3146 2.72 80000 0.5693 0.7985
0.3146 2.75 81000 0.5788 0.7982
0.3079 2.79 82000 0.5726 0.7978
0.3058 2.82 83000 0.5677 0.7988
0.3055 2.85 84000 0.5701 0.7982
0.3049 2.89 85000 0.5809 0.7970
0.3044 2.92 86000 0.5741 0.7986
0.3057 2.96 87000 0.5743 0.7980
0.3081 2.99 88000 0.5771 0.7978

Framework versions

  • Transformers 4.12.3
  • Pytorch 1.10.0+cu102
  • Datasets 1.15.1
  • Tokenizers 0.10.3