haseong8012's picture
End of training
ab5a948
|
raw
history blame
4.8 kB
metadata
license: apache-2.0
base_model: facebook/wav2vec2-large-xlsr-53
tags:
  - generated_from_trainer
datasets:
  - zeroth_korean
metrics:
  - wer
model-index:
  - name: wav2vec2-large-xlsr-53-fine-tune_korean_byAILAB2
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: zeroth_korean
          type: zeroth_korean
          config: clean
          split: test
          args: clean
        metrics:
          - name: Wer
            type: wer
            value: 0.9067911459117602

wav2vec2-large-xlsr-53-fine-tune_korean_byAILAB2

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the zeroth_korean dataset. It achieves the following results on the evaluation set:

  • Loss: 1.4929
  • Wer: 0.9068

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: 0.0002
  • train_batch_size: 16
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 32
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • num_epochs: 50

Training results

Training Loss Epoch Step Validation Loss Wer
No log 0.99 38 54.4059 1.0
No log 2.0 77 38.8388 1.0
No log 2.99 115 24.1740 1.0
No log 4.0 154 16.4733 1.0
No log 4.99 192 10.1900 1.0
No log 6.0 231 6.0076 1.0
No log 6.99 269 4.8990 1.0
No log 8.0 308 4.8442 1.0
No log 8.99 346 4.8284 1.0
No log 10.0 385 4.8316 1.0
16.886 10.99 423 4.8164 1.0
16.886 12.0 462 4.7815 1.0
16.886 12.99 500 4.7204 0.9989
16.886 14.0 539 4.6842 0.9989
16.886 14.99 577 4.6641 0.9994
16.886 16.0 616 4.6527 1.0
16.886 16.99 654 4.6745 0.9992
16.886 18.0 693 4.6591 1.0
16.886 18.99 731 4.6506 0.9997
16.886 20.0 770 4.6719 0.9967
4.4391 20.99 808 4.6067 0.9968
4.4391 22.0 847 4.5748 0.9968
4.4391 22.99 885 4.5166 0.9962
4.4391 24.0 924 4.3783 0.9926
4.4391 24.99 962 4.2711 0.9913
4.4391 26.0 1001 3.6515 1.0030
4.4391 26.99 1039 3.1057 1.0640
4.4391 28.0 1078 2.6593 1.0742
4.4391 28.99 1116 2.4071 1.0587
4.4391 30.0 1155 2.2041 1.0379
4.4391 30.99 1193 2.0495 1.0319
3.1722 32.0 1232 1.9754 1.0459
3.1722 32.99 1270 1.8658 0.9968
3.1722 34.0 1309 1.7887 0.9883
3.1722 34.99 1347 1.7560 0.9776
3.1722 36.0 1386 1.6987 0.9675
3.1722 36.99 1424 1.6513 0.9443
3.1722 38.0 1463 1.6187 0.9473
3.1722 38.99 1501 1.6210 0.9408
3.1722 40.0 1540 1.5957 0.9458
3.1722 40.99 1578 1.5673 0.9246
1.2364 42.0 1617 1.5748 0.9286
1.2364 42.99 1655 1.5333 0.9217
1.2364 44.0 1694 1.5138 0.9100
1.2364 44.99 1732 1.5244 0.9223
1.2364 46.0 1771 1.5041 0.9080
1.2364 46.99 1809 1.5151 0.9155
1.2364 48.0 1848 1.4955 0.9077
1.2364 48.99 1886 1.4924 0.9065
1.2364 49.35 1900 1.4929 0.9068

Framework versions

  • Transformers 4.33.2
  • Pytorch 1.12.1
  • Datasets 2.14.5
  • Tokenizers 0.13.3