artyomboyko's picture
Update README.md
973bf11
|
raw
history blame
8.35 kB
metadata
language:
  - ru
license: apache-2.0
tags:
  - hf-asr-leaderboard
  - generated_from_trainer
datasets:
  - mozilla-foundation/common_voice_13_0
metrics:
  - wer
model-index:
  - name: whisper-base-fine_tuned-ru
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: common_voice_13_0
          type: mozilla-foundation/common_voice_11_0
          args: 'config: ru, split: test'
        metrics:
          - name: Wer
            type: wer
            value: 41.216909250757055

whisper-base-fine_tuned-ru

This model is a fine-tuned version of openai/whisper-small on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.22031
  • Wer: 17.724332

Model description

Same as original model (see whisper-small). But! This model has been fine-tuned for the task of transcribing the Russian language.

Intended uses & limitations

Same as original model (see whisper-small).

Training and evaluation data

More information needed

Training procedure

The model is fine-tuned using the following notebook (available only in the Russian version): https://github.com/blademoon/Whisper_Train

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 1e-06
  • 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_steps: 250
  • training_steps: 50000

Training results

Training Loss Epoch Step Validation Loss Wer
0.344 0.22 500 0.3936 58.4474
0.1948 0.44 1000 0.2391 57.0232
0.1853 0.66 1500 0.2255 66.1826
0.186 0.88 2000 0.2180 65.3833
0.1532 1.1 2500 0.2135 50.6050
0.1374 1.32 3000 0.2107 47.9428
0.1359 1.54 3500 0.2082 60.0693
0.1387 1.76 4000 0.2052 58.8674
0.1212 1.97 4500 0.2027 51.9571
0.111 2.19 5000 0.2027 50.0780
0.1108 2.41 5500 0.2013 42.9664
0.1148 2.63 6000 0.2000 40.7882
0.114 2.85 6500 0.2002 32.6050
0.092 3.07 7000 0.2000 32.9307
0.0783 3.29 7500 0.2001 33.1413
0.0989 3.51 8000 0.1986 32.0313
0.0919 3.73 8500 0.1991 28.7199
0.0928 3.95 9000 0.1982 26.1798
0.0721 4.17 9500 0.2007 22.4960
0.078 4.39 10000 0.2012 26.0774
0.0764 4.61 10500 0.2004 24.7906
0.0812 4.83 11000 0.2003 24.8022
0.0531 5.05 11500 0.2022 21.3837
0.0587 5.27 12000 0.2038 21.1638
0.0553 5.48 12500 0.2039 21.9224
0.0537 5.7 13000 0.2042 20.9671
0.0608 5.92 13500 0.2049 21.1068
0.0467 6.14 14000 0.2073 18.6528
0.0533 6.36 14500 0.2088 18.7843
0.048 6.58 15000 0.2092 18.5609
0.0479 6.8 15500 0.2101 19.1648
0.0383 7.02 16000 0.2105 18.9379
0.0384 7.24 16500 0.2147 18.8018
0.0451 7.46 17000 0.2156 18.9170
0.0399 7.68 17500 0.2163 18.3806
0.0387 7.9 18000 0.2159 17.9605
0.0347 8.12 18500 0.2203 17.7243
0.0324 8.34 19000 0.2231 17.8163
0.035 8.56 19500 0.2231 17.8954
0.0338 8.78 20000 0.2234 17.7371
0.0305 9.0 20500 0.2244 17.8035
0.0244 9.21 21000 0.2305 17.8942
0.0249 9.43 21500 0.2321 17.9024
0.0242 9.65 22000 0.2328 18.2212
0.0269 9.87 22500 0.2327 17.8104
0.0198 10.09 23000 0.2380 17.7301
0.0191 10.31 23500 0.2396 17.8861
0.0218 10.53 24000 0.2412 17.7464
0.0219 10.75 24500 0.2406 17.7453
0.0206 10.97 25000 0.2427 17.9128
0.0182 11.19 25500 0.2482 18.0676
0.0143 11.41 26000 0.2506 17.9245
0.0162 11.63 26500 0.2501 18.1572
0.0172 11.85 27000 0.2535 18.1164
0.0148 12.07 27500 0.2558 18.1130
0.0123 12.29 28000 0.2573 18.4085
0.0129 12.51 28500 0.2603 18.0978
0.0136 12.72 29000 0.2615 18.1793
0.011 12.94 29500 0.2617 18.2247
0.0096 13.16 30000 0.2666 18.2712
0.01 13.38 30500 0.2667 18.4457
0.0122 13.6 31000 0.2690 18.1095
0.0121 13.82 31500 0.2700 18.1653
0.0088 14.04 32000 0.2720 18.4539
0.0076 14.26 32500 0.2746 18.2956
0.0086 14.48 33000 0.2764 18.5644
0.0086 14.7 33500 0.2771 18.5260
0.0085 14.92 34000 0.2788 18.4481
0.008 15.14 34500 0.2803 18.4923
0.0074 15.36 35000 0.2824 18.6028
0.0069 15.58 35500 0.2838 18.7692
0.008 15.8 36000 0.2848 18.6901
0.0065 16.02 36500 0.2864 18.7413
0.006 16.24 37000 0.2885 18.5458
0.0061 16.45 37500 0.2885 18.6470
0.0056 16.67 38000 0.2898 18.3736
0.0061 16.89 38500 0.2912 18.8064
0.0048 17.11 39000 0.2933 18.9018
0.0053 17.33 39500 0.2939 18.6168
0.006 17.55 40000 0.2954 18.7238
0.0045 17.77 40500 0.2952 18.8099
0.0059 17.99 41000 0.2964 18.5551
0.0053 18.21 41500 0.2980 18.7157
0.004 18.43 42000 0.2988 18.6412
0.0049 18.65 42500 0.2990 18.7099
0.0048 18.87 43000 0.3004 18.7552
0.0041 19.09 43500 0.3015 18.8169
0.0048 19.31 44000 0.3018 18.8518
0.0039 19.53 44500 0.3022 18.9437
0.0041 19.75 45000 0.3029 18.8239
0.0041 19.96 45500 0.3036 18.8169
0.004 20.18 46000 0.3045 18.8274
0.0044 20.4 46500 0.3048 18.8867
0.0042 20.62 47000 0.3054 18.8425
0.0044 20.84 47500 0.3058 18.8448
0.004 21.06 48000 0.3057 18.8425
0.0038 21.28 48500 0.3062 18.7029
0.0038 21.5 49000 0.3063 18.8413
0.0046 21.72 49500 0.3063 18.8227
0.0036 21.94 50000 0.3064 18.8483

Framework versions

  • Transformers 4.30.2
  • Pytorch 2.0.1+cu117
  • Datasets 2.13.1
  • Tokenizers 0.13.3