File size: 8,384 Bytes
b459b53 973bf11 de13070 973bf11 8e12f3a b459b53 973bf11 cf1c209 973bf11 800da3c 973bf11 800da3c 973bf11 2942510 973bf11 c26be60 973bf11 79637bc 973bf11 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 |
---
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-small-fine_tuned-ru
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: common_voice_13_0
type: mozilla-foundation/common_voice_13_0
args: 'config: ru, split: test'
metrics:
- name: Wer
type: wer
value: 17.724332
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# whisper-small-fine_tuned-ru
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the [common_voice_13_0](https://huggingface.co/datasets/mozilla-foundation/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](https://huggingface.co/openai/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](https://huggingface.co/openai/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: Pytorch 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
|