wav2vec2-large-xlsr-53-ru-snr10-commonvoice_train3000_val200
This model is a fine-tuned version of jonatasgrosman/wav2vec2-large-xlsr-53-russian on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:
- Loss: 1.3736
- Wer: 0.5075
- Cer: 0.2395
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: 0.0001
- 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: 100
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
---|---|---|---|---|---|
0.9298 | 0.96 | 180 | 1.0829 | 0.6153 | 0.2984 |
0.6185 | 1.91 | 360 | 1.1071 | 0.5924 | 0.2899 |
0.5395 | 2.87 | 540 | 1.0219 | 0.5558 | 0.2604 |
0.464 | 3.83 | 720 | 1.0042 | 0.5363 | 0.2564 |
0.4346 | 4.79 | 900 | 0.9817 | 0.5323 | 0.2403 |
0.4025 | 5.74 | 1080 | 1.0918 | 0.5558 | 0.2549 |
0.358 | 6.7 | 1260 | 1.0987 | 0.5336 | 0.2437 |
0.3466 | 7.66 | 1440 | 1.0802 | 0.5349 | 0.2437 |
0.3215 | 8.62 | 1620 | 1.1377 | 0.5467 | 0.2588 |
0.3247 | 9.57 | 1800 | 1.0324 | 0.5153 | 0.2350 |
0.287 | 10.53 | 1980 | 1.1466 | 0.5565 | 0.2603 |
0.2716 | 11.49 | 2160 | 1.2634 | 0.5532 | 0.2536 |
0.2555 | 12.45 | 2340 | 1.1859 | 0.5160 | 0.2318 |
0.2454 | 13.4 | 2520 | 1.1147 | 0.5186 | 0.2278 |
0.2299 | 14.36 | 2700 | 1.1287 | 0.5167 | 0.2282 |
0.2269 | 15.32 | 2880 | 1.2123 | 0.5042 | 0.2275 |
0.2132 | 16.28 | 3060 | 1.1219 | 0.5082 | 0.2297 |
0.1965 | 17.23 | 3240 | 1.2263 | 0.5167 | 0.2345 |
0.1943 | 18.19 | 3420 | 1.2679 | 0.5284 | 0.2353 |
0.1867 | 19.15 | 3600 | 1.2097 | 0.5186 | 0.2422 |
0.1851 | 20.11 | 3780 | 1.3118 | 0.5147 | 0.2330 |
0.1709 | 21.06 | 3960 | 1.1834 | 0.5193 | 0.2374 |
0.1757 | 22.02 | 4140 | 1.3010 | 0.5036 | 0.2272 |
0.1661 | 22.98 | 4320 | 1.2384 | 0.5075 | 0.2313 |
0.1607 | 23.94 | 4500 | 1.3642 | 0.5219 | 0.2421 |
0.1611 | 24.89 | 4680 | 1.3055 | 0.5108 | 0.2363 |
0.1567 | 25.85 | 4860 | 1.3666 | 0.5140 | 0.2383 |
0.1469 | 26.81 | 5040 | 1.3888 | 0.5101 | 0.2367 |
0.1432 | 27.77 | 5220 | 1.3478 | 0.5206 | 0.2333 |
0.1479 | 28.72 | 5400 | 1.3297 | 0.4918 | 0.2291 |
0.144 | 29.68 | 5580 | 1.3736 | 0.5075 | 0.2395 |
Framework versions
- Transformers 4.35.2
- Pytorch 2.1.0+cu118
- Datasets 2.15.0
- Tokenizers 0.15.0
- Downloads last month
- 22
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.