Hugging Face's logo

badrex
/
xlsr-polish

Automatic Speech Recognition
Transformers
Safetensors
wav2vec2
Generated from Trainer
Eval Results
Inference Endpoints
Model card Files Community
xlsr-polish / README.md
Badr Abdullah
Model save
5188ae8 verified
preview code
|
raw
history blame contribute delete
6.52 kB
metadata
license: apache-2.0
base_model: facebook/wav2vec2-xls-r-300m
tags:
  - generated_from_trainer
datasets:
  - common_voice_17_0
metrics:
  - wer
model-index:
  - name: xlsr-polish
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: common_voice_17_0
          type: common_voice_17_0
          config: pl
          split: validation
          args: pl
        metrics:
          - name: Wer
            type: wer
            value: 0.1443174034459139

Visualize in Weights & Biases

xlsr-polish

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1686
  • Wer: 0.1443
  • Cer: 0.0313

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: 0.0003
  • 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: 10
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Cer
4.7158 0.1543 100 4.0453 1.0 1.0
3.3469 0.3086 200 3.2544 1.0 1.0
2.9194 0.4630 300 2.7288 0.9985 0.8650
0.921 0.6173 400 0.5673 0.5449 0.1303
0.8196 0.7716 500 0.4311 0.4439 0.1025
0.7248 0.9259 600 0.3672 0.3894 0.0875
0.1727 1.0802 700 0.3141 0.3363 0.0739
0.1807 1.2346 800 0.3075 0.3463 0.0758
0.1683 1.3889 900 0.2969 0.3217 0.0707
0.1616 1.5432 1000 0.2650 0.3045 0.0675
0.1569 1.6975 1100 0.2718 0.2912 0.0658
0.1185 1.8519 1200 0.2647 0.3139 0.0672
0.1101 2.0062 1300 0.2476 0.2659 0.0576
0.1296 2.1605 1400 0.2493 0.2704 0.0590
0.0829 2.3148 1500 0.2299 0.2614 0.0576
0.0881 2.4691 1600 0.2434 0.2670 0.0601
0.125 2.6235 1700 0.2318 0.2745 0.0570
0.1227 2.7778 1800 0.2245 0.2527 0.0542
0.1128 2.9321 1900 0.2293 0.2600 0.0562
0.079 3.0864 2000 0.2227 0.2511 0.0530
0.0906 3.2407 2100 0.2289 0.2331 0.0515
0.09 3.3951 2200 0.2196 0.2486 0.0528
0.1113 3.5494 2300 0.2230 0.2392 0.0539
0.0867 3.7037 2400 0.2155 0.2237 0.0492
0.097 3.8580 2500 0.2120 0.2261 0.0493
0.0659 4.0123 2600 0.2073 0.2216 0.0493
0.0796 4.1667 2700 0.2135 0.2181 0.0468
0.0601 4.3210 2800 0.2034 0.2190 0.0480
0.0644 4.4753 2900 0.2115 0.2092 0.0456
0.0772 4.6296 3000 0.1986 0.2127 0.0461
0.066 4.7840 3100 0.1985 0.2027 0.0447
0.0633 4.9383 3200 0.2094 0.2115 0.0456
0.0579 5.0926 3300 0.2058 0.2169 0.0460
0.0709 5.2469 3400 0.1976 0.1973 0.0428
0.0405 5.4012 3500 0.2001 0.1965 0.0424
0.0515 5.5556 3600 0.2035 0.2014 0.0438
0.0785 5.7099 3700 0.1864 0.1928 0.0412
0.0514 5.8642 3800 0.1850 0.1858 0.0397
0.0355 6.0185 3900 0.1903 0.1837 0.0399
0.0621 6.1728 4000 0.1881 0.1798 0.0392
0.0525 6.3272 4100 0.1852 0.1881 0.0403
0.0497 6.4815 4200 0.1855 0.1770 0.0387
0.0362 6.6358 4300 0.1945 0.1899 0.0400
0.0399 6.7901 4400 0.1803 0.1742 0.0378
0.0483 6.9444 4500 0.1777 0.1723 0.0372
0.0293 7.0988 4600 0.1903 0.1697 0.0369
0.0635 7.2531 4700 0.1787 0.1726 0.0365
0.0199 7.4074 4800 0.1722 0.1682 0.0362
0.0393 7.5617 4900 0.1918 0.1641 0.0357
0.0357 7.7160 5000 0.1801 0.1649 0.0358
0.0444 7.8704 5100 0.1775 0.1626 0.0353
0.0266 8.0247 5200 0.1693 0.1592 0.0341
0.0381 8.1790 5300 0.1794 0.1571 0.0341
0.0308 8.3333 5400 0.1685 0.1551 0.0333
0.0304 8.4877 5500 0.1752 0.1519 0.0330
0.0316 8.6420 5600 0.1752 0.1507 0.0326
0.0377 8.7963 5700 0.1671 0.1523 0.0328
0.0588 8.9506 5800 0.1725 0.1550 0.0335
0.0487 9.1049 5900 0.1774 0.1531 0.0332
0.0169 9.2593 6000 0.1709 0.1470 0.0318
0.0274 9.4136 6100 0.1778 0.1468 0.0318
0.023 9.5679 6200 0.1718 0.1482 0.0322
0.0274 9.7222 6300 0.1700 0.1451 0.0315
0.0349 9.8765 6400 0.1686 0.1443 0.0313

Framework versions

  • Transformers 4.42.0.dev0
  • Pytorch 2.3.1+cu121
  • Datasets 2.19.2
  • Tokenizers 0.19.1