mHuBERT-147-br / README.md
gweltou's picture
Upload tokenizer
6799feb verified
|
raw
history blame
3.1 kB
metadata
license: cc-by-nc-4.0
tags:
  - generated_from_trainer
base_model: utter-project/mHuBERT-147
datasets:
  - common_voice_15_0
metrics:
  - wer
model-index:
  - name: mHuBERT-147-br
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: common_voice_15_0
          type: common_voice_15_0
          config: br
          split: None
          args: br
        metrics:
          - type: wer
            value: 53.76572908956329
            name: Wer

mHuBERT-147-br

This model is a fine-tuned version of utter-project/mHuBERT-147 on the common_voice_15_0 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.7650
  • Wer: 53.7657
  • Cer: 18.3841

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: 3.7e-05
  • train_batch_size: 8
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 16
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • num_epochs: 40
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Cer Validation Loss Wer
6.5746 2.18 1000 99.8848 3.8929 100.0
2.8591 4.36 2000 51.1549 1.8873 97.5296
1.4189 6.54 3000 27.4120 1.0985 77.2853
0.9787 8.71 4000 0.8995 71.3360 24.4590
0.803 10.89 5000 0.8429 67.1817 22.9902
0.718 13.07 6000 0.8035 63.8879 21.6750
0.6359 15.25 7000 0.7927 62.2502 21.1144
0.5832 17.43 8000 0.7508 60.3072 20.3406
0.555 19.61 9000 0.7509 58.7990 19.8568
0.5167 21.79 10000 0.7757 58.0218 19.7569
0.4917 23.97 11000 0.7588 56.9671 19.4574
0.4629 26.14 12000 0.7710 55.6255 19.0792
0.4454 28.32 13000 0.7546 55.0888 18.8257
0.4235 30.5 14000 0.7548 54.9963 18.7240
0.4135 32.68 15000 0.7689 54.6725 18.6222
0.411 34.86 16000 0.7619 54.4504 18.5320
0.3934 37.04 17000 0.7621 53.9323 18.4014
0.3912 39.22 18000 0.7650 53.7657 18.3841

Framework versions

  • Transformers 4.39.1
  • Pytorch 2.0.1+cu117
  • Datasets 2.18.0
  • Tokenizers 0.15.2