Hugging Face's logo

badrex
/
xls-r-300-cv17-bulgarian-adap-ru

Automatic Speech Recognition
Transformers
Safetensors
wav2vec2
Generated from Trainer
Eval Results
Inference Endpoints
Model card Files Community
xls-r-300-cv17-bulgarian-adap-ru / README.md
Badr Abdullah
Upload tokenizer
09fb966 verified
preview code
|
raw
history blame
3.65 kB
metadata
license: apache-2.0
tags:
  - generated_from_trainer
base_model: facebook/wav2vec2-xls-r-300m
datasets:
  - common_voice_17_0
metrics:
  - wer
model-index:
  - name: xls-r-300-cv17-bulgarian-adap-ru
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: common_voice_17_0
          type: common_voice_17_0
          config: bg
          split: validation
          args: bg
        metrics:
          - type: wer
            value: 0.3184421100534719
            name: Wer

Visualize in Weights & Biases

xls-r-300-cv17-bulgarian-adap-ru

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: 0.3848
  • Wer: 0.3184
  • Cer: 0.0766

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: 15
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Cer
3.1611 0.6579 100 3.1566 1.0 1.0
1.4834 1.3158 200 1.5156 0.9683 0.3419
0.5874 1.9737 300 0.5361 0.6018 0.1459
0.312 2.6316 400 0.3991 0.4526 0.1071
0.2139 3.2895 500 0.3913 0.4365 0.1053
0.2653 3.9474 600 0.3756 0.4114 0.0997
0.186 4.6053 700 0.3684 0.4057 0.0971
0.1569 5.2632 800 0.3831 0.4182 0.0996
0.1635 5.9211 900 0.3577 0.3803 0.0914
0.0962 6.5789 1000 0.3461 0.3620 0.0868
0.2232 7.2368 1100 0.3705 0.3596 0.0856
0.1456 7.8947 1200 0.3722 0.3643 0.0880
0.0846 8.5526 1300 0.3657 0.3565 0.0839
0.0874 9.2105 1400 0.3836 0.3418 0.0814
0.1059 9.8684 1500 0.3634 0.3397 0.0808
0.0719 10.5263 1600 0.3741 0.3468 0.0838
0.0681 11.1842 1700 0.3757 0.3396 0.0817
0.0701 11.8421 1800 0.3892 0.3324 0.0804
0.043 12.5 1900 0.3892 0.3315 0.0797
0.0482 13.1579 2000 0.3905 0.3213 0.0768
0.0279 13.8158 2100 0.3826 0.3185 0.0761
0.0609 14.4737 2200 0.3848 0.3184 0.0766

Framework versions

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