kingabzpro's picture
Update README.md
6404ab3
metadata
language:
  - sv-SE
license: apache-2.0
tags:
  - automatic-speech-recognition
  - robust-speech-event
datasets:
  - mozilla-foundation/common_voice_8_0
metrics:
  - wer
  - cer
model-index:
  - name: wav2vec2-xls-r-300m-swedish
    results:
      - task:
          type: automatic-speech-recognition
          name: Speech Recognition
        dataset:
          type: mozilla-foundation/common_voice_8_0
          name: Common Voice sv-SE
          args: sv-SE
        metrics:
          - type: wer
            value: 38.57
            name: Test WER
            args:
              - learning_rate: 0.0003
              - train_batch_size: 64
              - eval_batch_size: 8
              - seed: 42
              - gradient_accumulation_steps: 2
              - total_train_batch_size: 128
              - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
              - lr_scheduler_type: linear
              - lr_scheduler_warmup_steps: 1000
              - num_epochs: 30
              - mixed_precision_training: Native AMP
          - type: cer
            value: 10.98
            name: Test CER
            args:
              - learning_rate: 0.0003
              - train_batch_size: 64
              - eval_batch_size: 8
              - seed: 42
              - gradient_accumulation_steps: 2
              - total_train_batch_size: 128
              - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
              - lr_scheduler_type: linear
              - lr_scheduler_warmup_steps: 1000
              - num_epochs: 30
              - mixed_precision_training: Native AMP

wav2vec2-large-xls-r-300m-Swedish

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

  • Loss: 0.4286
  • Wer: 0.2729
  • Cer: 0.0858

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0003
  • train_batch_size: 64
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 128
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 1000
  • num_epochs: 30
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Cer
4.6203 5.49 500 2.8904 1.0 1.0
1.147 10.98 1000 0.5255 0.4107 0.1304
0.5246 16.48 1500 0.4598 0.3342 0.1058
0.378 21.97 2000 0.4316 0.2991 0.0949
0.298 27.47 2500 0.4286 0.2729 0.0858

Framework versions

  • Transformers 4.17.0.dev0
  • Pytorch 1.10.2+cu102
  • Datasets 1.18.2.dev0
  • Tokenizers 0.11.0