mmtg's picture
Upload tokenizer
aa09f8b verified
|
raw
history blame
2.06 kB
metadata
license: apache-2.0
tags:
  - generated_from_trainer
base_model: facebook/wav2vec2-large-xlsr-53
datasets:
  - common_voice_11_0
metrics:
  - wer
model-index:
  - name: wav2vec2-xlsr-cv-11-layer
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: common_voice_11_0
          type: common_voice_11_0
          config: nan-tw
          split: test
          args: nan-tw
        metrics:
          - type: wer
            value: 1.0165016501650166
            name: Wer

wav2vec2-xlsr-cv-11-layer

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:

  • Loss: 5.4908
  • Wer: 1.0165

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: 0.003
  • 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: 40
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
19.3425 8.0808 400 5.2991 1.0
4.0894 16.1616 800 5.5709 1.0
3.8676 24.2424 1200 5.5382 1.0055
3.7673 32.3232 1600 5.4908 1.0165

Framework versions

  • Transformers 4.41.2
  • Pytorch 2.1.2
  • Datasets 2.19.2
  • Tokenizers 0.19.1