metadata

language:
  - vi
license: apache-2.0
tags:
  - whisper-event
  - generated_from_trainer
datasets:
  - common_voice_11_0
metrics:
  - wer
model-index:
  - name: Whisper Medium VI - Multi - Augmented
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: common_voice_11_0
          type: common_voice_11_0
          config: vi
          split: test
          args: vi
        metrics:
          - type: wer
            value: 16.659355121737224
            name: Wer
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: mozilla-foundation/common_voice_11_0
          type: mozilla-foundation/common_voice_11_0
          config: vi
          split: test
        metrics:
          - type: wer
            value: 16.63
            name: WER
          - type: cer
            value: 7.74
            name: CER
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: google/fleurs
          type: google/fleurs
          config: vi_vn
          split: test
        metrics:
          - type: wer
            value: 9.04
            name: WER
          - type: cer
            value: 4.81
            name: CER
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: vivos
          type: vivos
          split: test
        metrics:
          - type: wer
            value: 8.53
            name: WER
          - type: cer
            value: 3.67
            name: CER

Whisper Medium VI - Multi - Augmented

This model is a fine-tuned version of openai/whisper-medium on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:

Loss: 0.3696
Wer: 16.6594
Cer: 7.7625

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: 1e-05
train_batch_size: 32
eval_batch_size: 16
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 5000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.1992	1.8	1000	0.2726	17.4929	8.2562
0.0402	3.6	2000	0.3317	17.4929	8.2588
0.0073	5.4	3000	0.3429	17.6793	8.8913
0.0014	7.19	4000	0.3599	19.0283	9.5103
0.0006	8.99	5000	0.3696	16.6594	7.7625

Framework versions

Transformers 4.26.0.dev0
Pytorch 1.13.1+cu117
Datasets 2.7.1.dev0
Tokenizers 0.13.2