metadata

language:
  - pl
tags:
  - whisper-event
  - generated_from_trainer
datasets:
  - mozilla-foundation/common_voice_11_0
  - google/fleurs
metrics:
  - wer
model-index:
  - name: Whisper Large v2 PL
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: Common Voice 11.0
          type: mozilla-foundation/common_voice_11_0
          config: pl
          split: test
          args: pl
        metrics:
          - type: wer
            value: 7.280175959972464
            name: WER
          - type: wer
            value: 7.31
            name: WER
          - type: wer_without_norm
            value: 20.18
            name: WER unnormalized
          - type: cer
            value: 2.08
            name: CER
          - type: mer
            value: 7.27
            name: MER
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: facebook/voxpopuli
          type: facebook/voxpopuli
          config: pl
          split: test
        metrics:
          - type: wer
            value: 9.61
            name: WER
          - type: wer_without_norm
            value: 30.33
            name: WER unnormalized
          - type: cer
            value: 5.5
            name: CER
          - type: mer
            value: 9.45
            name: MER
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: google/fleurs
          type: google/fleurs
          config: pl_pl
          split: test
        metrics:
          - type: wer
            value: 8.68
            name: WER
          - type: wer_without_norm
            value: 29.33
            name: WER unnormalized
          - type: cer
            value: 3.63
            name: CER
          - type: mer
            value: 8.62
            name: MER

Whisper Large v2 PL

This model is a fine-tuned version of bardsai/whisper-large-v2-pl on the Common Voice 11.0 and the FLEURS datasets. It achieves the following results on the evaluation set:

Loss: 0.3684
Wer: 7.2802

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e-05
train_batch_size: 8
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 64
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 2100
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0047	1.35	700	0.3428	8.5562
0.0011	2.7	1400	0.3605	7.5505
0.0003	4.05	2100	0.3684	7.2802

Framework versions

Transformers 4.26.0.dev0
Pytorch 1.13.0+cu117
Datasets 2.7.1.dev0
Tokenizers 0.13.2