metadata

license: apache-2.0
tags:
  - generated_from_trainer
datasets:
  - fleurs
metrics:
  - wer
model-index:
  - name: whisper-training-blog
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: fleurs
          type: fleurs
          config: sv_se
          split: validation
          args: sv_se
        metrics:
          - name: Wer
            type: wer
            value: 180.05748044068338

whisper-training-blog

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

Loss: 1.0050
Wer: 180.0575

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: 7.5e-06
train_batch_size: 16
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.3
training_steps: 448
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.4112	0.1	44	1.4919	245.3457
1.0502	0.2	88	1.2255	220.1501
0.9033	0.29	132	1.1203	206.2430
0.8141	1.06	176	1.0675	201.9639
0.8029	1.16	220	1.0394	178.3650
0.6324	1.25	264	1.0301	221.2997
0.6972	2.02	308	1.0134	176.6725
0.6052	2.12	352	1.0065	194.7150
0.6047	2.21	396	1.0030	160.9133
0.5849	2.31	440	1.0050	180.0575

Framework versions

Transformers 4.27.3
Pytorch 2.0.0+cu118
Datasets 2.10.1
Tokenizers 0.13.3