metadata

library_name: transformers
language:
  - gsw
license: apache-2.0
base_model: openai/whisper-large-v2
tags:
  - generated_from_trainer
datasets:
  - notebotIE/zh_split_preprocessed
metrics:
  - wer
model-index:
  - name: Whisper Large V2 - Swiss German
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: SwissDialDataset_ETH
          type: notebotIE/zh_split_preprocessed
        metrics:
          - name: Wer
            type: wer
            value: 0.1923865967631639

Whisper Large V2 - Swiss German

This model is a fine-tuned version of openai/whisper-large-v2 on the SwissDialDataset_ETH dataset. It achieves the following results on the evaluation set:

Loss: 0.2725
Wer Ortho: 0.2707
Wer: 0.1924
Cer: 0.0743

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: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: constant_with_warmup
lr_scheduler_warmup_steps: 50
training_steps: 1000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer	Cer
0.1277	1.2300	250	0.2468	0.2631	0.1855	0.0556
0.0726	2.4600	500	0.2414	0.9085	0.6793	0.5239
0.0405	3.6900	750	0.2519	0.5194	0.6558	0.4984
0.0296	4.9200	1000	0.2725	0.2707	0.1924	0.0743

Framework versions

Transformers 4.46.3
Pytorch 2.5.1+cu121
Datasets 3.2.0
Tokenizers 0.20.3