steja/whisper-small-german

Whisper Small german

This model is a fine-tuned version of openai/whisper-small on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3092
• Wer: 12.2134

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: 32
• seed: 42
• gradient_accumulation_steps: 2
• 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: 6000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1427	1.99	1000	0.2298	12.2134
0.032	3.98	2000	0.2521	12.4540
0.0066	5.96	3000	0.2766	12.3981
0.0036	7.95	4000	0.2932	12.5753
0.0023	9.94	5000	0.3041	12.5719
0.0019	11.93	6000	0.3092	12.6312

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.0+cu117
• Datasets 2.7.1
• Tokenizers 0.13.2