jstoone/whisper-medium-da-cv11

Whisper Medium Danish (Common Voice Corpus 11.0)

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

• Loss: 0.5081
• Wer: 16.1336

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: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0229	7.01	1000	0.4464	16.5124
0.0072	15.0	2000	0.5081	16.1336
0.0048	22.01	3000	0.5193	16.6597
0.0034	30.0	4000	0.5715	16.8964
0.0005	37.01	5000	0.5998	16.4440
0.0007	45.01	6000	0.5908	16.8017
0.0001	53.0	7000	0.6093	16.2336
0.0001	60.01	8000	0.6405	16.3335
0.0	68.0	9000	0.6524	16.2493
0.0	75.01	10000	0.6569	16.2283

Framework versions

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