jstoone/whisper-medium-da

Whisper Medium Danish (CV11 + FLEAURS)

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

• Loss: 0.5814
• Wer: 13.7086

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: 8e-06
• train_batch_size: 32
• 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_steps: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0265	3.14	1000	0.3690	14.7607
0.0063	6.29	2000	0.4342	14.0926
0.0016	9.43	3000	0.4847	14.3609
0.002	12.58	4000	0.4919	14.1715
0.0013	15.72	5000	0.5114	14.2294
0.0014	18.87	6000	0.5197	13.9137
0.0003	22.01	7000	0.5422	14.1978
0.0001	25.16	8000	0.5659	13.8716
0.0001	28.3	9000	0.5772	13.7296
0.0001	31.45	10000	0.5814	13.7086

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.1+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2