whisper-medium-uk

This model is a fine-tuned version of openai/whisper-medium on the Common Voice 16.1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3673
• Wer: 20.1065

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: 6e-06
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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
0.1947	0.94	1000	0.2269	22.7263
0.1034	1.89	2000	0.2102	20.6058
0.0572	2.83	3000	0.2192	20.3908
0.0261	3.77	4000	0.2483	21.0204
0.0112	4.72	5000	0.2758	21.1480
0.0058	5.66	6000	0.3166	20.3270
0.0026	6.6	7000	0.3268	20.5877
0.0017	7.55	8000	0.3483	20.0455
0.0006	8.49	9000	0.3635	20.0996
0.0005	9.43	10000	0.3673	20.1065

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 2.2.0+cu121
• Datasets 2.17.1
• Tokenizers 0.15.2