whisper-small-en-nonnative

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: 0.4024
• Wer: 40.7599

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: 16
• 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
0.2992	0.2	1000	0.4468	29.3898
0.2918	0.39	2000	0.4240	27.1733
0.3078	0.59	3000	0.4173	28.4974
0.2711	0.79	4000	0.4027	24.5538
0.2813	0.98	5000	0.4029	28.6413
0.1416	1.18	6000	0.4078	25.9931
0.1399	1.38	7000	0.4078	28.8140
0.1478	1.57	8000	0.4070	31.3759
0.1479	1.77	9000	0.4033	33.5636
0.1266	1.97	10000	0.4024	40.7599

Framework versions

• Transformers 4.37.2
• Pytorch 2.2.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1