abbenedekwhisper-small.en-finetuning2-D3K

This model is a fine-tuned version of openai/whisper-small.en on the None dataset. It achieves the following results on the evaluation set:

• Loss: 7.0682
• Cer: 53.2554
• Wer: 135.0993
• Ser: 100.0
• Cer Clean: 0.5008
• Wer Clean: 0.6623
• Ser Clean: 1.7544

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-08
• train_batch_size: 16
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5
• training_steps: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer	Ser	Cer Clean	Wer Clean	Ser Clean
7.4342	0.05	10	7.0727	53.2554	135.0993	100.0	0.5008	0.6623	1.7544
7.2902	0.11	20	7.0722	53.2554	135.0993	100.0	0.5008	0.6623	1.7544
6.9726	0.16	30	7.0711	53.2554	135.0993	100.0	0.5008	0.6623	1.7544
7.3598	0.21	40	7.0705	53.2554	135.0993	100.0	0.5008	0.6623	1.7544
7.0578	0.27	50	7.0682	53.2554	135.0993	100.0	0.5008	0.6623	1.7544

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.2+cu121
• Datasets 2.14.5
• Tokenizers 0.15.2