abbenedekwhisper-tiny.en-finetuning3-D3K

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

• Loss: 3.2102
• Cer: 48.9705
• Wer: 91.3907
• Ser: 100.0
• Cer Clean: 6.0657
• Wer Clean: 12.9139
• Ser Clean: 13.1579

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: 5e-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: 10
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer	Ser	Cer Clean	Wer Clean	Ser Clean
6.2196	1.06	200	5.5899	52.5320	112.9139	100.0	7.3456	14.2384	14.9123
5.2943	2.13	400	4.9201	52.4763	110.2649	100.0	7.6238	14.9007	15.7895
4.5662	3.19	600	4.4164	51.1964	105.6291	100.0	7.6238	14.9007	15.7895
4.0943	4.26	800	4.0825	50.5843	103.3113	100.0	7.1786	14.5695	14.9123
3.6948	5.32	1000	3.7923	51.5303	101.9868	100.0	6.3439	12.9139	13.1579
3.3742	6.38	1200	3.5565	50.3617	98.3444	100.0	6.3439	13.5762	14.0351
3.1519	7.45	1400	3.3895	49.0262	93.7086	100.0	6.3439	13.5762	14.0351
2.9995	8.51	1600	3.2845	48.6366	92.7152	100.0	6.3439	13.5762	14.0351
2.9152	9.57	1800	3.2282	47.9688	91.7219	100.0	6.0657	12.9139	13.1579
2.884	10.64	2000	3.2102	48.9705	91.3907	100.0	6.0657	12.9139	13.1579

Framework versions

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