naji02010101/whisper-tiny-en_v9

Whisper tiny En v9 Naji

This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6733
• Wer Ortho: 29.3915
• Wer: 20.8169

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-06
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 100
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.9714	0.16	100	1.4118	35.6127	25.7941
0.3325	0.32	200	0.7783	31.1177	22.9749
0.3148	0.48	300	0.7235	29.8563	21.5616
0.3247	0.64	400	0.7005	29.6142	21.2918
0.3384	0.8	500	0.6928	29.4267	21.2006
0.2685	0.96	600	0.6914	29.0518	20.4825
0.2862	1.12	700	0.6795	29.3642	21.0296
0.2466	1.28	800	0.6792	29.4540	21.1626
0.2803	1.44	900	0.6736	29.0596	20.4635
0.2414	1.6	1000	0.6755	29.1533	20.5737
0.2783	1.76	1100	0.6714	29.4618	20.9384
0.2696	1.92	1200	0.6775	29.6063	20.8777
0.2592	2.08	1300	0.6713	29.4540	20.8245
0.2193	2.24	1400	0.6733	29.3915	20.8169

Framework versions

• Transformers 4.52.4
• Pytorch 2.7.1+cu126
• Datasets 2.14.6
• Tokenizers 0.21.1