ahmedlh/whisper-tiny-en_v8

Whisper tiny En v6 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: 0.6790
• Wer Ortho: 29.6940
• Wer: 20.9923

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: 2e-06
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.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: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
2.0788	0.0824	50	2.4317	32.8217	23.5413
1.1756	0.1647	100	1.4161	33.8546	24.7759
0.3704	0.2471	150	0.7644	31.5185	23.9919
0.3139	0.3295	200	0.7199	30.0753	21.7199
0.3249	0.4119	250	0.6991	29.9208	21.6495
0.3264	0.4942	300	0.6930	29.6216	21.3773
0.3069	0.5766	350	0.6848	29.7616	21.5838
0.2899	0.6590	400	0.6806	29.2161	20.6591
0.293	0.7414	450	0.6765	29.4140	20.8468
0.3009	0.8237	500	0.6790	29.6940	20.9923

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 2.14.6
• Tokenizers 0.21.1