whisper-small-test

This model is a fine-tuned version of openai/whisper-small on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1505
• Wer: 17.6506

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: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 16000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5739	0.1088	1000	0.5302	46.0061
0.4512	0.2176	2000	0.4017	36.9299
0.3804	0.3264	3000	0.3504	33.6799
0.3382	0.4353	4000	0.3146	31.3559
0.3355	0.5441	5000	0.2815	28.3068
0.3085	0.6529	6000	0.2592	26.9714
0.2943	0.7617	7000	0.2387	25.6010
0.2802	0.8705	8000	0.2206	23.7800
0.2611	0.9793	9000	0.2027	22.4196
0.179	1.0881	10000	0.1917	21.3924
0.1712	1.1970	11000	0.1825	20.4576
0.1723	1.3058	12000	0.1737	19.8497
0.1536	1.4146	13000	0.1654	18.7739
0.1509	1.5234	14000	0.1576	18.2609
0.1511	1.6322	15000	0.1526	17.8403
0.1518	1.7410	16000	0.1505	17.6506

Framework versions

• Transformers 4.46.2
• Pytorch 2.4.0
• Datasets 3.1.0
• Tokenizers 0.20.3