whisper-tiny-ur-5h

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

• Loss: 0.8447
• Wer: 53.3013

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: 8
• 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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9092	0.8460	500	0.8859	59.1489
0.6742	1.6920	1000	0.8084	58.6685
0.4835	2.5381	1500	0.7822	52.2992
0.3839	3.3841	2000	0.7792	55.2505
0.3781	4.2301	2500	0.7976	57.1311
0.2321	5.0761	3000	0.8026	53.3425
0.2785	5.9222	3500	0.8129	54.4955
0.2542	6.7682	4000	0.8306	53.6033
0.2232	7.6142	4500	0.8451	53.9602
0.1776	8.4602	5000	0.8447	53.3013

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1