SmartWhatt/rangsitra-whisper-sm-replace-0.25

Rangsitra Whisper

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:

• Cer: 6.6204
• Wer: 22.1856
• Loss: 0.1924

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: 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: warmup_stable_decay
• lr_scheduler_warmup_steps: 1017
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Cer	Wer	Validation Loss
0.1918	1.0	2035	9.2807	32.4529	0.2132
0.1169	2.0	4070	7.9355	26.8911	0.1865
0.0653	3.0	6105	7.6426	25.3136	0.1859
0.0246	4.0	8140	6.8334	22.5960	0.1875
0.0044	5.0	10175	6.6204	22.1856	0.1924

Framework versions

• Transformers 4.52.4
• Pytorch 2.9.1+cu128
• Datasets 3.6.0
• Tokenizers 0.21.4