Curiousfox/whisper-small_new_ver2

Hokkien-to-Tai Lo Whisper ver 1.2

This model is a fine-tuned version of openai/whisper-small on the Common Voice 16.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6145
• Wer: 127.2585

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: 8
• 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: linear
• lr_scheduler_warmup_steps: 1000
• training_steps: 8000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9814	0.256	800	0.9274	132.0108
0.8335	0.512	1600	0.8062	128.4465
0.7856	0.768	2400	0.7318	131.0020
0.6938	1.024	3200	0.6892	127.7740
0.6089	1.28	4000	0.6609	125.5100
0.6078	1.536	4800	0.6377	128.8276
0.5756	1.792	5600	0.6244	126.0031
0.5764	2.048	6400	0.6145	127.2585

Framework versions

• Transformers 4.49.0
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0