whisper-finetuned-nan-tw-v3-torbo-20-epoch

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2611
• Wer: 100.0

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: 5e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• 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
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2238	1.0616	500	0.2174	100.0
0.1871	2.1233	1000	0.2099	100.0
0.1606	3.1849	1500	0.2050	100.0
0.1315	4.2465	2000	0.2134	100.0
0.1087	5.3082	2500	0.2247	100.0
0.0969	6.3698	3000	0.2269	100.0
0.0851	7.4315	3500	0.2381	100.0
0.0793	8.4931	4000	0.2412	100.0
0.0737	9.5547	4500	0.2406	100.0
0.07	10.6164	5000	0.2452	99.9569
0.0669	11.6780	5500	0.2435	100.0
0.0623	12.7396	6000	0.2447	100.0
0.0605	13.8013	6500	0.2490	100.0
0.0563	14.8629	7000	0.2524	99.9569
0.0525	15.9245	7500	0.2536	100.0
0.0485	16.9862	8000	0.2554	100.0
0.0435	18.0468	8500	0.2591	100.0
0.0386	19.1084	9000	0.2611	100.0

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.1
• Tokenizers 0.21.0