aslinguist/whisper-central-paiwan

whisper-coastal-paiwan

This model is a fine-tuned version of openai/whisper-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9448
• Wer: 37.2570

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use 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: 200
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2293	3.8760	500	0.7556	49.3880
0.0754	7.7519	1000	0.8284	53.9957
0.0354	11.6279	1500	0.8634	44.2405
0.0305	15.5039	2000	0.9155	39.8128
0.014	19.3798	2500	0.9610	41.5407
0.0119	23.2558	3000	0.9340	38.0490
0.002	27.1318	3500	0.9220	37.9050
0.0026	31.0078	4000	0.9313	36.8611
0.0018	34.8837	4500	0.9397	36.9690
0.0013	38.7597	5000	0.9448	37.2570

Framework versions

• Transformers 4.51.2
• Pytorch 2.2.2+cu118
• Datasets 3.5.0
• Tokenizers 0.21.1