andrewbawitlung/whisper-small-mizonal2-lus-v25.03.05

Whisper Small

This model is a fine-tuned version of openai/whisper-small on the MiZonal v2.0.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3063
• Wer: 18.1954

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: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 6000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4781	0.66	500	0.4381	31.6513
0.2315	1.31	1000	0.3232	26.3651
0.2077	1.97	1500	0.2701	21.4102
0.1065	2.62	2000	0.2621	20.4574
0.051	3.28	2500	0.2723	19.9022
0.0538	3.93	3000	0.2631	19.1897
0.0227	4.59	3500	0.2811	19.3305
0.0099	5.24	4000	0.2904	19.2477
0.0092	5.9	4500	0.2900	18.6511
0.004	6.55	5000	0.3005	18.1871
0.0021	7.21	5500	0.3043	18.2948
0.0022	7.86	6000	0.3063	18.1954

Framework versions

• Transformers 4.37.2
• Pytorch 2.3.1+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1