multiclass-fz-enc-tiny

This model is a fine-tuned version of openai/whisper-tiny on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1161
• Wer: 17.2365
• Cer: 10.5891

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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.1515	4.5872	500	0.2850	26.4482	18.5933
0.0139	9.1743	1000	0.1469	42.4976	27.6108
0.0015	13.7615	1500	0.1307	23.5043	15.8193
0.0008	18.3486	2000	0.1246	17.0940	10.2800
0.0005	22.9358	2500	0.1207	17.0465	10.4260
0.0004	27.5229	3000	0.1188	17.3789	10.5290
0.0003	32.1101	3500	0.1178	17.3314	10.5634
0.0003	36.6972	4000	0.1168	17.0465	10.4088
0.0002	41.2844	4500	0.1163	17.2365	10.5891
0.0002	45.8716	5000	0.1161	17.2365	10.5891

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0