whisper-multiclass-lang-en-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: 0.1349
• Wer: 7.7873
• Cer: 5.4535

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
0.0661	4.5872	500	0.2294	11.3485	7.0165
0.0033	9.1743	1000	0.1563	17.7113	11.5596
0.0005	13.7615	1500	0.1456	8.0247	5.3246
0.0003	18.3486	2000	0.1414	7.8348	5.1271
0.0002	22.9358	2500	0.1390	8.5945	5.7798
0.0002	27.5229	3000	0.1374	8.5470	5.9258
0.0001	32.1101	3500	0.1362	7.9772	5.5136
0.0001	36.6972	4000	0.1355	7.8348	5.4449
0.0001	41.2844	4500	0.1351	7.7873	5.4449
0.0001	45.8716	5000	0.1349	7.7873	5.4535

Framework versions

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