multiclass-fz-enc-tiny-en

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

• Loss: 0.1601
• Wer: 11.4435
• Cer: 7.4545

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
0.1904	4.5872	500	0.3361	511.2061	542.4081
0.0141	9.1743	1000	0.1815	17.2365	18.1553
0.0016	13.7615	1500	0.1727	12.1557	11.6798
0.0008	18.3486	2000	0.1674	12.3932	11.7657
0.0005	22.9358	2500	0.1645	12.1083	11.8945
0.0004	27.5229	3000	0.1628	11.9183	11.6970
0.0003	32.1101	3500	0.1620	11.8708	7.8409
0.0003	36.6972	4000	0.1610	21.3675	17.9320
0.0002	41.2844	4500	0.1601	11.4910	7.4631
0.0002	45.8716	5000	0.1601	11.4435	7.4545

Framework versions

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