fs-w-he-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: 2.7192
• Wer: 173.9316
• Cer: 146.1783

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
10.4685	4.5872	500	10.1251	536.1823	471.2470
3.0813	9.1743	1000	3.6378	239.3162	199.3988
1.4313	13.7615	1500	2.4350	250.0	207.9268
0.6499	18.3486	2000	1.9427	376.7331	318.6104
0.336	22.9358	2500	1.9422	245.7265	195.3538
0.0683	27.5229	3000	2.2195	265.3846	224.6908
0.0104	32.1101	3500	2.3854	177.7778	148.1621
0.0013	36.6972	4000	2.6033	175.7360	158.7513
0.0003	41.2844	4500	2.6985	155.6030	127.8169
0.0002	45.8716	5000	2.7192	173.9316	146.1783

Framework versions

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