Whisper-large-v3-turbo-tr-finetuned

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the commonvoice_17_tr_fixed dataset. It achieves the following results on the evaluation set:

• Loss: 0.1598
• Wer: 13.1893
• Cer: 2.8903

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
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 248
• training_steps: 1650

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.1376	0.2655	110	0.1511	13.7193	2.9344
0.1419	0.5311	220	0.1723	15.5702	3.6979
0.1439	0.7966	330	0.1752	15.6375	3.6395
0.1235	1.0604	440	0.1764	14.8004	3.2088
0.0863	1.3259	550	0.1701	14.8572	3.2624
0.0824	1.5914	660	0.1651	14.6595	3.3526
0.0852	1.8570	770	0.1664	14.5354	3.1855
0.042	2.1207	880	0.1743	14.9645	3.3255
0.0419	2.3862	990	0.1658	14.0853	3.1087
0.0475	2.6518	1100	0.1571	13.5216	2.8987
0.0425	2.9173	1210	0.1578	13.3429	2.8445
0.0247	3.1811	1320	0.1648	13.1220	2.8129
0.0224	3.4466	1430	0.1619	13.1325	2.9317
0.0223	3.7121	1540	0.1603	13.0737	2.9002
0.0183	3.9777	1650	0.1598	13.1893	2.8903

Framework versions

• Transformers 4.57.1
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.1