egemenakdeniz/whisper-medium-tr-finetunedV3

Whisper-medium-tr-finetunedV2

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

• Loss: 0.2645
• Wer: 16.2853

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: 4
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 500
• training_steps: 25600

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2293	0.3018	2000	0.2808	22.5592
0.2207	0.6037	4000	0.2697	21.5055
0.2	0.9055	6000	0.2532	20.2982
0.0704	1.2074	8000	0.2567	19.3854
0.073	1.5092	10000	0.2441	18.5946
0.0761	1.8110	12000	0.2469	18.6514
0.0307	2.1129	14000	0.2493	17.2212
0.025	2.4147	16000	0.2545	17.4547
0.0242	2.7166	18000	0.2535	17.0319
0.0077	3.0184	20000	0.2460	16.0834
0.0142	3.3203	22000	0.2572	16.3589
0.0053	3.6221	24000	0.2645	16.2853

Framework versions

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