volkan-aslan/whisper-small-tr-v4

Whisper Small TR V4 - Volkan ASLAN

This model is a fine-tuned version of openai/whisper-small on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4422
• Wer: 26.1688

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: 3.75e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3197	0.6894	1000	0.4490	34.3677
0.1477	1.3785	2000	0.4296	31.9782
0.0718	2.0676	3000	0.4213	31.0806
0.0747	2.7570	4000	0.4210	30.0775
0.038	3.4461	5000	0.4247	29.0658
0.0192	4.1351	6000	0.4267	28.1274
0.0121	4.8245	7000	0.4248	27.4376
0.0045	5.5136	8000	0.4347	26.9011
0.0009	6.2027	9000	0.4387	26.4294
0.0008	6.8921	10000	0.4422	26.1688

Framework versions

• Transformers 4.51.3
• Pytorch 2.5.1+cu121
• Datasets 3.5.1
• Tokenizers 0.21.1