alpcansoydas/whisper-large-v2-tr-ft-24-03-26-encoder-only-25ksamples-simulated

Whisper for Turkish Call Centers

This model is a fine-tuned version of openai/whisper-large-v2 on the Custom turkish call center simulated data dataset. It achieves the following results on the evaluation set:

• Loss: 0.2663

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: 5e-06
• train_batch_size: 16
• 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: 4500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.3106	0.1777	250	0.3197
0.2659	0.3554	500	0.2988
0.2773	0.5330	750	0.2874
0.2977	0.7107	1000	0.2810
0.2756	0.8884	1250	0.2781
0.2531	1.0661	1500	0.2753
0.2490	1.2438	1750	0.2741
0.2445	1.4215	2000	0.2706
0.2430	1.5991	2250	0.2694
0.3167	1.7768	2500	0.2686
0.2514	1.9545	2750	0.2665
0.2445	2.1322	3000	0.2702
0.2365	2.3099	3250	0.2675
0.1800	2.4876	3500	0.2667
0.2421	2.6652	3750	0.2660
0.2282	2.8429	4000	0.2656
0.2306	3.0206	4250	0.2653
0.2174	3.1983	4500	0.2663

Framework versions

• Transformers 5.3.0
• Pytorch 2.10.0+cu128
• Datasets 4.8.4
• Tokenizers 0.22.2