alpcansoydas/whisper-large-v2-tr-ft-31-03-26-encoder-only-50ksamples-simulated-data

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: 0.2606

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: 5600
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.3234	0.0889	250	0.3185
0.3021	0.1777	500	0.2982
0.3473	0.2666	750	0.2914
0.2801	0.3555	1000	0.2860
0.3699	0.4444	1250	0.2800
0.2698	0.5332	1500	0.2763
0.3133	0.6221	1750	0.2751
0.3027	0.7110	2000	0.2727
0.3042	0.7999	2250	0.2716
0.2616	0.8887	2500	0.2695
0.2800	0.9776	2750	0.2670
0.2515	1.0665	3000	0.2673
0.2804	1.1554	3250	0.2672
0.2644	1.2442	3500	0.2651
0.2243	1.3331	3750	0.2642
0.2685	1.4220	4000	0.2636
0.2434	1.5108	4250	0.2627
0.2065	1.5997	4500	0.2624
0.2786	1.6886	4750	0.2618
0.2829	1.7775	5000	0.2612
0.1872	1.8663	5250	0.2609
0.2964	1.9552	5500	0.2606
0.2305	1.9908	5600	0.2606

Framework versions

• Transformers 5.4.0
• Pytorch 2.10.0+cu128
• Datasets 4.8.4
• Tokenizers 0.22.2