whisper-base-turkish-1

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

• Loss: 0.2586

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-05
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• training_steps: 40000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.3966	0.0625	2500	0.4053
0.3632	0.125	5000	0.3775
0.3093	0.1875	7500	0.3485
0.2731	0.25	10000	0.3204
0.244	0.3125	12500	0.3061
0.2226	0.375	15000	0.2942
0.2025	0.4375	17500	0.2819
0.1873	0.5	20000	0.2763
0.1713	0.5625	22500	0.2687
0.1517	0.625	25000	0.2635
0.1465	0.6875	27500	0.2588
0.1016	1.0228	30000	0.2616
0.1025	1.0853	32500	0.2612
0.1033	1.1478	35000	0.2576
0.0804	1.2103	37500	0.2609
0.0918	1.2728	40000	0.2586

Framework versions

• Transformers 4.45.2
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.20.3