tylee4/lr_3e-5

Whisper Tiny Ko - 0.1

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

• Loss: 0.3568
• Cer: 11.5477

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: 3e-06
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• 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
• lr_scheduler_warmup_steps: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.5846	0.4488	500	0.5778	17.0684
0.5078	0.8977	1000	0.4889	14.7653
0.438	1.3465	1500	0.4454	13.8077
0.3945	1.7953	2000	0.4158	12.9504
0.3629	2.2442	2500	0.3952	12.4517
0.3363	2.6930	3000	0.3804	12.3059
0.3301	3.1418	3500	0.3699	11.8438
0.3291	3.5907	4000	0.3626	11.7568
0.2829	4.0395	4500	0.3582	11.5566
0.3056	4.4883	5000	0.3568	11.5477

Framework versions

• Transformers 4.53.0
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.21.2