whisper-tiny-kor_eng_tiny_ed_ev

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

• Loss: 0.5072
• Cer: 9.4241

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-05
• train_batch_size: 12
• eval_batch_size: 6
• seed: 42
• optimizer: Use OptimizerNames.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: 200
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.6251	18.1818	200	0.3865	8.6387
0.003	36.3636	400	0.4316	9.5550
0.0001	54.5455	600	0.4462	9.4241
0.0	72.7273	800	0.4552	9.6859
0.0	90.9091	1000	0.4627	9.1623
0.0	109.0909	1200	0.4680	9.1623
0.0	127.2727	1400	0.4729	9.1623
0.0	145.4545	1600	0.4776	9.1623
0.0	163.6364	1800	0.4809	9.1623
0.0	181.8182	2000	0.4849	9.2932
0.0	200.0	2200	0.4885	9.2932
0.0	218.1818	2400	0.4924	9.1623
0.0	236.3636	2600	0.4955	9.2932
0.0	254.5455	2800	0.4987	9.2932
0.0	272.7273	3000	0.5012	9.2932
0.0	290.9091	3200	0.5030	9.4241
0.0	309.0909	3400	0.5048	9.4241
0.0	327.2727	3600	0.5063	9.4241
0.0	345.4545	3800	0.5073	9.4241
0.0	363.6364	4000	0.5072	9.4241

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu126
• Datasets 3.5.0
• Tokenizers 0.21.1