gwanju_small_model

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

• Loss: 0.3975
• Cer: 116.4511

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.5232	0.2964	1000	0.5003	168.6294
0.4416	0.5928	2000	0.4517	414.7252
0.4279	0.8892	3000	0.4226	180.9793
0.2886	1.1855	4000	0.4153	95.8875
0.3188	1.4819	5000	0.4085	161.7609
0.2717	1.7783	6000	0.3988	158.5859
0.1885	2.0747	7000	0.3993	98.1544
0.1747	2.3711	8000	0.4010	155.7158
0.1852	2.6675	9000	0.3986	103.8166
0.1662	2.9638	10000	0.3975	116.4511

Framework versions

• Transformers 4.41.2
• Pytorch 2.2.2+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1