chungnam_large2_model

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

• Loss: 0.0603
• Cer: 21.8993

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: 100
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.1193	1.6556	250	0.1941	78.9065
0.0298	3.3113	500	0.0792	75.7698
0.0143	4.9669	750	0.0694	13.9568
0.0031	6.6225	1000	0.0602	8.8633
0.0005	8.2781	1250	0.0572	16.0
0.0027	9.9338	1500	0.0541	15.1942
0.0001	11.5894	1750	0.0563	16.9209
0.0003	13.2450	2000	0.0554	18.7914
0.0001	14.9007	2250	0.0572	18.9065
0.0001	16.5563	2500	0.0581	20.6619
0.0001	18.2119	2750	0.0588	19.7410
0.0	19.8675	3000	0.0593	21.0072
0.0	21.5232	3250	0.0598	22.2734
0.0	23.1788	3500	0.0601	21.9856
0.0	24.8344	3750	0.0603	21.8993
0.0	26.4901	4000	0.0603	21.8993

Framework versions

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