chungnam_firestation_tiny_model

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

• Loss: 0.0616
• Cer: 59.7122

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: 250
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
1.9425	0.6623	100	1.5318	164.8345
0.9552	1.3245	200	0.9569	148.0576
0.6638	1.9868	300	0.6359	126.2734
0.3385	2.6490	400	0.4440	113.9856
0.1794	3.3113	500	0.2976	96.2590
0.1331	3.9735	600	0.2152	184.6619
0.0666	4.6358	700	0.1515	97.4964
0.0354	5.2980	800	0.1093	77.2662
0.0311	5.9603	900	0.0887	77.2950
0.015	6.6225	1000	0.0749	102.1007
0.0089	7.2848	1100	0.0686	68.5180
0.0052	7.9470	1200	0.0646	65.3237
0.0036	8.6093	1300	0.0614	59.2518
0.0029	9.2715	1400	0.0607	64.6043
0.0025	9.9338	1500	0.0607	67.5396
0.003	10.5960	1600	0.0612	59.2806
0.0019	11.2583	1700	0.0614	56.1151
0.0018	11.9205	1800	0.0613	55.6547
0.0017	12.5828	1900	0.0616	60.1151
0.0017	13.2450	2000	0.0616	59.7122

Framework versions

• Transformers 4.41.2
• Pytorch 2.2.2+cu121
• Datasets 3.2.0
• Tokenizers 0.19.1