whisper-small-ver2

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

• Loss: 0.6016
• Cer: 15.2224

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: 20
• eval_batch_size: 16
• 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: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.1387	3.5	500	0.4537	16.3581
0.0218	6.99	1000	0.5024	16.0241
0.0041	10.49	1500	0.5310	15.2224
0.0028	13.99	2000	0.5554	18.7918
0.0009	17.48	2500	0.5743	14.9265
0.0012	20.98	3000	0.5883	15.2510
0.0006	24.48	3500	0.5982	15.2128
0.0006	27.97	4000	0.6016	15.2224

Framework versions

• Transformers 4.39.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.17.1
• Tokenizers 0.15.1