gyeongsan_1_6_model_50000WER

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

• Loss: 0.1396
• Wer: 18.0750

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: 32
• 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: 100
• training_steps: 50000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1764	0.2252	5000	0.1840	20.9807
0.1526	0.4503	10000	0.1684	19.8061
0.16	0.6755	15000	0.1586	19.3306
0.1449	0.9006	20000	0.1511	18.7879
0.1048	1.1258	25000	0.1485	18.9708
0.1045	1.3509	30000	0.1452	18.5721
0.0938	1.5761	35000	0.1410	18.1488
0.105	1.8012	40000	0.1379	17.8164
0.0629	2.0264	45000	0.1388	17.9810
0.0636	2.2515	50000	0.1396	18.0750

Framework versions

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