gwanju_medium_model

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

• Loss: 0.3568
• Cer: 47.6919

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
0.4521	0.2964	1000	0.4426	52.3770
0.404	0.5928	2000	0.4035	35.2990
0.3888	0.8892	3000	0.3768	26.0564
0.2392	1.1855	4000	0.3710	23.6799
0.2727	1.4819	5000	0.3638	36.0795
0.227	1.7783	6000	0.3565	27.6930
0.1401	2.0747	7000	0.3566	38.6322
0.124	2.3711	8000	0.3612	34.4801
0.1329	2.6675	9000	0.3592	47.7958
0.1204	2.9638	10000	0.3568	47.6919

Framework versions

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