gwanju_medium2_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: 0.3592
• Cer: 44.1724

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: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.4478	0.2964	1000	0.4342	48.5130
0.3965	0.5928	2000	0.3973	401.9235
0.3913	0.8892	3000	0.3777	23.1253
0.236	1.1855	4000	0.3685	27.0142
0.2617	1.4819	5000	0.3638	31.4542
0.2245	1.7783	6000	0.3556	24.0368
0.1363	2.0747	7000	0.3582	27.9426
0.1232	2.3711	8000	0.3629	29.3680
0.1299	2.6675	9000	0.3615	40.2485
0.1191	2.9638	10000	0.3592	44.1724

Framework versions

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