gwanju2_m__model

This model is a fine-tuned version of openai/whisper-medium on the Marcusxx/gwanju2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5285
• Cer: 31.2542

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: 20000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.4001	0.6046	2000	0.4109	181.6648
0.2492	1.2092	4000	0.3885	23.4726
0.25	1.8138	6000	0.3801	31.3821
0.1398	2.4184	8000	0.3955	23.6811
0.0733	3.0230	10000	0.4135	24.3813
0.0876	3.6276	12000	0.4193	23.1801
0.0402	4.2322	14000	0.4600	28.5421
0.0404	4.8368	16000	0.4663	26.3909
0.0152	5.4414	18000	0.5243	30.2673
0.0048	6.0459	20000	0.5285	31.2542

Framework versions

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