freshpearYoon/train2

whisper_finetune

This model is a fine-tuned version of openai/whisper-base on the aihub_2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3508
• Cer: 12.0915
• Wer: 35.6445

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: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Cer	Validation Loss	Wer
0.3416	0.32	500	12.9022	0.3510	36.6764
0.3344	0.64	1000	12.7008	0.3562	36.9463
0.3119	0.96	1500	12.4164	0.3517	36.5065
0.246	1.28	2000	12.4510	0.3569	36.4665
0.2437	1.6	2500	12.0823	0.3487	36.0543
0.2318	1.92	3000	0.3454	11.9698	35.7519
0.1861	2.24	3500	0.3501	11.9882	35.6895
0.1729	2.56	4000	0.3508	12.0915	35.6445

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 1.12.1+cu113
• Datasets 2.15.0
• Tokenizers 0.15.0