yoona-J/whisper_call_audio

Whisper_call

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

• Loss: 0.5226
• Cer: 81.2616

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: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.8139	2.0	200	0.7714	37.8221
0.3977	4.0	400	0.4928	33.9313
0.1864	6.0	600	0.4458	78.8698
0.0907	8.0	800	0.4532	72.4187
0.0357	10.0	1000	0.4712	89.8265
0.015	12.0	1200	0.4899	88.2432
0.007	14.0	1400	0.5035	85.6325
0.0043	16.0	1600	0.5162	83.3839
0.0037	18.0	1800	0.5205	82.0448
0.0034	20.0	2000	0.5226	81.2616

Framework versions

• Transformers 4.42.0.dev0
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1