mangoo111/stt_whisper_mixed

AIHub_non-face-to-face-care_data_model_synthesis

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

• Loss: 0.4661
• Cer: 86.6106
• Normalized Cer: 0.1083

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: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Normalized Cer
0.6939	2.5	1000	0.7123	100.3264	0.1254
0.4626	5.0	2000	0.4787	93.8130	0.1173
0.3828	7.5	3000	0.4661	86.6106	0.1083
0.3207	10.0	4000	0.4811	99.8114	0.1248

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1