younghoonKIM/exp1

whisper_large

This model is a fine-tuned version of openai/whisper-large on the MAICON2023_noise dataset. It achieves the following results on the evaluation set:

• Loss: 0.2609
• Cer: 27.9801

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: 4
• 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
0.6254	0.36	1000	0.5211	39.0406
0.3894	0.71	2000	0.3733	23.1574
0.0932	1.07	3000	0.2990	24.4794
0.0952	1.43	4000	0.2609	27.9801

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0