Whisper_Base

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

• Loss: 1.7755
• Wer: 68.7469

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: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 6000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4487	3.0143	500	1.2483	56.6216
0.1295	7.0057	1000	1.2956	57.3096
0.0439	10.02	1500	1.3855	63.5135
0.0128	14.0113	2000	1.4395	57.5799
0.003	18.0027	2500	1.6753	63.6118
0.001	21.017	3000	1.6112	63.0221
0.0005	25.0083	3500	1.7148	63.8206
0.0003	28.0227	4000	1.6958	72.4816
0.0003	32.014	4500	1.7403	70.8600
0.0002	36.0053	5000	1.7782	79.4840
0.0002	39.0197	5500	1.7614	72.3587
0.0002	43.011	6000	1.7755	68.7469

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu121
• Datasets 3.3.1
• Tokenizers 0.21.0