Graduation_Project_Distil_Whisper_base3

This model is a fine-tuned version of Baselhany/Graduation_Project_Distil_Whisper_base3 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1698
• Wer: 0.3720

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.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
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.4782	1.0	520	0.1344	0.4069
1.0405	2.0	1040	0.1495	0.4004
0.5482	3.0	1560	0.1573	0.3436
0.4041	4.0	2080	0.1587	0.4126
0.3115	5.0	2600	0.1569	0.3798
0.2612	6.0	3120	0.1515	0.4272
0.187	7.0	3640	0.1577	0.3917
0.1596	8.0	4160	0.1538	0.4334
0.1465	9.0	4680	0.1497	0.3771
0.1149	10.0	5200	0.1506	0.4192
0.0935	11.0	5720	0.1465	0.3974
0.0849	12.0	6240	0.1483	0.3979
0.0686	13.0	6760	0.1472	0.4237
0.0533	14.0	7280	0.1483	0.4251
0.0382	14.9726	7785	0.1498	0.4302

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1