whisper-base-aug-6-april-v2

This model is a fine-tuned version of PhanithLIM/whisper-base-aug-22-mar-v1 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1967
• Wer: 85.0444

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: 6.6666666666666675e-06
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• lr_scheduler_warmup_steps: 1000
• num_epochs: 9

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3557	1.0	694	0.2833	94.2047
0.264	2.0	1388	0.2454	91.6031
0.221	3.0	2082	0.2244	89.9050
0.1913	4.0	2776	0.2120	88.7833
0.1688	5.0	3470	0.2056	89.0170
0.1501	6.0	4164	0.1995	86.8360
0.1347	7.0	4858	0.1981	86.5555
0.1221	8.0	5552	0.1976	86.0570
0.1098	9.0	6246	0.1967	85.0444

Framework versions

• Transformers 4.51.1
• Pytorch 2.5.1+cu121
• Datasets 3.5.0
• Tokenizers 0.21.0