whisper-tiny-aug-1-april-v2

This model is a fine-tuned version of PhanithLIM/whisper-tiny-aug-1-april-v1 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2714
• Wer: 90.1919

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.615	1.0	320	0.5064	97.4550
0.4639	2.0	640	0.4207	95.7909
0.3884	3.0	960	0.3734	94.5184
0.3392	4.0	1280	0.3457	93.7157
0.3031	5.0	1600	0.3206	92.7565
0.2741	6.0	1920	0.3041	92.1104
0.2499	7.0	2240	0.2951	91.8168
0.23	8.0	2560	0.2843	91.3078
0.2125	9.0	2880	0.2783	90.8575
0.1969	9.9703	3190	0.2714	90.1919

Framework versions

• Transformers 4.50.3
• Pytorch 2.5.1+cu121
• Datasets 3.5.0
• Tokenizers 0.21.0