PhanithLIM/whisper-tiny-aug-14-april-lightning-v1

whisper-tiny-aug-14-april-liggtning-v1

This model is a fine-tuned version of openai/whisper-tiny on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3037
• Wer: 88.1560

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

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.5827	1.0	148	1.4064	107.6928
1.3101	2.0	296	1.2578	105.2291
1.1558	3.0	444	1.0699	103.2401
0.8778	4.0	592	0.6933	101.4244
0.5971	5.0	740	0.5113	99.2814
0.4665	6.0	888	0.4283	95.5088
0.3941	7.0	1036	0.3780	93.6161
0.3449	8.0	1184	0.3437	93.5327
0.308	9.0	1332	0.3190	89.7729
0.2817	9.9356	1470	0.3037	88.1560

Framework versions

• Transformers 4.51.3
• Pytorch 2.2.1+cu121
• Datasets 3.5.0
• Tokenizers 0.21.1