aipanjab/whisper-base-pa

Whisper-base

This model is a fine-tuned version of openai/whisper-base on various datasets. It achieves the following results on the evaluation set:

• Loss: 0.2729
• Wer: 38.4713

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: 16
• eval_batch_size: 8
• seed: 42
• 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: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1851	1.2804	1000	0.2017	47.7530
0.1354	2.5608	2000	0.1628	40.4972
0.1025	3.8412	3000	0.1542	38.4193
0.0558	5.1216	4000	0.1625	37.7120
0.0438	6.4020	5000	0.1783	38.2217
0.0311	7.6825	6000	0.1950	38.3855
0.0196	8.9629	7000	0.2156	38.3647
0.0092	10.2433	8000	0.2482	38.6430
0.0059	11.5237	9000	0.2635	38.6898
0.0051	12.8041	10000	0.2729	38.4713

Framework versions

• Transformers 4.46.3
• Pytorch 2.1.0+cu118
• Datasets 3.1.0
• Tokenizers 0.20.3