leenag/Malasar_Luke

This model is a fine-tuned version of openai/whisper-small on the Spoken Bible Corpus: Malasar dataset. It achieves the following results on the evaluation set:

Loss: 0.6211
Wer: 60.1371

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: 16
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 2000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1612	11.3636	250	0.3185	64.0206
0.0114	22.7273	500	0.4682	65.7339
0.0016	34.0909	750	0.5380	59.9657
0.0004	45.4545	1000	0.5761	59.8515
0.0003	56.8182	1250	0.5969	59.6802
0.0002	68.1818	1500	0.6104	60.3655
0.0002	79.5455	1750	0.6181	60.1942
0.0002	90.9091	2000	0.6211	60.1371

Framework versions

Transformers 4.41.2
Pytorch 2.1.2+cu121
Datasets 2.16.0
Tokenizers 0.19.1

Citation

@misc{multistage2024,
      title={Multistage Fine-tuning Strategies for Automatic Speech Recognition in Low-resource Languages}, 
      author={Leena G Pillai, Kavya Manohar, Basil K Raju, Elizabeth Sherly},
      year={2024},
      eprint={2411.04573},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2411.04573}, 
}