vrclc/Malasar_medium_MTF

leenag/Malasar_Luke

This model is a fine-tuned version of vasista22/whisper-tamil-medium on the Spoken Bible Corpus: Malasar dataset. It achieves the following results on the evaluation set:

• Loss: 0.5075
• Wer: 48.2010

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
0.018	11.3636	250	0.3526	50.2570
0.0059	22.7273	500	0.4000	49.5146
0.0002	34.0909	750	0.4418	48.3152
0.0001	45.4545	1000	0.4785	48.0868
0.0	56.8182	1250	0.4923	47.8013
0.0	68.1818	1500	0.5008	47.8013
0.0	79.5455	1750	0.5059	48.2010
0.0	90.9091	2000	0.5075	48.2010

Framework versions

• Transformers 4.41.2
• Pytorch 2.0.1+cu117
• Datasets 2.16.0
• Tokenizers 0.19.1