leenag/Malasar_Dict

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: 0.0139
• Wer: 7.6014

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
0.042	0.6410	250	0.0392	18.1869
0.023	1.2821	500	0.0318	14.5833
0.0158	1.9231	750	0.0215	10.5293
0.0106	2.5641	1000	0.0175	11.5428
0.0035	3.2051	1250	0.0145	7.5450
0.0027	3.8462	1500	0.0139	9.1779
0.0018	4.4872	1750	0.0144	7.5450
0.0016	5.1282	2000	0.0139	7.6014

Framework versions

• Transformers 4.41.2
• Pytorch 2.1.2+cu121
• Datasets 2.16.0
• Tokenizers 0.19.1