arjunshajitech/whisper-small-malayalam-v5

Whisper Small Malayalam - Arjun Shaji

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

• Loss: 0.0209
• Wer: 9.5456

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 8000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0433	0.5800	1000	0.0434	27.1379
0.02	1.1601	2000	0.0312	20.3733
0.0169	1.7401	3000	0.0242	15.4975
0.0071	2.3202	4000	0.0217	12.3555
0.0058	2.9002	5000	0.0197	11.0646
0.0022	3.4803	6000	0.0202	10.0881
0.0008	4.0603	7000	0.0204	9.7006
0.0005	4.6404	8000	0.0209	9.5456

Framework versions

• Transformers 4.41.0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1