marlhex/whisper-small-dv

Whisper Small - DV - Marlhex

This model is a fine-tuned version of openai/whisper-small on the Common Voice 13 dataset.

Model description

Whisper fine tunned for dv language (Maldivan language, Divehi) from Maldives

Intended uses & limitations

part of the AI portfolio to show to companies some of the work I've done in the Audio pilar.

Training and evaluation data

WER normalized.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 5
• training_steps: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
No log	0.0326	10	1.9577	100.0	100.0

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0

Next Steps

• Looking forward to training more languages that require more GB of storage, but my setup is limited.