kerker256/whisper-small-tl

Whisper Small Tagalog - Yohan Keraudren

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

• Loss: 0.5932
• Wer: 20.1879

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: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0087	8.4746	1000	0.5078	21.3190
0.0021	16.9492	2000	0.5594	20.6576
0.0005	25.4237	3000	0.5833	20.2646
0.0004	33.8983	4000	0.5932	20.1879

Framework versions

• Transformers 5.18.0
• Pytorch 2.11.0+cu130
• Datasets 5.0.1
• Tokenizers 0.23.2