CasperMuz/whisper-base-sna-cleaned

Whisper Medium Shona - Cleaned Data

This model is a fine-tuned version of openai/whisper-base on the Cleaned Google WAXAL Shona dataset. It achieves the following results on the evaluation set:

• Loss: 0.4569
• Wer: 40.4367

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 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: 200
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9170	0.2398	200	0.8819	63.4739
0.6427	0.4796	400	0.6353	52.0626
0.5638	0.7194	600	0.5593	47.5029
0.5299	0.9592	800	0.5188	45.2543
0.4432	1.1990	1000	0.4957	43.3099
0.4213	1.4388	1200	0.4811	42.0281
0.4076	1.6787	1400	0.4706	42.2915
0.3994	1.9185	1600	0.4616	41.1645
0.3501	2.1583	1800	0.4584	40.6838
0.3890	2.3981	2000	0.4569	40.4367

Framework versions

• Transformers 5.13.1
• Pytorch 2.12.0+cu130
• Datasets 5.0.0
• Tokenizers 0.22.2