misterkissi/whisper-small-umbundu

wisper-small-umbundu

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.5770
• Wer: 0.5981

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: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH 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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.2102	9.4340	500	1.1333	0.5466
0.0459	18.8679	1000	1.3514	0.6631
0.003	28.3019	1500	1.4324	0.6576
0.0025	37.7358	2000	1.4792	0.5691
0.0008	47.1698	2500	1.5070	0.5869
0.0003	56.6038	3000	1.5325	0.6020
0.0003	66.0377	3500	1.5505	0.5989
0.0002	75.4717	4000	1.5636	0.6033
0.0002	84.9057	4500	1.5732	0.5965
0.0002	94.3396	5000	1.5770	0.5981

Framework versions

• Transformers 4.53.2
• Pytorch 2.6.0+cu124
• Datasets 2.14.4
• Tokenizers 0.21.2