misterkissi/whisper-small-setswana

wisper-small-setswana

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

• Loss: 0.4474
• Wer: 0.4804

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 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
1.6799	7.4627	500	0.4266	0.3825
0.0114	14.9254	1000	0.3919	0.3980
0.0012	22.3881	1500	0.4005	0.5170
0.0004	29.8507	2000	0.4136	0.4883
0.0002	37.3134	2500	0.4237	0.4773
0.0002	44.7761	3000	0.4318	0.4650
0.0001	52.2388	3500	0.4376	0.4747
0.0001	59.7015	4000	0.4427	0.4822
0.0001	67.1642	4500	0.4459	0.4813
0.0001	74.6269	5000	0.4474	0.4804

Framework versions

• Transformers 4.53.2
• Pytorch 2.6.0+cu124
• Datasets 2.18.0
• Tokenizers 0.21.2