tshiluba

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

• Loss: 0.3210
• Wer: 0.1515

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: 3e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• 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: 50
• training_steps: 48000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3596	2.5641	500	0.2753	0.2459
0.0485	5.1282	1000	0.2929	0.1969
0.0172	7.6923	1500	0.2998	0.1808
0.0113	10.2564	2000	0.3029	0.1690
0.009	12.8205	2500	0.3094	0.1690
0.0064	15.3846	3000	0.3127	0.1542
0.0059	17.9487	3500	0.3208	0.1584
0.0055	20.5128	4000	0.3446	0.1586
0.0052	23.0769	4500	0.3210	0.1515
0.0045	25.6410	5000	0.3424	0.1679
0.0055	28.2051	5500	0.3413	0.1569
0.0048	30.7692	6000	0.3370	0.1554

Framework versions

• Transformers 4.46.0
• Pytorch 2.7.0+cu126
• Datasets 3.3.2
• Tokenizers 0.20.3