whisper-base-yoruba-colab-CV17.0

This model is a fine-tuned version of openai/whisper-base on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 1.0478
• Wer: 0.7074

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.15
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.7762	1.5385	200	0.9767	0.8377
0.5374	3.0769	400	0.8458	0.7728
0.2352	4.6154	600	0.8632	0.7407
0.1038	6.1538	800	0.9157	0.7078
0.0426	7.6923	1000	0.9531	0.7246
0.0201	9.2308	1200	0.9950	0.7219
0.0093	10.7692	1400	1.0347	0.6989
0.0037	12.3077	1600	1.0351	0.7063
0.0013	13.8462	1800	1.0429	0.7069
0.0008	15.3846	2000	1.0478	0.7074

Framework versions

• Transformers 4.42.4
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1