whisper-small-yoruba-07-15

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: 0.4112
• Wer Ortho: 68.9201
• Wer: 61.9420

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.7803	0.0805	250	0.9028	78.2451	72.5162
0.5693	0.1610	500	0.7365	85.1996	82.5191
0.504	0.2415	750	0.6444	73.1001	69.8969
0.463	0.3220	1000	0.5931	78.2923	71.3930
0.4036	0.4024	1250	0.5471	68.4638	62.0921
0.3496	0.4829	1500	0.5171	73.5459	71.9933
0.3346	0.5634	1750	0.4908	67.8109	65.7669
0.34	0.6439	2000	0.4612	70.3336	65.5394
0.3153	0.7244	2250	0.4380	66.8799	60.1118
0.3061	0.8049	2500	0.4228	67.9499	60.2982
0.2877	0.8854	2750	0.4164	67.9735	59.7051
0.2892	0.9659	3000	0.4112	68.9201	61.9420

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1