Whisper-WOLOF-5-hours-ALFFA-dataset

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

• Loss: 0.5488
• Wer: 25.3648
• Cer: 7.4292

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: 32
• 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_ratio: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.312	3.8760	500	0.4949	33.8409	10.1763
0.1965	7.7519	1000	0.4457	29.0445	8.7321
0.0184	11.6279	1500	0.4554	26.4433	7.7258
0.0049	15.5039	2000	0.4903	27.0143	8.0668
0.0024	19.3798	2500	0.4904	25.9865	7.8505
0.0012	23.2558	3000	0.5039	25.4917	7.5317
0.0005	27.1318	3500	0.5155	25.3140	7.5012
0.0003	31.0078	4000	0.5259	25.4790	7.5234
0.0002	34.8837	4500	0.5339	25.3394	7.4680
0.0002	38.7597	5000	0.5395	25.3267	7.4541
0.0002	42.6357	5500	0.5448	25.3267	7.4375
0.0001	46.5116	6000	0.5488	25.3648	7.4292

Framework versions

• Transformers 4.45.2
• Pytorch 2.1.0+cu118
• Datasets 3.0.1
• Tokenizers 0.20.1