misterkissi/whisper-small-fongbe

wisper-small-fongbe

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: 0.1111
• Wer: 0.2074

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
1.3253	1.2441	500	0.2596	0.5616
0.1357	2.4882	1000	0.1269	0.2757
0.0577	3.7323	1500	0.1112	0.2695
0.0278	4.9763	2000	0.1015	0.2262
0.0143	6.2192	2500	0.1062	0.2645
0.0087	7.4633	3000	0.1104	0.2590
0.0066	8.7073	3500	0.1153	0.2066
0.0049	9.9514	4000	0.1138	0.2364
0.0039	11.1943	4500	0.1100	0.2005
0.0034	12.4384	5000	0.1111	0.2074

Framework versions

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