whisper-tiny-oshiwambo-speech

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

• Loss: 0.1409
• Wer: 44.7619
• Cer: 30.8962
• Word Acc: 64.4444
• Sent Acc: 2.8571
• Precision: 0.6444
• Recall: 0.5524
• F1 Score: 0.5949

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: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 10000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Word Acc	Sent Acc	Precision	Recall	F1 Score
0.0098	117.65	1000	0.0976	37.1429	29.0094	66.6667	8.5714	0.6538	0.6476	0.6507
0.0105	235.29	2000	0.1061	41.9048	33.0189	63.6364	2.8571	0.6238	0.6	0.6117
0.0105	352.94	3000	0.1134	37.1429	26.8868	66.6667	5.7143	0.6667	0.6286	0.6471
0.0091	470.59	4000	0.1222	37.1429	25.7075	66.6667	5.7143	0.6667	0.6286	0.6471
0.0098	588.24	5000	0.1265	40.0	28.3019	65.625	2.8571	0.6562	0.6	0.6269
0.0094	705.88	6000	0.1314	42.8571	30.8962	64.5161	2.8571	0.6452	0.5714	0.6061
0.0093	823.53	7000	0.1366	42.8571	29.2453	64.5161	2.8571	0.6452	0.5714	0.6061
0.0094	941.18	8000	0.1360	45.7143	31.8396	63.3333	0.0	0.6333	0.5429	0.5846
0.01	1058.82	9000	0.1394	44.7619	30.8962	64.4444	2.8571	0.6444	0.5524	0.5949
0.0087	1176.47	10000	0.1409	44.7619	30.8962	64.4444	2.8571	0.6444	0.5524	0.5949

Framework versions

• Transformers 4.30.0.dev0
• Pytorch 2.0.0
• Datasets 2.12.0
• Tokenizers 0.13.3