b-brave/whisper-small-ft-balbus-sep28k-v1.5

miosipof/whisper-small-ft-balbus-sep28k-v1.6

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

• Loss: 0.1091
• Accuracy: 0.8101
• Precision: 0.8184
• Recall: 0.7261
• F1: 0.7695
• Roc-auc: 0.8006

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: 2e-06
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.5
• training_steps: 1200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1	Roc-auc
0.1683	0.2506	200	0.1682	0.5730	0.7364	0.0341	0.0652	0.5123
0.1494	0.5013	400	0.1446	0.7084	0.6603	0.6838	0.6718	0.7056
0.1212	0.7519	600	0.1236	0.7629	0.6917	0.8245	0.7523	0.7699
0.1088	1.0025	800	0.1107	0.8062	0.8337	0.6945	0.7578	0.7936
0.0955	1.2531	1000	0.1106	0.8081	0.8036	0.7416	0.7713	0.8006
0.0997	1.5038	1200	0.1091	0.8101	0.8184	0.7261	0.7695	0.8006

Framework versions

• Transformers 4.45.2
• Pytorch 2.2.0
• Datasets 3.2.0
• Tokenizers 0.20.3