whisper-og-audio-abuse-feature

This model is a fine-tuned version of openai/whisper-medium on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4737
• Accuracy: 0.8922
• Macro Precision: 0.8730
• Macro Recall: 0.8547
• Macro F1-score: 0.8631

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.01
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Macro Precision	Macro Recall	Macro F1-score
0.4515	0.4367	50	0.4025	0.8069	0.8192	0.8136	0.8066
0.3406	0.8734	100	0.3144	0.8807	0.8875	0.8760	0.8787
0.2698	1.3100	150	0.3496	0.8487	0.8667	0.8408	0.8441
0.2478	1.7467	200	0.2916	0.8942	0.8938	0.8935	0.8937
0.1959	2.1834	250	0.3660	0.8721	0.8795	0.8671	0.8698
0.1286	2.6201	300	0.3699	0.8881	0.8902	0.8853	0.8869
0.1314	3.0568	350	0.3474	0.8905	0.8898	0.8906	0.8901
0.0632	3.4934	400	0.4238	0.8844	0.8843	0.8831	0.8836
0.0761	3.9301	450	0.3957	0.8905	0.8927	0.8878	0.8894
0.0344	4.3668	500	0.4951	0.8918	0.8932	0.8895	0.8908
0.0318	4.8035	550	0.5091	0.8967	0.8969	0.8953	0.8960

Framework versions

• Transformers 4.41.2
• Pytorch 2.1.2
• Datasets 2.19.2
• Tokenizers 0.19.1