whisper-small-ft-balbus-sep28k-multiclass_v3

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: 0.5221
• Accuracy: 0.6704
• Precision: 0.6905
• Recall: 0.6573
• F1: 0.6614
• Roc-auc: None

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: 12
• eval_batch_size: 6
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 48
• 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
0.7619	0.2051	100	0.7516	0.4035	0.4054	0.3432	0.2112	None
0.6854	0.4103	200	0.6178	0.5138	0.6335	0.4817	0.4125	None
0.5066	0.6154	300	0.5008	0.6231	0.6986	0.5986	0.5812	None
0.4635	0.8205	400	0.4840	0.6331	0.6968	0.6098	0.5959	None
0.4525	1.0256	500	0.5023	0.6469	0.6885	0.6255	0.6287	None
0.4164	1.2308	600	0.5023	0.6719	0.6784	0.6697	0.6730	None
0.3798	1.4359	700	0.4758	0.6412	0.6931	0.6192	0.6161	None
0.4174	1.6410	800	0.4708	0.66	0.6998	0.6395	0.6410	None
0.4212	1.8462	900	0.4717	0.6796	0.6951	0.6696	0.6758	None
0.3413	2.0513	1000	0.5100	0.6769	0.6904	0.6666	0.6699	None
0.2757	2.2564	1100	0.5297	0.6735	0.6892	0.6630	0.6682	None
0.2894	2.4615	1200	0.5221	0.6704	0.6905	0.6573	0.6614	None

Framework versions

• Transformers 4.45.2
• Pytorch 2.2.0
• Datasets 3.6.0
• Tokenizers 0.20.3