b-brave/whisper-small-ft-balbus-sep28k-multiclass_v3a

whisper-small-ft-balbus-sep28k-multiclass_v3b

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: 0.4681
• Accuracy: 0.655
• Precision: 0.7027
• Recall: 0.6335
• F1: 0.6299
• 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: 3e-06
• train_batch_size: 16
• eval_batch_size: 8
• 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: 0.5
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1	Roc-auc
0.8055	0.2734	100	0.7699	0.3638	0.1213	0.3333	0.1779	None
0.7434	0.5468	200	0.7339	0.4385	0.2928	0.4117	0.3400	None
0.6728	0.8202	300	0.6096	0.5523	0.6058	0.5276	0.5052	None
0.4928	1.0936	400	0.5101	0.6362	0.6856	0.6166	0.6175	None
0.4613	1.3671	500	0.4971	0.6262	0.7070	0.6003	0.5895	None
0.4563	1.6405	600	0.4731	0.6642	0.6937	0.6475	0.6498	None
0.449	1.9139	700	0.4681	0.655	0.7027	0.6335	0.6299	None

Framework versions

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