Whisper-small-speechocean

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

• Loss: 0.6821

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: 0.001
• train_batch_size: 6
• eval_batch_size: 6
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.5214	1.0	417	1.3231
0.6304	2.0	834	0.6180
0.532	3.0	1251	0.5340
0.4258	4.0	1668	0.5058
0.3192	5.0	2085	0.5050
0.288	6.0	2502	0.4952
0.2097	7.0	2919	0.5252
0.1986	8.0	3336	0.5281
0.1185	9.0	3753	0.5534
0.091	10.0	4170	0.5695
0.0548	11.0	4587	0.5935
0.0423	12.0	5004	0.6130
0.031	13.0	5421	0.6170
0.0169	14.0	5838	0.6234
0.0193	15.0	6255	0.6416
0.0125	16.0	6672	0.6478
0.0055	17.0	7089	0.6602
0.0064	18.0	7506	0.6736
0.004	19.0	7923	0.6785
0.003	20.0	8340	0.6821

Framework versions

• PEFT 0.8.0
• Transformers 4.40.0
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1