Kisson/whisper-small-hi-test

Whisper Small Hi Test - Kisson

This model is a fine-tuned version of openai/whisper-small on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4560
• Wer: 48.4890

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: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5
• training_steps: 40
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3267	0.0098	4	0.4986	53.1152
0.3203	0.0196	8	0.4992	52.1544
0.2772	0.0293	12	0.4968	52.0782
0.303	0.0391	16	0.4895	50.7534
0.3085	0.0489	20	0.4802	50.9015
0.3306	0.0587	24	0.4730	50.6222
0.3181	0.0685	28	0.4664	50.1989
0.3726	0.0782	32	0.4598	49.2932
0.3088	0.0880	36	0.4573	48.6202
0.3347	0.0978	40	0.4560	48.4890

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1