datdo2717/whisper-small-vi-23-11

Whisper Small Ori vi

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: 0.5081
• Wer: 19.1274

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: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• training_steps: 1100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.6002	0.2222	100	0.5593	21.0556
0.4758	0.4444	200	0.5197	20.4011
0.4916	0.6667	300	0.5082	21.7101
0.4612	0.8889	400	0.4973	19.7467
0.2709	1.1111	500	0.4971	20.9500
0.2823	1.3333	600	0.4974	19.4300
0.2819	1.5556	700	0.4943	19.2892
0.2817	1.7778	800	0.4930	19.5496
0.2752	2.0	900	0.4885	19.3878
0.1722	2.2222	1000	0.5053	19.1414
0.1383	2.4444	1100	0.5081	19.1274

Framework versions

• Transformers 4.46.3
• Pytorch 2.4.0
• Datasets 3.1.0
• Tokenizers 0.20.0