datdo2717/11

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: 0.4981
• Wer: 16.5519

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: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5019	0.2222	100	0.4649	17.3540
0.4235	0.4444	200	0.4257	16.7932
0.4364	0.6667	300	0.4184	16.5164
0.4106	0.8889	400	0.4043	15.6434
0.2338	1.1111	500	0.4064	15.7286
0.2286	1.3333	600	0.4066	15.9699
0.2185	1.5556	700	0.4058	15.7428
0.212	1.7778	800	0.3999	15.6079
0.2308	2.0	900	0.3991	17.2617
0.0983	2.2222	1000	0.4233	15.9415
0.1183	2.4444	1100	0.4286	16.0409
0.1003	2.6667	1200	0.4304	16.0764
0.1005	2.8889	1300	0.4332	15.7641
0.048	3.1111	1400	0.4636	16.3248
0.0475	3.3333	1500	0.4684	16.2041
0.0516	3.5556	1600	0.4679	16.2254
0.058	3.7778	1700	0.4691	16.2538
0.0457	4.0	1800	0.4693	16.2041
0.028	4.2222	1900	0.4940	16.4880
0.0235	4.4444	2000	0.4981	16.5519

Framework versions

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